diff --git "a/checkpoint-53500/trainer_state.json" "b/checkpoint-53500/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-53500/trainer_state.json"
@@ -0,0 +1,32116 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.982876839257255,
+  "global_step": 53500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6999999999999996e-05,
+      "loss": 2.1657,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.6999999999999996e-05,
+      "loss": 2.0891,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.699999999999999e-05,
+      "loss": 1.8828,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000117,
+      "loss": 1.5145,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000147,
+      "loss": 1.1219,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00017699999999999997,
+      "loss": 1.05,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00020699999999999996,
+      "loss": 0.8439,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000237,
+      "loss": 0.7854,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000267,
+      "loss": 0.8191,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029699999999999996,
+      "loss": 0.8437,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999498718948424,
+      "loss": 0.8548,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002998941740002228,
+      "loss": 0.7763,
+      "step": 120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002998384761056032,
+      "loss": 0.751,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002997827782109836,
+      "loss": 0.8973,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029972708031636403,
+      "loss": 0.7949,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029967138242174444,
+      "loss": 0.7542,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029961568452712485,
+      "loss": 0.7928,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029955998663250527,
+      "loss": 1.0169,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002995042887378857,
+      "loss": 0.8765,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002994485908432661,
+      "loss": 0.8155,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002993928929486465,
+      "loss": 0.7217,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002993371950540269,
+      "loss": 0.8845,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002992814971594073,
+      "loss": 0.8273,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002992257992647878,
+      "loss": 0.7669,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002991701013701682,
+      "loss": 0.8067,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002991144034755486,
+      "loss": 0.7247,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299058705580929,
+      "loss": 0.7256,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029900300768630943,
+      "loss": 0.8089,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029894730979168984,
+      "loss": 0.714,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029889161189707025,
+      "loss": 0.852,
+      "step": 300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002988359140024507,
+      "loss": 0.8713,
+      "step": 310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029878021610783113,
+      "loss": 0.9165,
+      "step": 320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029872451821321154,
+      "loss": 0.8829,
+      "step": 330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029866882031859195,
+      "loss": 0.8993,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029861312242397236,
+      "loss": 0.7899,
+      "step": 350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002985574245293528,
+      "loss": 0.8127,
+      "step": 360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002985017266347332,
+      "loss": 0.7791,
+      "step": 370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002984460287401136,
+      "loss": 0.801,
+      "step": 380
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000298390330845494,
+      "loss": 0.9644,
+      "step": 390
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002983346329508744,
+      "loss": 0.8106,
+      "step": 400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029827893505625483,
+      "loss": 0.7146,
+      "step": 410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029822323716163524,
+      "loss": 0.809,
+      "step": 420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029816753926701565,
+      "loss": 0.7411,
+      "step": 430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002981118413723961,
+      "loss": 0.8484,
+      "step": 440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029805614347777653,
+      "loss": 0.8111,
+      "step": 450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029800044558315694,
+      "loss": 0.8426,
+      "step": 460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029794474768853735,
+      "loss": 0.8323,
+      "step": 470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029788904979391776,
+      "loss": 0.8651,
+      "step": 480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002978333518992982,
+      "loss": 0.7317,
+      "step": 490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029777765400467864,
+      "loss": 0.8968,
+      "step": 500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029772195611005905,
+      "loss": 0.8402,
+      "step": 510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029766625821543946,
+      "loss": 0.8242,
+      "step": 520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029761056032081987,
+      "loss": 0.8364,
+      "step": 530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002975548624262003,
+      "loss": 0.6825,
+      "step": 540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002974991645315807,
+      "loss": 0.7819,
+      "step": 550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002974434666369611,
+      "loss": 0.7209,
+      "step": 560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002973877687423415,
+      "loss": 0.6985,
+      "step": 570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029733207084772193,
+      "loss": 0.8092,
+      "step": 580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029727637295310234,
+      "loss": 0.7812,
+      "step": 590
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029722067505848275,
+      "loss": 0.8383,
+      "step": 600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029716497716386316,
+      "loss": 0.676,
+      "step": 610
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029710927926924357,
+      "loss": 0.8265,
+      "step": 620
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029705358137462404,
+      "loss": 0.8141,
+      "step": 630
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029699788348000445,
+      "loss": 0.9589,
+      "step": 640
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029694218558538486,
+      "loss": 0.7259,
+      "step": 650
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029688648769076527,
+      "loss": 0.9564,
+      "step": 660
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002968307897961457,
+      "loss": 0.7736,
+      "step": 670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002967750919015261,
+      "loss": 0.8092,
+      "step": 680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002967193940069065,
+      "loss": 0.6447,
+      "step": 690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002966636961122869,
+      "loss": 0.797,
+      "step": 700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002966079982176673,
+      "loss": 0.8556,
+      "step": 710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029655230032304774,
+      "loss": 0.7671,
+      "step": 720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002964966024284282,
+      "loss": 0.7346,
+      "step": 730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002964409045338086,
+      "loss": 0.8341,
+      "step": 740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000296385206639189,
+      "loss": 0.7651,
+      "step": 750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029632950874456944,
+      "loss": 0.6583,
+      "step": 760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029627381084994985,
+      "loss": 0.9385,
+      "step": 770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029621811295533026,
+      "loss": 0.7709,
+      "step": 780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029616241506071067,
+      "loss": 0.7106,
+      "step": 790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002961067171660911,
+      "loss": 0.8967,
+      "step": 800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002960510192714715,
+      "loss": 0.7392,
+      "step": 810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002959953213768519,
+      "loss": 0.7959,
+      "step": 820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029593962348223237,
+      "loss": 0.8163,
+      "step": 830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002958839255876128,
+      "loss": 0.886,
+      "step": 840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002958282276929932,
+      "loss": 0.7094,
+      "step": 850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002957725297983736,
+      "loss": 0.7543,
+      "step": 860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000295716831903754,
+      "loss": 0.7067,
+      "step": 870
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002956611340091344,
+      "loss": 0.7805,
+      "step": 880
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029560543611451484,
+      "loss": 0.7247,
+      "step": 890
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029554973821989525,
+      "loss": 0.748,
+      "step": 900
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029549404032527566,
+      "loss": 0.8429,
+      "step": 910
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029543834243065607,
+      "loss": 0.7323,
+      "step": 920
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002953826445360365,
+      "loss": 0.6964,
+      "step": 930
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002953269466414169,
+      "loss": 0.7016,
+      "step": 940
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002952712487467973,
+      "loss": 0.9425,
+      "step": 950
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029521555085217777,
+      "loss": 0.7805,
+      "step": 960
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002951598529575582,
+      "loss": 0.7907,
+      "step": 970
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002951041550629386,
+      "loss": 0.6496,
+      "step": 980
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000295048457168319,
+      "loss": 0.8093,
+      "step": 990
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002949927592736994,
+      "loss": 0.7786,
+      "step": 1000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002949370613790798,
+      "loss": 0.8045,
+      "step": 1010
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002948813634844603,
+      "loss": 0.7455,
+      "step": 1020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002948256655898407,
+      "loss": 0.8247,
+      "step": 1030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002947699676952211,
+      "loss": 0.732,
+      "step": 1040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002947142698006015,
+      "loss": 0.7978,
+      "step": 1050
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029465857190598193,
+      "loss": 0.8016,
+      "step": 1060
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029460287401136234,
+      "loss": 0.7567,
+      "step": 1070
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029454717611674276,
+      "loss": 0.5406,
+      "step": 1080
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029449147822212317,
+      "loss": 0.9177,
+      "step": 1090
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002944357803275036,
+      "loss": 0.7351,
+      "step": 1100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000294380082432884,
+      "loss": 0.7508,
+      "step": 1110
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002943243845382644,
+      "loss": 0.7642,
+      "step": 1120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002942686866436448,
+      "loss": 0.7415,
+      "step": 1130
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002942129887490252,
+      "loss": 0.8158,
+      "step": 1140
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002941572908544057,
+      "loss": 0.8524,
+      "step": 1150
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002941015929597861,
+      "loss": 0.741,
+      "step": 1160
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002940458950651665,
+      "loss": 0.6836,
+      "step": 1170
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002939901971705469,
+      "loss": 0.6417,
+      "step": 1180
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029393449927592733,
+      "loss": 0.8078,
+      "step": 1190
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029387880138130774,
+      "loss": 0.6682,
+      "step": 1200
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002938231034866882,
+      "loss": 0.8078,
+      "step": 1210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002937674055920686,
+      "loss": 0.7706,
+      "step": 1220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029371170769744903,
+      "loss": 0.7169,
+      "step": 1230
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029365600980282944,
+      "loss": 0.8704,
+      "step": 1240
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029360031190820985,
+      "loss": 0.8781,
+      "step": 1250
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029354461401359026,
+      "loss": 0.7668,
+      "step": 1260
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002934889161189707,
+      "loss": 0.9439,
+      "step": 1270
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002934332182243511,
+      "loss": 0.7886,
+      "step": 1280
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002933775203297315,
+      "loss": 0.6752,
+      "step": 1290
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002933218224351119,
+      "loss": 0.9126,
+      "step": 1300
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002932661245404923,
+      "loss": 0.7678,
+      "step": 1310
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029321042664587273,
+      "loss": 0.7099,
+      "step": 1320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029315472875125314,
+      "loss": 0.7767,
+      "step": 1330
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002930990308566336,
+      "loss": 0.8892,
+      "step": 1340
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000293043332962014,
+      "loss": 0.7937,
+      "step": 1350
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029298763506739443,
+      "loss": 0.6858,
+      "step": 1360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029293193717277484,
+      "loss": 0.7289,
+      "step": 1370
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029287623927815525,
+      "loss": 0.747,
+      "step": 1380
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029282054138353566,
+      "loss": 0.7755,
+      "step": 1390
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002927648434889161,
+      "loss": 0.8012,
+      "step": 1400
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029270914559429654,
+      "loss": 0.7436,
+      "step": 1410
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029265344769967695,
+      "loss": 0.7274,
+      "step": 1420
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029259774980505736,
+      "loss": 0.7739,
+      "step": 1430
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002925420519104378,
+      "loss": 0.6398,
+      "step": 1440
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002924863540158182,
+      "loss": 0.8671,
+      "step": 1450
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002924306561211986,
+      "loss": 0.7438,
+      "step": 1460
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000292374958226579,
+      "loss": 0.7123,
+      "step": 1470
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002923192603319594,
+      "loss": 0.7453,
+      "step": 1480
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029226356243733983,
+      "loss": 0.7307,
+      "step": 1490
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029220786454272024,
+      "loss": 0.8455,
+      "step": 1500
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029215216664810065,
+      "loss": 0.7856,
+      "step": 1510
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029209646875348106,
+      "loss": 0.8438,
+      "step": 1520
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002920407708588615,
+      "loss": 0.7787,
+      "step": 1530
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029198507296424194,
+      "loss": 0.7922,
+      "step": 1540
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029192937506962235,
+      "loss": 0.7894,
+      "step": 1550
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029187367717500276,
+      "loss": 0.6718,
+      "step": 1560
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029181797928038317,
+      "loss": 0.7892,
+      "step": 1570
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002917622813857636,
+      "loss": 1.0537,
+      "step": 1580
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000291706583491144,
+      "loss": 0.823,
+      "step": 1590
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029165088559652446,
+      "loss": 0.6297,
+      "step": 1600
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029159518770190487,
+      "loss": 0.6563,
+      "step": 1610
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002915394898072853,
+      "loss": 0.6895,
+      "step": 1620
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002914837919126657,
+      "loss": 0.7973,
+      "step": 1630
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002914280940180461,
+      "loss": 0.7643,
+      "step": 1640
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002913723961234265,
+      "loss": 0.8495,
+      "step": 1650
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002913166982288069,
+      "loss": 0.9305,
+      "step": 1660
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029126100033418734,
+      "loss": 0.768,
+      "step": 1670
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029120530243956775,
+      "loss": 0.7529,
+      "step": 1680
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029114960454494816,
+      "loss": 0.8489,
+      "step": 1690
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029109390665032857,
+      "loss": 0.6501,
+      "step": 1700
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000291038208755709,
+      "loss": 0.7523,
+      "step": 1710
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002909825108610894,
+      "loss": 0.7792,
+      "step": 1720
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029092681296646986,
+      "loss": 0.7997,
+      "step": 1730
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029087111507185027,
+      "loss": 0.8561,
+      "step": 1740
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002908154171772307,
+      "loss": 0.8504,
+      "step": 1750
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002907597192826111,
+      "loss": 0.7216,
+      "step": 1760
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002907040213879915,
+      "loss": 0.7572,
+      "step": 1770
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002906483234933719,
+      "loss": 0.6785,
+      "step": 1780
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002905926255987523,
+      "loss": 0.8522,
+      "step": 1790
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002905369277041328,
+      "loss": 0.622,
+      "step": 1800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002904812298095132,
+      "loss": 0.7532,
+      "step": 1810
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002904255319148936,
+      "loss": 0.7477,
+      "step": 1820
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000290369834020274,
+      "loss": 0.737,
+      "step": 1830
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029031413612565444,
+      "loss": 0.8347,
+      "step": 1840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029025843823103485,
+      "loss": 0.6236,
+      "step": 1850
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029020274033641526,
+      "loss": 0.894,
+      "step": 1860
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029014704244179567,
+      "loss": 0.829,
+      "step": 1870
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002900913445471761,
+      "loss": 0.655,
+      "step": 1880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002900356466525565,
+      "loss": 0.8142,
+      "step": 1890
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002899799487579369,
+      "loss": 0.8483,
+      "step": 1900
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002899242508633173,
+      "loss": 0.8201,
+      "step": 1910
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002898685529686977,
+      "loss": 0.6601,
+      "step": 1920
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002898128550740782,
+      "loss": 0.8202,
+      "step": 1930
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002897571571794586,
+      "loss": 0.6678,
+      "step": 1940
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000289701459284839,
+      "loss": 0.7731,
+      "step": 1950
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002896457613902194,
+      "loss": 0.7277,
+      "step": 1960
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00028959563328506184,
+      "loss": 0.6507,
+      "step": 1970
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00028953993539044225,
+      "loss": 0.7655,
+      "step": 1980
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00028948423749582266,
+      "loss": 0.7873,
+      "step": 1990
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00028942853960120307,
+      "loss": 0.9066,
+      "step": 2000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002893728417065835,
+      "loss": 0.8656,
+      "step": 2010
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002893171438119639,
+      "loss": 0.7119,
+      "step": 2020
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002892614459173443,
+      "loss": 0.76,
+      "step": 2030
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002892057480227247,
+      "loss": 0.659,
+      "step": 2040
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002891500501281051,
+      "loss": 0.6439,
+      "step": 2050
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00028909435223348554,
+      "loss": 0.6981,
+      "step": 2060
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00028903865433886595,
+      "loss": 0.7457,
+      "step": 2070
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00028898295644424636,
+      "loss": 0.7324,
+      "step": 2080
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00028892725854962677,
+      "loss": 0.8091,
+      "step": 2090
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00028887156065500724,
+      "loss": 0.7322,
+      "step": 2100
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00028881586276038765,
+      "loss": 0.7112,
+      "step": 2110
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00028876016486576806,
+      "loss": 0.8487,
+      "step": 2120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00028870446697114847,
+      "loss": 0.6818,
+      "step": 2130
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002886487690765289,
+      "loss": 0.7988,
+      "step": 2140
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002885930711819093,
+      "loss": 0.7921,
+      "step": 2150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00028853737328728976,
+      "loss": 0.7544,
+      "step": 2160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00028848167539267017,
+      "loss": 0.7076,
+      "step": 2170
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002884259774980506,
+      "loss": 0.7264,
+      "step": 2180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000288370279603431,
+      "loss": 0.8993,
+      "step": 2190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002883145817088114,
+      "loss": 0.7293,
+      "step": 2200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002882588838141918,
+      "loss": 0.7917,
+      "step": 2210
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002882031859195722,
+      "loss": 0.6635,
+      "step": 2220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00028814748802495263,
+      "loss": 0.7215,
+      "step": 2230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00028809179013033305,
+      "loss": 0.7617,
+      "step": 2240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00028803609223571346,
+      "loss": 0.8387,
+      "step": 2250
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00028798039434109387,
+      "loss": 0.665,
+      "step": 2260
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002879246964464743,
+      "loss": 0.8129,
+      "step": 2270
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002878689985518547,
+      "loss": 0.5688,
+      "step": 2280
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00028781330065723516,
+      "loss": 0.7847,
+      "step": 2290
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00028775760276261557,
+      "loss": 0.6735,
+      "step": 2300
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000287701904867996,
+      "loss": 0.5909,
+      "step": 2310
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002876462069733764,
+      "loss": 0.788,
+      "step": 2320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002875905090787568,
+      "loss": 0.783,
+      "step": 2330
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002875348111841372,
+      "loss": 0.7952,
+      "step": 2340
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002874791132895176,
+      "loss": 0.6714,
+      "step": 2350
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00028742341539489803,
+      "loss": 0.7481,
+      "step": 2360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00028736771750027844,
+      "loss": 0.6156,
+      "step": 2370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002873120196056589,
+      "loss": 0.7312,
+      "step": 2380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002872563217110393,
+      "loss": 0.7561,
+      "step": 2390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00028720062381641973,
+      "loss": 0.7927,
+      "step": 2400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00028714492592180014,
+      "loss": 0.7223,
+      "step": 2410
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00028708922802718055,
+      "loss": 0.7335,
+      "step": 2420
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00028703353013256097,
+      "loss": 0.6564,
+      "step": 2430
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002869778322379414,
+      "loss": 0.6481,
+      "step": 2440
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002869221343433218,
+      "loss": 0.9089,
+      "step": 2450
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002868664364487022,
+      "loss": 0.7339,
+      "step": 2460
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002868107385540826,
+      "loss": 0.6527,
+      "step": 2470
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000286755040659463,
+      "loss": 0.6925,
+      "step": 2480
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002866993427648435,
+      "loss": 0.691,
+      "step": 2490
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002866436448702239,
+      "loss": 0.6942,
+      "step": 2500
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002865879469756043,
+      "loss": 0.8531,
+      "step": 2510
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002865322490809847,
+      "loss": 0.9173,
+      "step": 2520
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00028647655118636513,
+      "loss": 0.7681,
+      "step": 2530
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00028642085329174554,
+      "loss": 0.7211,
+      "step": 2540
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00028636515539712595,
+      "loss": 0.6427,
+      "step": 2550
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00028630945750250636,
+      "loss": 0.7184,
+      "step": 2560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002862537596078868,
+      "loss": 0.6873,
+      "step": 2570
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002861980617132672,
+      "loss": 0.6968,
+      "step": 2580
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002861423638186476,
+      "loss": 0.788,
+      "step": 2590
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000286086665924028,
+      "loss": 0.7222,
+      "step": 2600
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002860309680294084,
+      "loss": 0.8198,
+      "step": 2610
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002859752701347889,
+      "loss": 0.5913,
+      "step": 2620
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002859195722401693,
+      "loss": 0.686,
+      "step": 2630
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002858638743455497,
+      "loss": 0.7338,
+      "step": 2640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002858081764509301,
+      "loss": 0.7901,
+      "step": 2650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00028575247855631053,
+      "loss": 0.6926,
+      "step": 2660
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00028569678066169094,
+      "loss": 0.7756,
+      "step": 2670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002856410827670714,
+      "loss": 0.7851,
+      "step": 2680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002855853848724518,
+      "loss": 0.9493,
+      "step": 2690
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00028552968697783223,
+      "loss": 0.8183,
+      "step": 2700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00028547398908321264,
+      "loss": 0.769,
+      "step": 2710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00028541829118859305,
+      "loss": 0.7693,
+      "step": 2720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00028536259329397346,
+      "loss": 0.649,
+      "step": 2730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002853068953993539,
+      "loss": 0.6816,
+      "step": 2740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002852511975047343,
+      "loss": 0.8367,
+      "step": 2750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002851954996101147,
+      "loss": 0.6146,
+      "step": 2760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002851398017154951,
+      "loss": 0.7475,
+      "step": 2770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002850841038208755,
+      "loss": 0.7175,
+      "step": 2780
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00028502840592625593,
+      "loss": 0.642,
+      "step": 2790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00028497270803163634,
+      "loss": 0.7133,
+      "step": 2800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002849170101370168,
+      "loss": 0.5929,
+      "step": 2810
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002848613122423972,
+      "loss": 0.7594,
+      "step": 2820
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00028480561434777763,
+      "loss": 0.7461,
+      "step": 2830
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00028474991645315804,
+      "loss": 0.7089,
+      "step": 2840
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00028469421855853845,
+      "loss": 0.805,
+      "step": 2850
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00028463852066391886,
+      "loss": 0.8134,
+      "step": 2860
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00028458282276929927,
+      "loss": 0.7319,
+      "step": 2870
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00028452712487467974,
+      "loss": 0.7903,
+      "step": 2880
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00028447142698006015,
+      "loss": 0.7304,
+      "step": 2890
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00028441572908544056,
+      "loss": 0.7112,
+      "step": 2900
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00028436003119082097,
+      "loss": 0.7919,
+      "step": 2910
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002843043332962014,
+      "loss": 0.6882,
+      "step": 2920
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002842486354015818,
+      "loss": 0.7139,
+      "step": 2930
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002841929375069622,
+      "loss": 0.7818,
+      "step": 2940
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002841372396123426,
+      "loss": 0.6862,
+      "step": 2950
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000284081541717723,
+      "loss": 0.8077,
+      "step": 2960
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00028402584382310344,
+      "loss": 0.8587,
+      "step": 2970
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00028397014592848385,
+      "loss": 0.6373,
+      "step": 2980
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00028391444803386426,
+      "loss": 0.7463,
+      "step": 2990
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002838587501392447,
+      "loss": 0.7491,
+      "step": 3000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00028380305224462514,
+      "loss": 0.8382,
+      "step": 3010
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00028374735435000555,
+      "loss": 0.7828,
+      "step": 3020
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00028369165645538596,
+      "loss": 0.8557,
+      "step": 3030
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00028363595856076637,
+      "loss": 0.723,
+      "step": 3040
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002835802606661468,
+      "loss": 0.8736,
+      "step": 3050
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002835245627715272,
+      "loss": 0.6935,
+      "step": 3060
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00028346886487690766,
+      "loss": 0.67,
+      "step": 3070
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00028341316698228807,
+      "loss": 0.8309,
+      "step": 3080
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002833574690876685,
+      "loss": 0.7515,
+      "step": 3090
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002833017711930489,
+      "loss": 0.7052,
+      "step": 3100
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002832460732984293,
+      "loss": 0.8169,
+      "step": 3110
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002831903754038097,
+      "loss": 0.6467,
+      "step": 3120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002831346775091901,
+      "loss": 0.6708,
+      "step": 3130
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028307897961457054,
+      "loss": 0.7488,
+      "step": 3140
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028302328171995095,
+      "loss": 0.6892,
+      "step": 3150
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028296758382533136,
+      "loss": 0.7342,
+      "step": 3160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028291188593071177,
+      "loss": 0.6201,
+      "step": 3170
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002828561880360922,
+      "loss": 0.881,
+      "step": 3180
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002828004901414726,
+      "loss": 0.7862,
+      "step": 3190
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028274479224685306,
+      "loss": 0.7323,
+      "step": 3200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028268909435223347,
+      "loss": 0.708,
+      "step": 3210
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002826333964576139,
+      "loss": 0.7473,
+      "step": 3220
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002825776985629943,
+      "loss": 0.7208,
+      "step": 3230
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002825220006683747,
+      "loss": 0.8586,
+      "step": 3240
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002824663027737551,
+      "loss": 0.6439,
+      "step": 3250
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002824106048791356,
+      "loss": 0.7187,
+      "step": 3260
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000282354906984516,
+      "loss": 0.6736,
+      "step": 3270
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002822992090898964,
+      "loss": 0.812,
+      "step": 3280
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002822435111952768,
+      "loss": 0.7541,
+      "step": 3290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002821878133006572,
+      "loss": 0.7432,
+      "step": 3300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028213211540603763,
+      "loss": 0.7788,
+      "step": 3310
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028207641751141804,
+      "loss": 0.8141,
+      "step": 3320
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028202071961679846,
+      "loss": 0.7325,
+      "step": 3330
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028196502172217887,
+      "loss": 0.7932,
+      "step": 3340
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002819093238275593,
+      "loss": 0.8532,
+      "step": 3350
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002818536259329397,
+      "loss": 0.7628,
+      "step": 3360
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002817979280383201,
+      "loss": 0.7139,
+      "step": 3370
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002817422301437005,
+      "loss": 0.7749,
+      "step": 3380
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000281686532249081,
+      "loss": 0.6444,
+      "step": 3390
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002816308343544614,
+      "loss": 0.6748,
+      "step": 3400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002815751364598418,
+      "loss": 0.7428,
+      "step": 3410
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002815194385652222,
+      "loss": 0.7912,
+      "step": 3420
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002814637406706026,
+      "loss": 0.8587,
+      "step": 3430
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028140804277598303,
+      "loss": 0.7408,
+      "step": 3440
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028135234488136344,
+      "loss": 0.7386,
+      "step": 3450
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002812966469867439,
+      "loss": 0.798,
+      "step": 3460
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002812409490921243,
+      "loss": 0.6816,
+      "step": 3470
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028118525119750473,
+      "loss": 0.8592,
+      "step": 3480
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028112955330288514,
+      "loss": 0.5379,
+      "step": 3490
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028107385540826555,
+      "loss": 0.721,
+      "step": 3500
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028101815751364596,
+      "loss": 0.7638,
+      "step": 3510
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002809624596190264,
+      "loss": 0.7621,
+      "step": 3520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002809067617244068,
+      "loss": 0.7362,
+      "step": 3530
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002808510638297872,
+      "loss": 0.7464,
+      "step": 3540
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002807953659351676,
+      "loss": 0.7857,
+      "step": 3550
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000280739668040548,
+      "loss": 0.6152,
+      "step": 3560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028068397014592843,
+      "loss": 0.6198,
+      "step": 3570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028062827225130884,
+      "loss": 0.7537,
+      "step": 3580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002805725743566893,
+      "loss": 0.8577,
+      "step": 3590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002805168764620697,
+      "loss": 0.649,
+      "step": 3600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028046117856745013,
+      "loss": 0.6553,
+      "step": 3610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028040548067283054,
+      "loss": 0.7438,
+      "step": 3620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028034978277821095,
+      "loss": 0.7611,
+      "step": 3630
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028029408488359136,
+      "loss": 0.7373,
+      "step": 3640
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028023838698897183,
+      "loss": 0.7569,
+      "step": 3650
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028018268909435224,
+      "loss": 0.8564,
+      "step": 3660
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028012699119973265,
+      "loss": 0.7556,
+      "step": 3670
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028007129330511306,
+      "loss": 0.6129,
+      "step": 3680
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002800155954104935,
+      "loss": 0.7645,
+      "step": 3690
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002799598975158739,
+      "loss": 0.8212,
+      "step": 3700
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002799041996212543,
+      "loss": 0.652,
+      "step": 3710
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002798485017266347,
+      "loss": 0.783,
+      "step": 3720
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002797928038320151,
+      "loss": 0.7443,
+      "step": 3730
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00027973710593739553,
+      "loss": 0.7401,
+      "step": 3740
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00027968140804277594,
+      "loss": 0.6446,
+      "step": 3750
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00027962571014815635,
+      "loss": 0.7434,
+      "step": 3760
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00027957001225353676,
+      "loss": 0.6295,
+      "step": 3770
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00027951431435891723,
+      "loss": 0.6988,
+      "step": 3780
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00027945861646429764,
+      "loss": 0.741,
+      "step": 3790
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00027940291856967805,
+      "loss": 0.8125,
+      "step": 3800
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00027934722067505846,
+      "loss": 0.8193,
+      "step": 3810
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00027929152278043887,
+      "loss": 0.734,
+      "step": 3820
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002792358248858193,
+      "loss": 0.7431,
+      "step": 3830
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002791801269911997,
+      "loss": 0.7123,
+      "step": 3840
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00027912442909658016,
+      "loss": 0.7364,
+      "step": 3850
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00027906873120196057,
+      "loss": 0.6552,
+      "step": 3860
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000279013033307341,
+      "loss": 0.5693,
+      "step": 3870
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002789573354127214,
+      "loss": 0.7448,
+      "step": 3880
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002789016375181018,
+      "loss": 0.846,
+      "step": 3890
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002788459396234822,
+      "loss": 0.6561,
+      "step": 3900
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002787902417288626,
+      "loss": 0.6593,
+      "step": 3910
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00027873454383424304,
+      "loss": 0.6321,
+      "step": 3920
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00027867884593962345,
+      "loss": 0.7445,
+      "step": 3930
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00027862314804500386,
+      "loss": 0.7248,
+      "step": 3940
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00027856745015038427,
+      "loss": 0.7501,
+      "step": 3950
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002785117522557647,
+      "loss": 0.7191,
+      "step": 3960
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002784560543611451,
+      "loss": 0.7647,
+      "step": 3970
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00027840035646652556,
+      "loss": 0.7046,
+      "step": 3980
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00027834465857190597,
+      "loss": 0.7628,
+      "step": 3990
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002782889606772864,
+      "loss": 0.6385,
+      "step": 4000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002782332627826668,
+      "loss": 0.786,
+      "step": 4010
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002781775648880472,
+      "loss": 0.7781,
+      "step": 4020
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002781218669934276,
+      "loss": 0.6636,
+      "step": 4030
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002780661690988081,
+      "loss": 0.8073,
+      "step": 4040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002780104712041885,
+      "loss": 0.6326,
+      "step": 4050
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002779547733095689,
+      "loss": 0.5519,
+      "step": 4060
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002778990754149493,
+      "loss": 0.7921,
+      "step": 4070
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002778433775203297,
+      "loss": 0.7914,
+      "step": 4080
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00027778767962571014,
+      "loss": 0.744,
+      "step": 4090
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00027773198173109055,
+      "loss": 0.7316,
+      "step": 4100
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00027767628383647096,
+      "loss": 0.8279,
+      "step": 4110
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00027762058594185137,
+      "loss": 0.7324,
+      "step": 4120
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002775648880472318,
+      "loss": 0.718,
+      "step": 4130
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002775091901526122,
+      "loss": 0.621,
+      "step": 4140
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002774534922579926,
+      "loss": 0.7068,
+      "step": 4150
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000277397794363373,
+      "loss": 0.6845,
+      "step": 4160
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002773420964687535,
+      "loss": 0.785,
+      "step": 4170
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002772863985741339,
+      "loss": 0.6549,
+      "step": 4180
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002772307006795143,
+      "loss": 0.7149,
+      "step": 4190
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002771750027848947,
+      "loss": 0.687,
+      "step": 4200
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002771193048902751,
+      "loss": 0.6946,
+      "step": 4210
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00027706360699565553,
+      "loss": 0.6804,
+      "step": 4220
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00027700790910103595,
+      "loss": 0.6528,
+      "step": 4230
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002769522112064164,
+      "loss": 0.6694,
+      "step": 4240
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002768965133117968,
+      "loss": 0.7115,
+      "step": 4250
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00027684081541717723,
+      "loss": 0.8429,
+      "step": 4260
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00027678511752255764,
+      "loss": 0.7695,
+      "step": 4270
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00027672941962793806,
+      "loss": 0.6824,
+      "step": 4280
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00027667372173331847,
+      "loss": 0.6731,
+      "step": 4290
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002766180238386989,
+      "loss": 0.922,
+      "step": 4300
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002765623259440793,
+      "loss": 0.8287,
+      "step": 4310
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002765066280494597,
+      "loss": 0.8205,
+      "step": 4320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002764509301548401,
+      "loss": 0.7776,
+      "step": 4330
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002763952322602205,
+      "loss": 0.7258,
+      "step": 4340
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00027633953436560093,
+      "loss": 0.8068,
+      "step": 4350
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00027628383647098134,
+      "loss": 0.7317,
+      "step": 4360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002762281385763618,
+      "loss": 0.7747,
+      "step": 4370
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002761724406817422,
+      "loss": 0.7258,
+      "step": 4380
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00027611674278712263,
+      "loss": 0.689,
+      "step": 4390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00027606104489250304,
+      "loss": 0.7344,
+      "step": 4400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00027600534699788345,
+      "loss": 0.6484,
+      "step": 4410
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00027594964910326387,
+      "loss": 0.6977,
+      "step": 4420
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002758939512086443,
+      "loss": 0.6636,
+      "step": 4430
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00027583825331402474,
+      "loss": 0.75,
+      "step": 4440
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00027578255541940515,
+      "loss": 0.6786,
+      "step": 4450
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027572685752478556,
+      "loss": 0.7433,
+      "step": 4460
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000275671159630166,
+      "loss": 0.6137,
+      "step": 4470
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002756154617355464,
+      "loss": 0.6681,
+      "step": 4480
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002755597638409268,
+      "loss": 0.7152,
+      "step": 4490
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002755040659463072,
+      "loss": 0.769,
+      "step": 4500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002754483680516876,
+      "loss": 0.7291,
+      "step": 4510
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027539267015706803,
+      "loss": 0.7176,
+      "step": 4520
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002753425420519104,
+      "loss": 0.8482,
+      "step": 4530
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027528684415729086,
+      "loss": 0.7211,
+      "step": 4540
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027523114626267127,
+      "loss": 0.653,
+      "step": 4550
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002751754483680517,
+      "loss": 0.7194,
+      "step": 4560
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002751197504734321,
+      "loss": 0.8425,
+      "step": 4570
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002750640525788125,
+      "loss": 0.6682,
+      "step": 4580
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002750083546841929,
+      "loss": 0.7796,
+      "step": 4590
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002749526567895733,
+      "loss": 0.8755,
+      "step": 4600
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027489695889495373,
+      "loss": 0.6759,
+      "step": 4610
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027484126100033414,
+      "loss": 0.8197,
+      "step": 4620
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027478556310571456,
+      "loss": 0.7042,
+      "step": 4630
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027472986521109497,
+      "loss": 0.765,
+      "step": 4640
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002746741673164754,
+      "loss": 0.8929,
+      "step": 4650
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002746184694218558,
+      "loss": 0.614,
+      "step": 4660
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027456277152723625,
+      "loss": 0.7963,
+      "step": 4670
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027450707363261667,
+      "loss": 0.6607,
+      "step": 4680
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002744513757379971,
+      "loss": 0.7462,
+      "step": 4690
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002743956778433775,
+      "loss": 0.8325,
+      "step": 4700
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002743399799487579,
+      "loss": 0.765,
+      "step": 4710
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002742842820541383,
+      "loss": 0.7377,
+      "step": 4720
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002742285841595188,
+      "loss": 0.8078,
+      "step": 4730
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002741728862648992,
+      "loss": 0.7291,
+      "step": 4740
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002741171883702796,
+      "loss": 0.6825,
+      "step": 4750
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027406149047566,
+      "loss": 0.5591,
+      "step": 4760
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002740057925810404,
+      "loss": 0.6499,
+      "step": 4770
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027395009468642083,
+      "loss": 0.74,
+      "step": 4780
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027389439679180124,
+      "loss": 0.66,
+      "step": 4790
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027383869889718165,
+      "loss": 0.8379,
+      "step": 4800
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027378300100256206,
+      "loss": 0.7177,
+      "step": 4810
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002737273031079425,
+      "loss": 0.5488,
+      "step": 4820
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002736716052133229,
+      "loss": 0.7356,
+      "step": 4830
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002736159073187033,
+      "loss": 0.7999,
+      "step": 4840
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002735602094240837,
+      "loss": 0.8105,
+      "step": 4850
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002735045115294642,
+      "loss": 0.6913,
+      "step": 4860
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002734488136348446,
+      "loss": 0.6566,
+      "step": 4870
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000273393115740225,
+      "loss": 0.7586,
+      "step": 4880
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002733374178456054,
+      "loss": 0.6784,
+      "step": 4890
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002732817199509858,
+      "loss": 0.7357,
+      "step": 4900
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027322602205636623,
+      "loss": 0.9303,
+      "step": 4910
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002731703241617467,
+      "loss": 0.737,
+      "step": 4920
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002731146262671271,
+      "loss": 0.7615,
+      "step": 4930
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002730589283725075,
+      "loss": 0.7686,
+      "step": 4940
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027300323047788793,
+      "loss": 0.6899,
+      "step": 4950
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027294753258326834,
+      "loss": 0.6733,
+      "step": 4960
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027289183468864875,
+      "loss": 0.7312,
+      "step": 4970
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027283613679402916,
+      "loss": 0.5332,
+      "step": 4980
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002727804388994096,
+      "loss": 0.764,
+      "step": 4990
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027272474100479,
+      "loss": 0.8746,
+      "step": 5000
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002726690431101704,
+      "loss": 0.7566,
+      "step": 5010
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002726133452155508,
+      "loss": 0.665,
+      "step": 5020
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002725576473209312,
+      "loss": 0.6802,
+      "step": 5030
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027250194942631163,
+      "loss": 0.7794,
+      "step": 5040
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002724462515316921,
+      "loss": 0.7101,
+      "step": 5050
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002723905536370725,
+      "loss": 0.8295,
+      "step": 5060
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002723348557424529,
+      "loss": 0.6303,
+      "step": 5070
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027227915784783333,
+      "loss": 0.7234,
+      "step": 5080
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027222345995321374,
+      "loss": 0.6497,
+      "step": 5090
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027216776205859415,
+      "loss": 0.6319,
+      "step": 5100
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027211206416397456,
+      "loss": 0.8571,
+      "step": 5110
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000272056366269355,
+      "loss": 0.7513,
+      "step": 5120
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027200066837473544,
+      "loss": 0.6934,
+      "step": 5130
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027194497048011585,
+      "loss": 0.739,
+      "step": 5140
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027188927258549626,
+      "loss": 0.7325,
+      "step": 5150
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027183357469087667,
+      "loss": 0.6974,
+      "step": 5160
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002717778767962571,
+      "loss": 0.6632,
+      "step": 5170
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002717221789016375,
+      "loss": 0.6815,
+      "step": 5180
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002716664810070179,
+      "loss": 0.6149,
+      "step": 5190
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002716107831123983,
+      "loss": 0.9114,
+      "step": 5200
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027155508521777873,
+      "loss": 0.6524,
+      "step": 5210
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027149938732315914,
+      "loss": 0.6619,
+      "step": 5220
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027144368942853955,
+      "loss": 0.6642,
+      "step": 5230
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027138799153391996,
+      "loss": 0.6233,
+      "step": 5240
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002713322936393004,
+      "loss": 0.6848,
+      "step": 5250
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027127659574468084,
+      "loss": 0.7088,
+      "step": 5260
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027122089785006125,
+      "loss": 0.6855,
+      "step": 5270
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027116519995544166,
+      "loss": 0.7053,
+      "step": 5280
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027110950206082207,
+      "loss": 0.5642,
+      "step": 5290
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002710538041662025,
+      "loss": 0.6653,
+      "step": 5300
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027099810627158295,
+      "loss": 0.7754,
+      "step": 5310
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027094240837696336,
+      "loss": 0.9231,
+      "step": 5320
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027088671048234377,
+      "loss": 0.6057,
+      "step": 5330
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002708310125877242,
+      "loss": 0.7264,
+      "step": 5340
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002707753146931046,
+      "loss": 0.7683,
+      "step": 5350
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000270719616798485,
+      "loss": 0.8335,
+      "step": 5360
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002706639189038654,
+      "loss": 0.7295,
+      "step": 5370
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002706082210092458,
+      "loss": 0.6963,
+      "step": 5380
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027055252311462624,
+      "loss": 0.6913,
+      "step": 5390
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027049682522000665,
+      "loss": 0.7111,
+      "step": 5400
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027044112732538706,
+      "loss": 0.7202,
+      "step": 5410
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027038542943076747,
+      "loss": 0.7426,
+      "step": 5420
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002703297315361479,
+      "loss": 0.733,
+      "step": 5430
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027027403364152835,
+      "loss": 0.8221,
+      "step": 5440
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027021833574690876,
+      "loss": 0.6752,
+      "step": 5450
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027016263785228917,
+      "loss": 0.7539,
+      "step": 5460
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002701069399576696,
+      "loss": 0.8046,
+      "step": 5470
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027005124206305,
+      "loss": 0.6616,
+      "step": 5480
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002699955441684304,
+      "loss": 0.6685,
+      "step": 5490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002699398462738108,
+      "loss": 0.5998,
+      "step": 5500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002698841483791913,
+      "loss": 0.6017,
+      "step": 5510
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002698284504845717,
+      "loss": 0.7894,
+      "step": 5520
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002697727525899521,
+      "loss": 0.7175,
+      "step": 5530
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002697170546953325,
+      "loss": 0.6036,
+      "step": 5540
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002696613568007129,
+      "loss": 0.8645,
+      "step": 5550
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026960565890609333,
+      "loss": 0.7887,
+      "step": 5560
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026954996101147374,
+      "loss": 0.792,
+      "step": 5570
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026949426311685416,
+      "loss": 0.74,
+      "step": 5580
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026943856522223457,
+      "loss": 0.6056,
+      "step": 5590
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000269382867327615,
+      "loss": 0.7574,
+      "step": 5600
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002693271694329954,
+      "loss": 0.7071,
+      "step": 5610
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002692714715383758,
+      "loss": 0.6793,
+      "step": 5620
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002692157736437562,
+      "loss": 0.5326,
+      "step": 5630
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002691600757491367,
+      "loss": 0.7632,
+      "step": 5640
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002691043778545171,
+      "loss": 0.6762,
+      "step": 5650
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002690486799598975,
+      "loss": 0.7766,
+      "step": 5660
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002689929820652779,
+      "loss": 0.7887,
+      "step": 5670
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002689372841706583,
+      "loss": 0.6496,
+      "step": 5680
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026888158627603873,
+      "loss": 0.7849,
+      "step": 5690
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002688258883814192,
+      "loss": 0.7911,
+      "step": 5700
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002687701904867996,
+      "loss": 0.8263,
+      "step": 5710
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026871449259218,
+      "loss": 0.7121,
+      "step": 5720
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026865879469756043,
+      "loss": 0.76,
+      "step": 5730
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026860309680294084,
+      "loss": 0.9227,
+      "step": 5740
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026854739890832125,
+      "loss": 0.6894,
+      "step": 5750
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026849170101370166,
+      "loss": 0.7586,
+      "step": 5760
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002684360031190821,
+      "loss": 0.7109,
+      "step": 5770
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002683803052244625,
+      "loss": 0.6974,
+      "step": 5780
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002683246073298429,
+      "loss": 0.6584,
+      "step": 5790
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002682689094352233,
+      "loss": 0.8414,
+      "step": 5800
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002682132115406037,
+      "loss": 0.6988,
+      "step": 5810
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026815751364598413,
+      "loss": 0.8674,
+      "step": 5820
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002681018157513646,
+      "loss": 0.726,
+      "step": 5830
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000268046117856745,
+      "loss": 0.7057,
+      "step": 5840
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002679904199621254,
+      "loss": 0.6789,
+      "step": 5850
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026793472206750583,
+      "loss": 0.6601,
+      "step": 5860
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026787902417288624,
+      "loss": 0.812,
+      "step": 5870
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026782332627826665,
+      "loss": 0.6259,
+      "step": 5880
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026776762838364706,
+      "loss": 0.5619,
+      "step": 5890
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026771193048902753,
+      "loss": 0.7351,
+      "step": 5900
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026765623259440794,
+      "loss": 0.557,
+      "step": 5910
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026760053469978835,
+      "loss": 0.717,
+      "step": 5920
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026754483680516876,
+      "loss": 0.72,
+      "step": 5930
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002674891389105492,
+      "loss": 0.752,
+      "step": 5940
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002674334410159296,
+      "loss": 0.7719,
+      "step": 5950
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026737774312131,
+      "loss": 0.7094,
+      "step": 5960
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002673220452266904,
+      "loss": 0.7023,
+      "step": 5970
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002672663473320708,
+      "loss": 0.5916,
+      "step": 5980
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026721064943745123,
+      "loss": 0.6855,
+      "step": 5990
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026715495154283164,
+      "loss": 0.7081,
+      "step": 6000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026709925364821205,
+      "loss": 0.7359,
+      "step": 6010
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026704355575359246,
+      "loss": 0.8214,
+      "step": 6020
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026698785785897293,
+      "loss": 0.5878,
+      "step": 6030
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026693215996435334,
+      "loss": 0.6977,
+      "step": 6040
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026687646206973375,
+      "loss": 0.7332,
+      "step": 6050
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026682076417511416,
+      "loss": 0.7437,
+      "step": 6060
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026676506628049457,
+      "loss": 0.7216,
+      "step": 6070
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000266709368385875,
+      "loss": 0.6907,
+      "step": 6080
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026665367049125545,
+      "loss": 0.7815,
+      "step": 6090
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026659797259663586,
+      "loss": 0.7536,
+      "step": 6100
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026654227470201627,
+      "loss": 0.6969,
+      "step": 6110
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002664865768073967,
+      "loss": 0.6722,
+      "step": 6120
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002664308789127771,
+      "loss": 0.7757,
+      "step": 6130
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002663751810181575,
+      "loss": 0.7461,
+      "step": 6140
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002663194831235379,
+      "loss": 0.7681,
+      "step": 6150
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026626378522891833,
+      "loss": 0.6563,
+      "step": 6160
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026620808733429874,
+      "loss": 0.6421,
+      "step": 6170
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026615238943967915,
+      "loss": 0.7113,
+      "step": 6180
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026609669154505956,
+      "loss": 0.6726,
+      "step": 6190
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026604099365043997,
+      "loss": 0.9037,
+      "step": 6200
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002659852957558204,
+      "loss": 0.7431,
+      "step": 6210
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026592959786120085,
+      "loss": 0.644,
+      "step": 6220
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026587389996658126,
+      "loss": 0.7136,
+      "step": 6230
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026581820207196167,
+      "loss": 0.5934,
+      "step": 6240
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002657625041773421,
+      "loss": 0.7996,
+      "step": 6250
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002657068062827225,
+      "loss": 0.6854,
+      "step": 6260
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002656511083881029,
+      "loss": 0.704,
+      "step": 6270
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002655954104934833,
+      "loss": 0.7104,
+      "step": 6280
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002655397125988637,
+      "loss": 0.6173,
+      "step": 6290
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026548401470424414,
+      "loss": 0.6304,
+      "step": 6300
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026542831680962455,
+      "loss": 0.7566,
+      "step": 6310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026537261891500496,
+      "loss": 0.8521,
+      "step": 6320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002653169210203854,
+      "loss": 0.6862,
+      "step": 6330
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026526122312576584,
+      "loss": 0.6939,
+      "step": 6340
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026520552523114625,
+      "loss": 0.6616,
+      "step": 6350
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026514982733652666,
+      "loss": 0.6732,
+      "step": 6360
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026509412944190707,
+      "loss": 0.6309,
+      "step": 6370
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002650384315472875,
+      "loss": 0.6804,
+      "step": 6380
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002649827336526679,
+      "loss": 0.5404,
+      "step": 6390
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002649270357580483,
+      "loss": 0.7586,
+      "step": 6400
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002648713378634287,
+      "loss": 0.7994,
+      "step": 6410
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002648156399688092,
+      "loss": 0.6768,
+      "step": 6420
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002647599420741896,
+      "loss": 0.6574,
+      "step": 6430
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026470424417957,
+      "loss": 0.7092,
+      "step": 6440
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002646485462849504,
+      "loss": 0.6939,
+      "step": 6450
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002645928483903308,
+      "loss": 0.737,
+      "step": 6460
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026453715049571124,
+      "loss": 0.7043,
+      "step": 6470
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026448145260109165,
+      "loss": 0.6901,
+      "step": 6480
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026442575470647206,
+      "loss": 0.6087,
+      "step": 6490
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026437005681185247,
+      "loss": 0.7057,
+      "step": 6500
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002643143589172329,
+      "loss": 0.6944,
+      "step": 6510
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002642586610226133,
+      "loss": 0.6548,
+      "step": 6520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002642029631279937,
+      "loss": 0.8376,
+      "step": 6530
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002641528350228361,
+      "loss": 0.7284,
+      "step": 6540
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002640971371282165,
+      "loss": 0.6533,
+      "step": 6550
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026404143923359694,
+      "loss": 0.6211,
+      "step": 6560
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026398574133897735,
+      "loss": 0.5828,
+      "step": 6570
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026393004344435776,
+      "loss": 0.6483,
+      "step": 6580
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002638743455497382,
+      "loss": 0.6245,
+      "step": 6590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026381864765511864,
+      "loss": 0.6072,
+      "step": 6600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026376294976049905,
+      "loss": 0.7017,
+      "step": 6610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00026370725186587946,
+      "loss": 0.7893,
+      "step": 6620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00026365155397125987,
+      "loss": 0.6475,
+      "step": 6630
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002635958560766403,
+      "loss": 0.6877,
+      "step": 6640
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002635401581820207,
+      "loss": 0.7875,
+      "step": 6650
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002634844602874011,
+      "loss": 0.6411,
+      "step": 6660
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002634287623927815,
+      "loss": 0.673,
+      "step": 6670
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002633730644981619,
+      "loss": 0.7801,
+      "step": 6680
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00026331736660354234,
+      "loss": 0.6334,
+      "step": 6690
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00026326166870892275,
+      "loss": 0.5618,
+      "step": 6700
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00026320597081430316,
+      "loss": 0.6909,
+      "step": 6710
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002631502729196836,
+      "loss": 0.6804,
+      "step": 6720
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00026309457502506404,
+      "loss": 0.7555,
+      "step": 6730
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00026303887713044445,
+      "loss": 0.7498,
+      "step": 6740
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00026298317923582486,
+      "loss": 0.6513,
+      "step": 6750
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00026292748134120527,
+      "loss": 0.7818,
+      "step": 6760
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002628717834465857,
+      "loss": 0.7264,
+      "step": 6770
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00026281608555196615,
+      "loss": 0.5854,
+      "step": 6780
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00026276038765734656,
+      "loss": 0.8884,
+      "step": 6790
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00026270468976272697,
+      "loss": 0.7449,
+      "step": 6800
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002626489918681074,
+      "loss": 0.7061,
+      "step": 6810
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002625932939734878,
+      "loss": 0.6957,
+      "step": 6820
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002625375960788682,
+      "loss": 0.6842,
+      "step": 6830
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002624818981842486,
+      "loss": 0.5894,
+      "step": 6840
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000262426200289629,
+      "loss": 0.7376,
+      "step": 6850
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00026237050239500943,
+      "loss": 0.6895,
+      "step": 6860
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00026231480450038985,
+      "loss": 0.7339,
+      "step": 6870
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00026225910660577026,
+      "loss": 0.6097,
+      "step": 6880
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00026220340871115067,
+      "loss": 0.6427,
+      "step": 6890
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002621477108165311,
+      "loss": 0.7468,
+      "step": 6900
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00026209201292191154,
+      "loss": 0.5823,
+      "step": 6910
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00026203631502729196,
+      "loss": 0.6749,
+      "step": 6920
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00026198061713267237,
+      "loss": 0.7436,
+      "step": 6930
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002619249192380528,
+      "loss": 0.7514,
+      "step": 6940
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002618692213434332,
+      "loss": 0.7576,
+      "step": 6950
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002618135234488136,
+      "loss": 0.6901,
+      "step": 6960
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00026175782555419407,
+      "loss": 0.6756,
+      "step": 6970
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002617021276595745,
+      "loss": 0.6949,
+      "step": 6980
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002616464297649549,
+      "loss": 0.7156,
+      "step": 6990
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002615907318703353,
+      "loss": 0.7414,
+      "step": 7000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002615350339757157,
+      "loss": 0.7717,
+      "step": 7010
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002614793360810961,
+      "loss": 0.7553,
+      "step": 7020
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00026142363818647653,
+      "loss": 0.628,
+      "step": 7030
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00026136794029185694,
+      "loss": 0.6224,
+      "step": 7040
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00026131224239723735,
+      "loss": 0.79,
+      "step": 7050
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00026125654450261777,
+      "loss": 0.6752,
+      "step": 7060
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002612008466079982,
+      "loss": 0.6976,
+      "step": 7070
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002611451487133786,
+      "loss": 0.6752,
+      "step": 7080
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000261089450818759,
+      "loss": 0.7177,
+      "step": 7090
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00026103375292413946,
+      "loss": 0.5774,
+      "step": 7100
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002609780550295199,
+      "loss": 0.7178,
+      "step": 7110
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002609223571349003,
+      "loss": 0.7939,
+      "step": 7120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002608666592402807,
+      "loss": 0.7736,
+      "step": 7130
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002608109613456611,
+      "loss": 0.6933,
+      "step": 7140
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002607552634510415,
+      "loss": 0.6997,
+      "step": 7150
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00026069956555642193,
+      "loss": 0.6604,
+      "step": 7160
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002606438676618024,
+      "loss": 0.7637,
+      "step": 7170
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002605881697671828,
+      "loss": 0.6724,
+      "step": 7180
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002605324718725632,
+      "loss": 0.7258,
+      "step": 7190
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00026047677397794363,
+      "loss": 0.7205,
+      "step": 7200
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00026042107608332404,
+      "loss": 0.7266,
+      "step": 7210
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00026036537818870445,
+      "loss": 0.842,
+      "step": 7220
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00026030968029408486,
+      "loss": 0.7843,
+      "step": 7230
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002602539823994653,
+      "loss": 0.6904,
+      "step": 7240
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002601982845048457,
+      "loss": 0.6767,
+      "step": 7250
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002601425866102261,
+      "loss": 0.7035,
+      "step": 7260
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002600868887156065,
+      "loss": 0.7647,
+      "step": 7270
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002600311908209869,
+      "loss": 0.6093,
+      "step": 7280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025997549292636733,
+      "loss": 0.6147,
+      "step": 7290
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002599197950317478,
+      "loss": 0.7109,
+      "step": 7300
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002598640971371282,
+      "loss": 0.7692,
+      "step": 7310
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002598083992425086,
+      "loss": 0.654,
+      "step": 7320
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025975270134788903,
+      "loss": 0.7219,
+      "step": 7330
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025969700345326944,
+      "loss": 0.7073,
+      "step": 7340
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025964130555864985,
+      "loss": 0.5878,
+      "step": 7350
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002595856076640303,
+      "loss": 0.6496,
+      "step": 7360
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025952990976941073,
+      "loss": 0.7443,
+      "step": 7370
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025947421187479114,
+      "loss": 0.7504,
+      "step": 7380
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025941851398017155,
+      "loss": 0.7563,
+      "step": 7390
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025936281608555196,
+      "loss": 0.71,
+      "step": 7400
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025930711819093237,
+      "loss": 0.7584,
+      "step": 7410
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002592514202963128,
+      "loss": 0.6944,
+      "step": 7420
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002591957224016932,
+      "loss": 0.7867,
+      "step": 7430
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002591400245070736,
+      "loss": 0.6939,
+      "step": 7440
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000259084326612454,
+      "loss": 0.7867,
+      "step": 7450
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00025902862871783443,
+      "loss": 0.8446,
+      "step": 7460
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00025897293082321484,
+      "loss": 0.782,
+      "step": 7470
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00025891723292859525,
+      "loss": 0.818,
+      "step": 7480
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002588615350339757,
+      "loss": 0.8101,
+      "step": 7490
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002588058371393561,
+      "loss": 0.6433,
+      "step": 7500
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00025875013924473654,
+      "loss": 0.8316,
+      "step": 7510
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00025869444135011695,
+      "loss": 0.8513,
+      "step": 7520
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00025863874345549736,
+      "loss": 0.6972,
+      "step": 7530
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00025858304556087777,
+      "loss": 0.8297,
+      "step": 7540
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002585273476662582,
+      "loss": 0.6686,
+      "step": 7550
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00025847164977163865,
+      "loss": 0.6718,
+      "step": 7560
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00025841595187701906,
+      "loss": 0.7151,
+      "step": 7570
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00025836025398239947,
+      "loss": 0.6683,
+      "step": 7580
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002583045560877799,
+      "loss": 0.8528,
+      "step": 7590
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002582488581931603,
+      "loss": 0.8118,
+      "step": 7600
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002581931602985407,
+      "loss": 0.7875,
+      "step": 7610
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002581374624039211,
+      "loss": 0.6881,
+      "step": 7620
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002580817645093015,
+      "loss": 0.6904,
+      "step": 7630
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00025802606661468194,
+      "loss": 0.7501,
+      "step": 7640
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00025797036872006235,
+      "loss": 0.5125,
+      "step": 7650
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00025791467082544276,
+      "loss": 0.6751,
+      "step": 7660
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00025785897293082317,
+      "loss": 0.6571,
+      "step": 7670
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002578032750362036,
+      "loss": 0.6458,
+      "step": 7680
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00025774757714158405,
+      "loss": 0.5871,
+      "step": 7690
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00025769187924696446,
+      "loss": 0.6758,
+      "step": 7700
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00025763618135234487,
+      "loss": 0.8249,
+      "step": 7710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002575804834577253,
+      "loss": 0.7329,
+      "step": 7720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002575247855631057,
+      "loss": 0.9108,
+      "step": 7730
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002574690876684861,
+      "loss": 0.7389,
+      "step": 7740
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00025741338977386657,
+      "loss": 0.7714,
+      "step": 7750
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000257357691879247,
+      "loss": 0.7028,
+      "step": 7760
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002573019939846274,
+      "loss": 0.7169,
+      "step": 7770
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002572462960900078,
+      "loss": 0.6585,
+      "step": 7780
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002571905981953882,
+      "loss": 0.6287,
+      "step": 7790
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002571349003007686,
+      "loss": 0.6507,
+      "step": 7800
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00025707920240614903,
+      "loss": 0.7753,
+      "step": 7810
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00025702350451152945,
+      "loss": 0.7194,
+      "step": 7820
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00025696780661690986,
+      "loss": 0.568,
+      "step": 7830
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00025691210872229027,
+      "loss": 0.6429,
+      "step": 7840
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002568564108276707,
+      "loss": 0.7006,
+      "step": 7850
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002568007129330511,
+      "loss": 0.6556,
+      "step": 7860
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002567450150384315,
+      "loss": 0.7786,
+      "step": 7870
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00025668931714381197,
+      "loss": 0.6277,
+      "step": 7880
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002566336192491924,
+      "loss": 0.6289,
+      "step": 7890
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002565779213545728,
+      "loss": 0.6624,
+      "step": 7900
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002565222234599532,
+      "loss": 0.6011,
+      "step": 7910
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002564665255653336,
+      "loss": 0.6754,
+      "step": 7920
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000256410827670714,
+      "loss": 0.6884,
+      "step": 7930
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00025635512977609443,
+      "loss": 0.5884,
+      "step": 7940
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00025629943188147484,
+      "loss": 0.6332,
+      "step": 7950
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00025624373398685526,
+      "loss": 0.7384,
+      "step": 7960
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00025618803609223567,
+      "loss": 0.8223,
+      "step": 7970
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002561323381976161,
+      "loss": 0.6917,
+      "step": 7980
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00025607664030299654,
+      "loss": 0.6033,
+      "step": 7990
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00025602094240837695,
+      "loss": 0.644,
+      "step": 8000
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00025596524451375737,
+      "loss": 0.8084,
+      "step": 8010
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002559095466191378,
+      "loss": 0.7939,
+      "step": 8020
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00025585941851398014,
+      "loss": 0.7747,
+      "step": 8030
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00025580372061936055,
+      "loss": 0.6367,
+      "step": 8040
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000255748022724741,
+      "loss": 0.7061,
+      "step": 8050
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002556923248301214,
+      "loss": 0.6948,
+      "step": 8060
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00025563662693550183,
+      "loss": 0.706,
+      "step": 8070
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00025558092904088225,
+      "loss": 0.6919,
+      "step": 8080
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00025552523114626266,
+      "loss": 0.614,
+      "step": 8090
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00025546953325164307,
+      "loss": 0.7104,
+      "step": 8100
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002554138353570235,
+      "loss": 0.6637,
+      "step": 8110
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002553581374624039,
+      "loss": 0.7115,
+      "step": 8120
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002553024395677843,
+      "loss": 0.7232,
+      "step": 8130
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002552467416731647,
+      "loss": 0.7147,
+      "step": 8140
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002551910437785451,
+      "loss": 0.6032,
+      "step": 8150
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00025513534588392553,
+      "loss": 0.6833,
+      "step": 8160
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00025507964798930595,
+      "loss": 0.783,
+      "step": 8170
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002550239500946864,
+      "loss": 0.6464,
+      "step": 8180
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002549682522000668,
+      "loss": 0.6631,
+      "step": 8190
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00025491255430544723,
+      "loss": 0.6535,
+      "step": 8200
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00025485685641082764,
+      "loss": 0.7087,
+      "step": 8210
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00025480115851620806,
+      "loss": 0.6446,
+      "step": 8220
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00025474546062158847,
+      "loss": 0.6835,
+      "step": 8230
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002546897627269689,
+      "loss": 0.7274,
+      "step": 8240
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00025463406483234934,
+      "loss": 0.7993,
+      "step": 8250
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00025457836693772975,
+      "loss": 0.7761,
+      "step": 8260
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00025452266904311017,
+      "loss": 0.6708,
+      "step": 8270
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002544669711484906,
+      "loss": 0.6827,
+      "step": 8280
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000254411273253871,
+      "loss": 0.6779,
+      "step": 8290
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002543555753592514,
+      "loss": 0.7011,
+      "step": 8300
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002542998774646318,
+      "loss": 0.6129,
+      "step": 8310
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002542441795700122,
+      "loss": 0.635,
+      "step": 8320
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00025418848167539263,
+      "loss": 0.6126,
+      "step": 8330
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00025413278378077304,
+      "loss": 0.6761,
+      "step": 8340
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00025407708588615345,
+      "loss": 0.7954,
+      "step": 8350
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00025402138799153387,
+      "loss": 0.5889,
+      "step": 8360
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002539656900969143,
+      "loss": 0.7154,
+      "step": 8370
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00025390999220229474,
+      "loss": 0.7932,
+      "step": 8380
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00025385429430767515,
+      "loss": 0.7512,
+      "step": 8390
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00025379859641305556,
+      "loss": 0.6542,
+      "step": 8400
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000253742898518436,
+      "loss": 0.7078,
+      "step": 8410
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002536872006238164,
+      "loss": 0.7645,
+      "step": 8420
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002536315027291968,
+      "loss": 0.7741,
+      "step": 8430
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00025357580483457726,
+      "loss": 0.791,
+      "step": 8440
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002535201069399577,
+      "loss": 0.6548,
+      "step": 8450
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002534644090453381,
+      "loss": 0.5454,
+      "step": 8460
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002534087111507185,
+      "loss": 0.681,
+      "step": 8470
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002533530132560989,
+      "loss": 0.7249,
+      "step": 8480
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002532973153614793,
+      "loss": 0.6408,
+      "step": 8490
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00025324161746685973,
+      "loss": 0.6231,
+      "step": 8500
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00025318591957224014,
+      "loss": 0.6256,
+      "step": 8510
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00025313022167762055,
+      "loss": 0.7,
+      "step": 8520
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00025307452378300096,
+      "loss": 0.6913,
+      "step": 8530
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002530188258883814,
+      "loss": 0.6713,
+      "step": 8540
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002529631279937618,
+      "loss": 0.692,
+      "step": 8550
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002529074300991422,
+      "loss": 0.6492,
+      "step": 8560
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00025285173220452266,
+      "loss": 0.7063,
+      "step": 8570
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002527960343099031,
+      "loss": 0.6779,
+      "step": 8580
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002527403364152835,
+      "loss": 0.6732,
+      "step": 8590
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002526846385206639,
+      "loss": 0.6546,
+      "step": 8600
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002526289406260443,
+      "loss": 0.6366,
+      "step": 8610
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002525732427314247,
+      "loss": 0.6969,
+      "step": 8620
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00025251754483680513,
+      "loss": 0.7025,
+      "step": 8630
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002524618469421856,
+      "loss": 0.7001,
+      "step": 8640
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000252406149047566,
+      "loss": 0.6574,
+      "step": 8650
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002523504511529464,
+      "loss": 0.7517,
+      "step": 8660
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00025229475325832683,
+      "loss": 0.6742,
+      "step": 8670
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00025223905536370724,
+      "loss": 0.7066,
+      "step": 8680
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00025218335746908765,
+      "loss": 0.7134,
+      "step": 8690
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00025212765957446806,
+      "loss": 0.5786,
+      "step": 8700
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00025207196167984847,
+      "loss": 0.702,
+      "step": 8710
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002520162637852289,
+      "loss": 0.6488,
+      "step": 8720
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002519605658906093,
+      "loss": 0.7283,
+      "step": 8730
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002519048679959897,
+      "loss": 0.7225,
+      "step": 8740
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002518491701013701,
+      "loss": 0.6083,
+      "step": 8750
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00025179347220675053,
+      "loss": 0.6344,
+      "step": 8760
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000251737774312131,
+      "loss": 0.5753,
+      "step": 8770
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002516820764175114,
+      "loss": 0.8158,
+      "step": 8780
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002516263785228918,
+      "loss": 0.7025,
+      "step": 8790
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002515706806282722,
+      "loss": 0.6105,
+      "step": 8800
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00025151498273365264,
+      "loss": 0.7146,
+      "step": 8810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00025145928483903305,
+      "loss": 0.7568,
+      "step": 8820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002514035869444135,
+      "loss": 0.7616,
+      "step": 8830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002513478890497939,
+      "loss": 0.7579,
+      "step": 8840
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00025129219115517434,
+      "loss": 0.6883,
+      "step": 8850
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00025123649326055475,
+      "loss": 0.7468,
+      "step": 8860
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00025118079536593516,
+      "loss": 0.654,
+      "step": 8870
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00025112509747131557,
+      "loss": 0.709,
+      "step": 8880
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000251069399576696,
+      "loss": 0.9533,
+      "step": 8890
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002510137016820764,
+      "loss": 0.6727,
+      "step": 8900
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002509580037874568,
+      "loss": 0.6831,
+      "step": 8910
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002509023058928372,
+      "loss": 0.7606,
+      "step": 8920
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002508466079982176,
+      "loss": 0.7022,
+      "step": 8930
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00025079091010359804,
+      "loss": 0.7217,
+      "step": 8940
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00025073521220897845,
+      "loss": 0.7169,
+      "step": 8950
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002506795143143589,
+      "loss": 0.6716,
+      "step": 8960
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002506238164197393,
+      "loss": 0.6086,
+      "step": 8970
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00025056811852511974,
+      "loss": 0.5452,
+      "step": 8980
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00025051242063050015,
+      "loss": 0.5555,
+      "step": 8990
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00025045672273588056,
+      "loss": 0.6223,
+      "step": 9000
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00025040102484126097,
+      "loss": 0.5115,
+      "step": 9010
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00025034532694664143,
+      "loss": 0.6867,
+      "step": 9020
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00025028962905202185,
+      "loss": 0.6013,
+      "step": 9030
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00025023393115740226,
+      "loss": 0.6702,
+      "step": 9040
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00025017823326278267,
+      "loss": 0.7934,
+      "step": 9050
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002501225353681631,
+      "loss": 0.613,
+      "step": 9060
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002500668374735435,
+      "loss": 0.7321,
+      "step": 9070
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002500111395789239,
+      "loss": 0.6458,
+      "step": 9080
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002499554416843043,
+      "loss": 0.6609,
+      "step": 9090
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002498997437896847,
+      "loss": 0.5997,
+      "step": 9100
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00024984404589506513,
+      "loss": 0.8273,
+      "step": 9110
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00024978834800044555,
+      "loss": 0.6497,
+      "step": 9120
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00024973265010582596,
+      "loss": 0.6347,
+      "step": 9130
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00024967695221120637,
+      "loss": 0.7549,
+      "step": 9140
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00024962125431658683,
+      "loss": 0.7474,
+      "step": 9150
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00024956555642196724,
+      "loss": 0.6484,
+      "step": 9160
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00024950985852734766,
+      "loss": 0.774,
+      "step": 9170
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00024945416063272807,
+      "loss": 0.6239,
+      "step": 9180
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002493984627381085,
+      "loss": 0.6637,
+      "step": 9190
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002493427648434889,
+      "loss": 0.703,
+      "step": 9200
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002492870669488693,
+      "loss": 0.6582,
+      "step": 9210
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00024923136905424977,
+      "loss": 0.6731,
+      "step": 9220
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002491756711596302,
+      "loss": 0.7721,
+      "step": 9230
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002491199732650106,
+      "loss": 0.662,
+      "step": 9240
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000249064275370391,
+      "loss": 0.6991,
+      "step": 9250
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002490085774757714,
+      "loss": 0.7294,
+      "step": 9260
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002489528795811518,
+      "loss": 0.666,
+      "step": 9270
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00024889718168653223,
+      "loss": 0.7909,
+      "step": 9280
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00024884148379191264,
+      "loss": 0.8497,
+      "step": 9290
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00024878578589729305,
+      "loss": 0.7551,
+      "step": 9300
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00024873008800267347,
+      "loss": 0.6879,
+      "step": 9310
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002486743901080539,
+      "loss": 0.7592,
+      "step": 9320
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002486186922134343,
+      "loss": 0.6669,
+      "step": 9330
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002485629943188147,
+      "loss": 0.5728,
+      "step": 9340
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00024850729642419516,
+      "loss": 0.6912,
+      "step": 9350
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002484515985295756,
+      "loss": 0.82,
+      "step": 9360
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000248395900634956,
+      "loss": 0.8179,
+      "step": 9370
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002483402027403364,
+      "loss": 0.59,
+      "step": 9380
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002482845048457168,
+      "loss": 0.6049,
+      "step": 9390
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002482288069510972,
+      "loss": 0.6329,
+      "step": 9400
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002481731090564777,
+      "loss": 0.7197,
+      "step": 9410
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002481174111618581,
+      "loss": 0.6503,
+      "step": 9420
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002480617132672385,
+      "loss": 0.7675,
+      "step": 9430
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002480060153726189,
+      "loss": 0.6315,
+      "step": 9440
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00024795031747799933,
+      "loss": 0.613,
+      "step": 9450
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00024789461958337974,
+      "loss": 0.7848,
+      "step": 9460
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00024783892168876015,
+      "loss": 0.7219,
+      "step": 9470
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00024778322379414056,
+      "loss": 0.5723,
+      "step": 9480
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000247727525899521,
+      "loss": 0.7077,
+      "step": 9490
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002476718280049014,
+      "loss": 0.5787,
+      "step": 9500
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002476161301102818,
+      "loss": 0.5726,
+      "step": 9510
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002475604322156622,
+      "loss": 0.6345,
+      "step": 9520
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002475047343210426,
+      "loss": 0.7327,
+      "step": 9530
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002474490364264231,
+      "loss": 0.7187,
+      "step": 9540
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002473933385318035,
+      "loss": 0.7135,
+      "step": 9550
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002473376406371839,
+      "loss": 0.7937,
+      "step": 9560
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002472819427425643,
+      "loss": 0.7391,
+      "step": 9570
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00024722624484794473,
+      "loss": 0.6772,
+      "step": 9580
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00024717054695332514,
+      "loss": 0.7811,
+      "step": 9590
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00024711484905870555,
+      "loss": 0.7126,
+      "step": 9600
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00024705915116408596,
+      "loss": 0.6437,
+      "step": 9610
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002470034532694664,
+      "loss": 0.6263,
+      "step": 9620
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002469477553748468,
+      "loss": 0.7226,
+      "step": 9630
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00024689205748022725,
+      "loss": 0.7301,
+      "step": 9640
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00024683635958560766,
+      "loss": 0.6653,
+      "step": 9650
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00024678066169098807,
+      "loss": 0.7842,
+      "step": 9660
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002467249637963685,
+      "loss": 0.6937,
+      "step": 9670
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002466692659017489,
+      "loss": 0.6652,
+      "step": 9680
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002466135680071293,
+      "loss": 0.771,
+      "step": 9690
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002465578701125097,
+      "loss": 0.756,
+      "step": 9700
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00024650217221789013,
+      "loss": 0.7209,
+      "step": 9710
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00024644647432327054,
+      "loss": 0.7682,
+      "step": 9720
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00024639077642865095,
+      "loss": 0.7506,
+      "step": 9730
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002463350785340314,
+      "loss": 0.5626,
+      "step": 9740
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002462793806394118,
+      "loss": 0.8043,
+      "step": 9750
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00024622368274479224,
+      "loss": 0.6046,
+      "step": 9760
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00024616798485017265,
+      "loss": 0.8198,
+      "step": 9770
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00024611228695555306,
+      "loss": 0.6326,
+      "step": 9780
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00024605658906093347,
+      "loss": 0.7361,
+      "step": 9790
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002460008911663139,
+      "loss": 0.7939,
+      "step": 9800
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002459451932716943,
+      "loss": 0.6577,
+      "step": 9810
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002458894953770747,
+      "loss": 0.6407,
+      "step": 9820
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002458337974824551,
+      "loss": 0.7183,
+      "step": 9830
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002457780995878355,
+      "loss": 0.7458,
+      "step": 9840
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00024572240169321594,
+      "loss": 0.6925,
+      "step": 9850
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00024566670379859635,
+      "loss": 0.7651,
+      "step": 9860
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002456110059039768,
+      "loss": 0.662,
+      "step": 9870
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002455553080093572,
+      "loss": 0.6219,
+      "step": 9880
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00024549961011473764,
+      "loss": 0.6743,
+      "step": 9890
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00024544391222011805,
+      "loss": 0.6198,
+      "step": 9900
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00024538821432549846,
+      "loss": 0.7147,
+      "step": 9910
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00024533251643087887,
+      "loss": 0.8246,
+      "step": 9920
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00024527681853625934,
+      "loss": 0.702,
+      "step": 9930
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00024522112064163975,
+      "loss": 0.7672,
+      "step": 9940
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00024516542274702016,
+      "loss": 0.7388,
+      "step": 9950
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00024510972485240057,
+      "loss": 0.7253,
+      "step": 9960
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000245054026957781,
+      "loss": 0.6616,
+      "step": 9970
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002449983290631614,
+      "loss": 0.617,
+      "step": 9980
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002449426311685418,
+      "loss": 0.7662,
+      "step": 9990
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002448869332739222,
+      "loss": 0.8083,
+      "step": 10000
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002448312353793026,
+      "loss": 0.7273,
+      "step": 10010
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00024477553748468304,
+      "loss": 0.61,
+      "step": 10020
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00024471983959006345,
+      "loss": 0.8243,
+      "step": 10030
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00024466414169544386,
+      "loss": 0.7192,
+      "step": 10040
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00024460844380082427,
+      "loss": 0.4855,
+      "step": 10050
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00024455274590620473,
+      "loss": 0.707,
+      "step": 10060
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00024449704801158515,
+      "loss": 0.7075,
+      "step": 10070
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00024444135011696556,
+      "loss": 0.6282,
+      "step": 10080
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00024438565222234597,
+      "loss": 0.6695,
+      "step": 10090
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002443299543277264,
+      "loss": 0.5489,
+      "step": 10100
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002442742564331068,
+      "loss": 0.7393,
+      "step": 10110
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002442185585384872,
+      "loss": 0.6557,
+      "step": 10120
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00024416286064386767,
+      "loss": 0.6952,
+      "step": 10130
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00024410716274924805,
+      "loss": 0.7737,
+      "step": 10140
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002440514648546285,
+      "loss": 0.711,
+      "step": 10150
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002439957669600089,
+      "loss": 0.6748,
+      "step": 10160
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002439400690653893,
+      "loss": 0.773,
+      "step": 10170
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00024388437117076972,
+      "loss": 0.8506,
+      "step": 10180
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00024382867327615016,
+      "loss": 0.6262,
+      "step": 10190
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00024377297538153057,
+      "loss": 0.7233,
+      "step": 10200
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00024371727748691098,
+      "loss": 0.6016,
+      "step": 10210
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002436615795922914,
+      "loss": 0.6963,
+      "step": 10220
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002436058816976718,
+      "loss": 0.6437,
+      "step": 10230
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00024355018380305222,
+      "loss": 0.645,
+      "step": 10240
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00024349448590843263,
+      "loss": 0.6856,
+      "step": 10250
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00024343878801381307,
+      "loss": 0.6308,
+      "step": 10260
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00024338309011919348,
+      "loss": 0.6025,
+      "step": 10270
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002433273922245739,
+      "loss": 0.6491,
+      "step": 10280
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002432716943299543,
+      "loss": 0.6953,
+      "step": 10290
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002432159964353347,
+      "loss": 0.641,
+      "step": 10300
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00024316029854071512,
+      "loss": 0.6919,
+      "step": 10310
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00024310460064609556,
+      "loss": 0.7419,
+      "step": 10320
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00024304890275147597,
+      "loss": 0.639,
+      "step": 10330
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00024299320485685638,
+      "loss": 0.6623,
+      "step": 10340
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002429375069622368,
+      "loss": 0.5978,
+      "step": 10350
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002428818090676172,
+      "loss": 0.5833,
+      "step": 10360
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00024282611117299762,
+      "loss": 0.6002,
+      "step": 10370
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00024277041327837803,
+      "loss": 0.7811,
+      "step": 10380
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002427147153837585,
+      "loss": 0.5564,
+      "step": 10390
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002426590174891389,
+      "loss": 0.8002,
+      "step": 10400
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00024260331959451931,
+      "loss": 0.685,
+      "step": 10410
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00024254762169989973,
+      "loss": 0.766,
+      "step": 10420
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00024249192380528014,
+      "loss": 0.7783,
+      "step": 10430
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00024243622591066055,
+      "loss": 0.8144,
+      "step": 10440
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00024238052801604099,
+      "loss": 0.6686,
+      "step": 10450
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002423248301214214,
+      "loss": 0.7232,
+      "step": 10460
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002422691322268018,
+      "loss": 0.7746,
+      "step": 10470
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00024221343433218222,
+      "loss": 0.8329,
+      "step": 10480
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00024215773643756263,
+      "loss": 0.6367,
+      "step": 10490
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00024210203854294304,
+      "loss": 0.7414,
+      "step": 10500
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00024204634064832345,
+      "loss": 0.7459,
+      "step": 10510
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002419906427537039,
+      "loss": 0.5564,
+      "step": 10520
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002419349448590843,
+      "loss": 0.7468,
+      "step": 10530
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002418792469644647,
+      "loss": 0.6507,
+      "step": 10540
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00024182354906984512,
+      "loss": 0.7085,
+      "step": 10550
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00024176785117522554,
+      "loss": 0.6424,
+      "step": 10560
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00024171215328060595,
+      "loss": 0.7305,
+      "step": 10570
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002416564553859864,
+      "loss": 0.7546,
+      "step": 10580
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00024160075749136682,
+      "loss": 0.705,
+      "step": 10590
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00024154505959674723,
+      "loss": 0.6509,
+      "step": 10600
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00024148936170212765,
+      "loss": 0.6708,
+      "step": 10610
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00024143366380750806,
+      "loss": 0.7686,
+      "step": 10620
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00024137796591288847,
+      "loss": 0.714,
+      "step": 10630
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00024132226801826888,
+      "loss": 0.6741,
+      "step": 10640
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00024126657012364932,
+      "loss": 0.6817,
+      "step": 10650
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00024121087222902973,
+      "loss": 0.6172,
+      "step": 10660
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00024115517433441014,
+      "loss": 0.6626,
+      "step": 10670
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00024109947643979055,
+      "loss": 0.6571,
+      "step": 10680
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00024104377854517096,
+      "loss": 0.7406,
+      "step": 10690
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00024098808065055137,
+      "loss": 0.7141,
+      "step": 10700
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002409323827559318,
+      "loss": 0.7329,
+      "step": 10710
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00024087668486131222,
+      "loss": 0.7647,
+      "step": 10720
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00024082098696669263,
+      "loss": 0.6834,
+      "step": 10730
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00024076528907207304,
+      "loss": 0.7599,
+      "step": 10740
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00024070959117745346,
+      "loss": 0.6467,
+      "step": 10750
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00024065389328283387,
+      "loss": 0.7246,
+      "step": 10760
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00024059819538821433,
+      "loss": 0.657,
+      "step": 10770
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00024054249749359474,
+      "loss": 0.7094,
+      "step": 10780
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00024048679959897515,
+      "loss": 0.584,
+      "step": 10790
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00024043110170435557,
+      "loss": 0.8219,
+      "step": 10800
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00024037540380973598,
+      "loss": 0.6759,
+      "step": 10810
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002403197059151164,
+      "loss": 0.8561,
+      "step": 10820
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002402640080204968,
+      "loss": 0.724,
+      "step": 10830
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00024020831012587724,
+      "loss": 0.7839,
+      "step": 10840
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00024015261223125765,
+      "loss": 0.7722,
+      "step": 10850
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00024009691433663806,
+      "loss": 0.7929,
+      "step": 10860
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00024004121644201847,
+      "loss": 0.6527,
+      "step": 10870
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00023998551854739888,
+      "loss": 0.708,
+      "step": 10880
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002399298206527793,
+      "loss": 0.6704,
+      "step": 10890
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00023987412275815973,
+      "loss": 0.8032,
+      "step": 10900
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00023981842486354014,
+      "loss": 0.6628,
+      "step": 10910
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00023976272696892055,
+      "loss": 0.7161,
+      "step": 10920
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00023970702907430096,
+      "loss": 0.7405,
+      "step": 10930
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00023965133117968138,
+      "loss": 0.6299,
+      "step": 10940
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002395956332850618,
+      "loss": 0.641,
+      "step": 10950
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002395399353904422,
+      "loss": 0.788,
+      "step": 10960
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00023948423749582266,
+      "loss": 0.7103,
+      "step": 10970
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00023942853960120307,
+      "loss": 0.7394,
+      "step": 10980
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00023937284170658349,
+      "loss": 0.6484,
+      "step": 10990
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002393171438119639,
+      "loss": 0.6061,
+      "step": 11000
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002392614459173443,
+      "loss": 0.7456,
+      "step": 11010
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00023920574802272472,
+      "loss": 0.6627,
+      "step": 11020
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00023915005012810516,
+      "loss": 0.5394,
+      "step": 11030
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00023909435223348557,
+      "loss": 0.7376,
+      "step": 11040
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00023903865433886598,
+      "loss": 0.841,
+      "step": 11050
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002389829564442464,
+      "loss": 0.5351,
+      "step": 11060
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002389272585496268,
+      "loss": 0.7919,
+      "step": 11070
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002388715606550072,
+      "loss": 0.6437,
+      "step": 11080
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00023881586276038762,
+      "loss": 0.6065,
+      "step": 11090
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00023876016486576806,
+      "loss": 0.7801,
+      "step": 11100
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00023870446697114847,
+      "loss": 0.6902,
+      "step": 11110
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00023864876907652888,
+      "loss": 0.5621,
+      "step": 11120
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002385930711819093,
+      "loss": 0.6552,
+      "step": 11130
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002385373732872897,
+      "loss": 0.6235,
+      "step": 11140
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00023848167539267012,
+      "loss": 0.6562,
+      "step": 11150
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00023842597749805056,
+      "loss": 0.7234,
+      "step": 11160
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00023837027960343097,
+      "loss": 0.6843,
+      "step": 11170
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002383145817088114,
+      "loss": 0.7269,
+      "step": 11180
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00023825888381419182,
+      "loss": 0.7298,
+      "step": 11190
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00023820318591957223,
+      "loss": 0.7556,
+      "step": 11200
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00023814748802495264,
+      "loss": 0.5612,
+      "step": 11210
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00023809179013033305,
+      "loss": 0.7553,
+      "step": 11220
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002380360922357135,
+      "loss": 0.7956,
+      "step": 11230
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002379803943410939,
+      "loss": 0.6301,
+      "step": 11240
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002379246964464743,
+      "loss": 0.5737,
+      "step": 11250
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00023786899855185472,
+      "loss": 0.6406,
+      "step": 11260
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00023781330065723513,
+      "loss": 0.7076,
+      "step": 11270
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00023775760276261554,
+      "loss": 0.7455,
+      "step": 11280
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00023770190486799598,
+      "loss": 0.7002,
+      "step": 11290
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002376462069733764,
+      "loss": 0.6286,
+      "step": 11300
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002375905090787568,
+      "loss": 0.756,
+      "step": 11310
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00023753481118413722,
+      "loss": 0.6186,
+      "step": 11320
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00023747911328951763,
+      "loss": 0.7172,
+      "step": 11330
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00023742341539489804,
+      "loss": 0.7271,
+      "step": 11340
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00023736771750027845,
+      "loss": 0.7748,
+      "step": 11350
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002373120196056589,
+      "loss": 0.5905,
+      "step": 11360
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002372563217110393,
+      "loss": 0.6501,
+      "step": 11370
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002372006238164197,
+      "loss": 0.6974,
+      "step": 11380
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00023714492592180012,
+      "loss": 0.7366,
+      "step": 11390
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00023708922802718053,
+      "loss": 0.6285,
+      "step": 11400
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00023703353013256094,
+      "loss": 0.655,
+      "step": 11410
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002369778322379414,
+      "loss": 0.6715,
+      "step": 11420
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00023692213434332182,
+      "loss": 0.7881,
+      "step": 11430
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00023686643644870223,
+      "loss": 0.7588,
+      "step": 11440
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00023681073855408264,
+      "loss": 0.733,
+      "step": 11450
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00023675504065946305,
+      "loss": 0.7324,
+      "step": 11460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023669934276484346,
+      "loss": 0.7302,
+      "step": 11470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023664364487022387,
+      "loss": 0.6896,
+      "step": 11480
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002365879469756043,
+      "loss": 0.7044,
+      "step": 11490
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023653224908098472,
+      "loss": 0.6727,
+      "step": 11500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023647655118636514,
+      "loss": 0.593,
+      "step": 11510
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023642085329174555,
+      "loss": 0.5973,
+      "step": 11520
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023636515539712596,
+      "loss": 0.672,
+      "step": 11530
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023630945750250637,
+      "loss": 0.6067,
+      "step": 11540
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002362537596078868,
+      "loss": 0.6839,
+      "step": 11550
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023619806171326722,
+      "loss": 0.6733,
+      "step": 11560
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023614236381864763,
+      "loss": 0.7261,
+      "step": 11570
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023608666592402804,
+      "loss": 0.8264,
+      "step": 11580
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023603096802940845,
+      "loss": 0.8673,
+      "step": 11590
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023597527013478886,
+      "loss": 0.7077,
+      "step": 11600
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023591957224016927,
+      "loss": 0.7507,
+      "step": 11610
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023586387434554974,
+      "loss": 0.5535,
+      "step": 11620
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023580817645093015,
+      "loss": 0.6683,
+      "step": 11630
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023575247855631056,
+      "loss": 0.5623,
+      "step": 11640
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023569678066169097,
+      "loss": 0.6956,
+      "step": 11650
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023564108276707138,
+      "loss": 0.6795,
+      "step": 11660
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002355853848724518,
+      "loss": 0.6907,
+      "step": 11670
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023552968697783223,
+      "loss": 0.6775,
+      "step": 11680
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023547398908321264,
+      "loss": 0.8199,
+      "step": 11690
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023541829118859306,
+      "loss": 0.637,
+      "step": 11700
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023536259329397347,
+      "loss": 0.779,
+      "step": 11710
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00023530689539935388,
+      "loss": 0.5617,
+      "step": 11720
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002352511975047343,
+      "loss": 0.562,
+      "step": 11730
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002351954996101147,
+      "loss": 0.585,
+      "step": 11740
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023513980171549514,
+      "loss": 0.6664,
+      "step": 11750
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023508410382087555,
+      "loss": 0.6776,
+      "step": 11760
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023502840592625596,
+      "loss": 0.5982,
+      "step": 11770
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023497270803163637,
+      "loss": 0.6264,
+      "step": 11780
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023491701013701678,
+      "loss": 0.6981,
+      "step": 11790
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002348613122423972,
+      "loss": 0.5879,
+      "step": 11800
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023480561434777766,
+      "loss": 0.6335,
+      "step": 11810
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023474991645315807,
+      "loss": 0.7219,
+      "step": 11820
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023469421855853848,
+      "loss": 0.6749,
+      "step": 11830
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002346385206639189,
+      "loss": 0.6922,
+      "step": 11840
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002345828227692993,
+      "loss": 0.5973,
+      "step": 11850
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023452712487467971,
+      "loss": 0.7362,
+      "step": 11860
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023447142698006013,
+      "loss": 0.758,
+      "step": 11870
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023441572908544056,
+      "loss": 0.6899,
+      "step": 11880
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023436003119082098,
+      "loss": 0.6532,
+      "step": 11890
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002343043332962014,
+      "loss": 0.7569,
+      "step": 11900
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002342486354015818,
+      "loss": 0.6823,
+      "step": 11910
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002341929375069622,
+      "loss": 0.662,
+      "step": 11920
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023413723961234262,
+      "loss": 0.6535,
+      "step": 11930
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023408154171772306,
+      "loss": 0.7318,
+      "step": 11940
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023402584382310347,
+      "loss": 0.702,
+      "step": 11950
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023397014592848388,
+      "loss": 0.8139,
+      "step": 11960
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002339144480338643,
+      "loss": 0.5758,
+      "step": 11970
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002338587501392447,
+      "loss": 0.6429,
+      "step": 11980
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023380305224462511,
+      "loss": 0.7322,
+      "step": 11990
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023374735435000553,
+      "loss": 0.5917,
+      "step": 12000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000233691656455386,
+      "loss": 0.6779,
+      "step": 12010
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002336359585607664,
+      "loss": 0.5901,
+      "step": 12020
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002335802606661468,
+      "loss": 0.6642,
+      "step": 12030
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023352456277152722,
+      "loss": 0.6308,
+      "step": 12040
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023346886487690763,
+      "loss": 0.7566,
+      "step": 12050
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023341316698228805,
+      "loss": 0.6343,
+      "step": 12060
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023335746908766848,
+      "loss": 0.6906,
+      "step": 12070
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002333017711930489,
+      "loss": 0.5773,
+      "step": 12080
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002332460732984293,
+      "loss": 0.7105,
+      "step": 12090
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023319037540380972,
+      "loss": 0.7163,
+      "step": 12100
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002331402472986521,
+      "loss": 0.6924,
+      "step": 12110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023308454940403252,
+      "loss": 0.6674,
+      "step": 12120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023302885150941293,
+      "loss": 0.6818,
+      "step": 12130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023297315361479334,
+      "loss": 0.6678,
+      "step": 12140
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023291745572017375,
+      "loss": 0.7061,
+      "step": 12150
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023286175782555416,
+      "loss": 0.6423,
+      "step": 12160
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023280605993093457,
+      "loss": 0.698,
+      "step": 12170
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000232750362036315,
+      "loss": 0.6791,
+      "step": 12180
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023269466414169542,
+      "loss": 0.6042,
+      "step": 12190
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023263896624707583,
+      "loss": 0.6702,
+      "step": 12200
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023258326835245624,
+      "loss": 0.6973,
+      "step": 12210
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023252757045783665,
+      "loss": 0.6986,
+      "step": 12220
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023247187256321706,
+      "loss": 0.6119,
+      "step": 12230
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023241617466859753,
+      "loss": 0.7303,
+      "step": 12240
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023236047677397794,
+      "loss": 0.8158,
+      "step": 12250
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023230477887935835,
+      "loss": 0.625,
+      "step": 12260
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023224908098473876,
+      "loss": 0.675,
+      "step": 12270
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023219338309011917,
+      "loss": 0.6737,
+      "step": 12280
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023213768519549959,
+      "loss": 0.8781,
+      "step": 12290
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023208198730088,
+      "loss": 0.6374,
+      "step": 12300
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023202628940626044,
+      "loss": 0.6831,
+      "step": 12310
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023197059151164085,
+      "loss": 0.6503,
+      "step": 12320
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023191489361702126,
+      "loss": 0.607,
+      "step": 12330
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023185919572240167,
+      "loss": 0.6086,
+      "step": 12340
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023180349782778208,
+      "loss": 0.736,
+      "step": 12350
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0002317477999331625,
+      "loss": 0.6346,
+      "step": 12360
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023169210203854293,
+      "loss": 0.7117,
+      "step": 12370
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023163640414392334,
+      "loss": 0.6165,
+      "step": 12380
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023158070624930375,
+      "loss": 0.7097,
+      "step": 12390
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023152500835468416,
+      "loss": 0.7594,
+      "step": 12400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023146931046006457,
+      "loss": 0.662,
+      "step": 12410
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023141361256544498,
+      "loss": 0.6289,
+      "step": 12420
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0002313579146708254,
+      "loss": 0.7139,
+      "step": 12430
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023130221677620586,
+      "loss": 0.7751,
+      "step": 12440
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023124651888158627,
+      "loss": 0.7929,
+      "step": 12450
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023119082098696668,
+      "loss": 0.5422,
+      "step": 12460
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0002311351230923471,
+      "loss": 0.7644,
+      "step": 12470
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0002310794251977275,
+      "loss": 0.5648,
+      "step": 12480
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023102372730310792,
+      "loss": 0.715,
+      "step": 12490
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023096802940848836,
+      "loss": 0.8608,
+      "step": 12500
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023091233151386877,
+      "loss": 0.6886,
+      "step": 12510
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023085663361924918,
+      "loss": 0.5238,
+      "step": 12520
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0002308009357246296,
+      "loss": 0.7138,
+      "step": 12530
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023074523783001,
+      "loss": 0.6431,
+      "step": 12540
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002306895399353904,
+      "loss": 0.8476,
+      "step": 12550
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00023063384204077082,
+      "loss": 0.6158,
+      "step": 12560
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00023057814414615126,
+      "loss": 0.678,
+      "step": 12570
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00023052244625153167,
+      "loss": 0.7354,
+      "step": 12580
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00023046674835691208,
+      "loss": 0.775,
+      "step": 12590
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002304110504622925,
+      "loss": 0.724,
+      "step": 12600
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002303553525676729,
+      "loss": 0.5052,
+      "step": 12610
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00023029965467305332,
+      "loss": 0.7101,
+      "step": 12620
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00023024395677843378,
+      "loss": 0.7345,
+      "step": 12630
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002301882588838142,
+      "loss": 0.7054,
+      "step": 12640
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002301325609891946,
+      "loss": 0.7076,
+      "step": 12650
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00023007686309457501,
+      "loss": 0.6755,
+      "step": 12660
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00023002116519995543,
+      "loss": 0.6861,
+      "step": 12670
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022996546730533584,
+      "loss": 0.6241,
+      "step": 12680
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022990976941071628,
+      "loss": 0.8885,
+      "step": 12690
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022985407151609669,
+      "loss": 0.6576,
+      "step": 12700
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002297983736214771,
+      "loss": 0.6509,
+      "step": 12710
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002297426757268575,
+      "loss": 0.6496,
+      "step": 12720
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022968697783223792,
+      "loss": 0.6482,
+      "step": 12730
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022963127993761833,
+      "loss": 0.5704,
+      "step": 12740
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022957558204299874,
+      "loss": 0.6875,
+      "step": 12750
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022951988414837918,
+      "loss": 0.5182,
+      "step": 12760
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002294641862537596,
+      "loss": 0.603,
+      "step": 12770
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022940848835914,
+      "loss": 0.6699,
+      "step": 12780
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022935279046452041,
+      "loss": 0.9298,
+      "step": 12790
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022929709256990082,
+      "loss": 0.7229,
+      "step": 12800
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022924139467528124,
+      "loss": 0.7195,
+      "step": 12810
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022918569678066167,
+      "loss": 0.643,
+      "step": 12820
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022912999888604209,
+      "loss": 0.6373,
+      "step": 12830
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022907430099142252,
+      "loss": 0.8089,
+      "step": 12840
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022901860309680293,
+      "loss": 0.6898,
+      "step": 12850
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022896290520218335,
+      "loss": 0.7715,
+      "step": 12860
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022890720730756376,
+      "loss": 0.7475,
+      "step": 12870
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022885150941294417,
+      "loss": 0.6232,
+      "step": 12880
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002287958115183246,
+      "loss": 0.6331,
+      "step": 12890
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022874011362370502,
+      "loss": 0.6767,
+      "step": 12900
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022868441572908543,
+      "loss": 0.7686,
+      "step": 12910
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022862871783446584,
+      "loss": 0.6391,
+      "step": 12920
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022857301993984625,
+      "loss": 0.7066,
+      "step": 12930
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022851732204522666,
+      "loss": 0.6514,
+      "step": 12940
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002284616241506071,
+      "loss": 0.7157,
+      "step": 12950
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002284059262559875,
+      "loss": 0.6704,
+      "step": 12960
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022835579815082987,
+      "loss": 0.6819,
+      "step": 12970
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022830010025621028,
+      "loss": 0.7704,
+      "step": 12980
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022824440236159072,
+      "loss": 0.7403,
+      "step": 12990
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022818870446697113,
+      "loss": 0.716,
+      "step": 13000
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022813300657235154,
+      "loss": 0.5772,
+      "step": 13010
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022807730867773195,
+      "loss": 0.6466,
+      "step": 13020
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022802161078311236,
+      "loss": 0.5935,
+      "step": 13030
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022796591288849278,
+      "loss": 0.7362,
+      "step": 13040
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002279102149938732,
+      "loss": 0.7085,
+      "step": 13050
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022785451709925365,
+      "loss": 0.7303,
+      "step": 13060
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022779881920463406,
+      "loss": 0.6836,
+      "step": 13070
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022774312131001447,
+      "loss": 0.6263,
+      "step": 13080
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00022768742341539489,
+      "loss": 0.6741,
+      "step": 13090
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002276317255207753,
+      "loss": 0.6999,
+      "step": 13100
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002275760276261557,
+      "loss": 0.6771,
+      "step": 13110
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00022752032973153615,
+      "loss": 0.7091,
+      "step": 13120
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00022746463183691656,
+      "loss": 0.6451,
+      "step": 13130
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00022740893394229697,
+      "loss": 0.7333,
+      "step": 13140
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00022735323604767738,
+      "loss": 0.69,
+      "step": 13150
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002272975381530578,
+      "loss": 0.775,
+      "step": 13160
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002272418402584382,
+      "loss": 0.7927,
+      "step": 13170
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002271861423638186,
+      "loss": 0.7457,
+      "step": 13180
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00022713044446919905,
+      "loss": 0.7266,
+      "step": 13190
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00022707474657457946,
+      "loss": 0.5273,
+      "step": 13200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00022701904867995987,
+      "loss": 0.6495,
+      "step": 13210
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00022696335078534028,
+      "loss": 0.7933,
+      "step": 13220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002269076528907207,
+      "loss": 0.6314,
+      "step": 13230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002268519549961011,
+      "loss": 0.7547,
+      "step": 13240
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00022679625710148157,
+      "loss": 0.716,
+      "step": 13250
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00022674055920686198,
+      "loss": 0.8462,
+      "step": 13260
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002266848613122424,
+      "loss": 0.8108,
+      "step": 13270
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002266291634176228,
+      "loss": 0.5796,
+      "step": 13280
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00022657346552300322,
+      "loss": 0.6584,
+      "step": 13290
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00022651776762838363,
+      "loss": 0.7407,
+      "step": 13300
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00022646206973376404,
+      "loss": 0.6505,
+      "step": 13310
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00022640637183914448,
+      "loss": 0.6075,
+      "step": 13320
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002263506739445249,
+      "loss": 0.6258,
+      "step": 13330
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002262949760499053,
+      "loss": 0.5988,
+      "step": 13340
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002262392781552857,
+      "loss": 0.7777,
+      "step": 13350
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022618358026066612,
+      "loss": 0.763,
+      "step": 13360
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022612788236604653,
+      "loss": 0.7065,
+      "step": 13370
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022607218447142697,
+      "loss": 0.7195,
+      "step": 13380
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022601648657680738,
+      "loss": 0.5759,
+      "step": 13390
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002259607886821878,
+      "loss": 0.727,
+      "step": 13400
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002259050907875682,
+      "loss": 0.5891,
+      "step": 13410
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022584939289294862,
+      "loss": 0.565,
+      "step": 13420
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022579369499832903,
+      "loss": 0.655,
+      "step": 13430
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022573799710370944,
+      "loss": 0.7448,
+      "step": 13440
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022568229920908988,
+      "loss": 0.7091,
+      "step": 13450
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022562660131447031,
+      "loss": 0.6339,
+      "step": 13460
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022557090341985073,
+      "loss": 0.7307,
+      "step": 13470
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022551520552523114,
+      "loss": 0.6675,
+      "step": 13480
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022545950763061155,
+      "loss": 0.7426,
+      "step": 13490
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022540380973599196,
+      "loss": 0.7154,
+      "step": 13500
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002253481118413724,
+      "loss": 0.6243,
+      "step": 13510
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002252924139467528,
+      "loss": 0.5902,
+      "step": 13520
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022523671605213322,
+      "loss": 0.6257,
+      "step": 13530
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022518101815751363,
+      "loss": 0.6915,
+      "step": 13540
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022512532026289404,
+      "loss": 0.6206,
+      "step": 13550
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022506962236827445,
+      "loss": 0.5492,
+      "step": 13560
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022501392447365486,
+      "loss": 0.7209,
+      "step": 13570
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002249582265790353,
+      "loss": 0.6182,
+      "step": 13580
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002249025286844157,
+      "loss": 0.7192,
+      "step": 13590
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022484683078979612,
+      "loss": 0.7341,
+      "step": 13600
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022479113289517654,
+      "loss": 0.732,
+      "step": 13610
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00022473543500055695,
+      "loss": 0.6638,
+      "step": 13620
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022467973710593736,
+      "loss": 0.4943,
+      "step": 13630
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0002246240392113178,
+      "loss": 0.7328,
+      "step": 13640
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0002245683413166982,
+      "loss": 0.7483,
+      "step": 13650
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022451264342207862,
+      "loss": 0.6602,
+      "step": 13660
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022445694552745903,
+      "loss": 0.7906,
+      "step": 13670
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022440124763283944,
+      "loss": 0.6396,
+      "step": 13680
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022434554973821985,
+      "loss": 0.639,
+      "step": 13690
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0002242898518436003,
+      "loss": 0.6533,
+      "step": 13700
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022423415394898073,
+      "loss": 0.694,
+      "step": 13710
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022417845605436114,
+      "loss": 0.8023,
+      "step": 13720
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022412275815974155,
+      "loss": 0.6403,
+      "step": 13730
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022406706026512196,
+      "loss": 0.7514,
+      "step": 13740
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022401136237050237,
+      "loss": 0.7525,
+      "step": 13750
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022395566447588278,
+      "loss": 0.7266,
+      "step": 13760
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022389996658126322,
+      "loss": 0.6511,
+      "step": 13770
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022384426868664363,
+      "loss": 0.5995,
+      "step": 13780
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022378857079202404,
+      "loss": 0.6895,
+      "step": 13790
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022373287289740446,
+      "loss": 0.623,
+      "step": 13800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022367717500278487,
+      "loss": 0.6552,
+      "step": 13810
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022362147710816528,
+      "loss": 0.7724,
+      "step": 13820
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0002235657792135457,
+      "loss": 0.6573,
+      "step": 13830
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022351008131892613,
+      "loss": 0.656,
+      "step": 13840
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022345438342430654,
+      "loss": 0.6983,
+      "step": 13850
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022339868552968695,
+      "loss": 0.7186,
+      "step": 13860
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022334298763506736,
+      "loss": 0.7065,
+      "step": 13870
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022328728974044777,
+      "loss": 0.6109,
+      "step": 13880
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00022323159184582818,
+      "loss": 0.717,
+      "step": 13890
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00022317589395120865,
+      "loss": 0.6701,
+      "step": 13900
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00022312019605658906,
+      "loss": 0.6823,
+      "step": 13910
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00022306449816196947,
+      "loss": 0.4733,
+      "step": 13920
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00022300880026734988,
+      "loss": 0.773,
+      "step": 13930
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002229531023727303,
+      "loss": 0.6302,
+      "step": 13940
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002228974044781107,
+      "loss": 0.6834,
+      "step": 13950
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00022284170658349112,
+      "loss": 0.7129,
+      "step": 13960
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00022278600868887155,
+      "loss": 0.6505,
+      "step": 13970
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00022273031079425196,
+      "loss": 0.6433,
+      "step": 13980
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00022267461289963238,
+      "loss": 0.7738,
+      "step": 13990
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002226189150050128,
+      "loss": 0.6346,
+      "step": 14000
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002225632171103932,
+      "loss": 0.6472,
+      "step": 14010
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002225075192157736,
+      "loss": 0.8198,
+      "step": 14020
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00022245182132115405,
+      "loss": 0.7712,
+      "step": 14030
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00022239612342653446,
+      "loss": 0.614,
+      "step": 14040
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00022234042553191487,
+      "loss": 0.6847,
+      "step": 14050
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00022228472763729528,
+      "loss": 0.5931,
+      "step": 14060
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002222290297426757,
+      "loss": 0.5816,
+      "step": 14070
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002221733318480561,
+      "loss": 0.6594,
+      "step": 14080
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00022211763395343651,
+      "loss": 0.6311,
+      "step": 14090
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00022206193605881698,
+      "loss": 0.7299,
+      "step": 14100
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002220062381641974,
+      "loss": 0.6198,
+      "step": 14110
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002219505402695778,
+      "loss": 0.8008,
+      "step": 14120
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002218948423749582,
+      "loss": 0.6623,
+      "step": 14130
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00022183914448033862,
+      "loss": 0.6649,
+      "step": 14140
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00022178344658571904,
+      "loss": 0.6897,
+      "step": 14150
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00022172774869109947,
+      "loss": 0.5862,
+      "step": 14160
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00022167205079647988,
+      "loss": 0.6233,
+      "step": 14170
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002216163529018603,
+      "loss": 0.6554,
+      "step": 14180
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002215606550072407,
+      "loss": 0.6932,
+      "step": 14190
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00022150495711262112,
+      "loss": 0.719,
+      "step": 14200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00022144925921800153,
+      "loss": 0.6083,
+      "step": 14210
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00022139356132338194,
+      "loss": 0.6489,
+      "step": 14220
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00022133786342876238,
+      "loss": 0.6216,
+      "step": 14230
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002212821655341428,
+      "loss": 0.7405,
+      "step": 14240
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002212264676395232,
+      "loss": 0.6268,
+      "step": 14250
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002211707697449036,
+      "loss": 0.7664,
+      "step": 14260
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00022111507185028402,
+      "loss": 0.8237,
+      "step": 14270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00022105937395566443,
+      "loss": 0.7146,
+      "step": 14280
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002210036760610449,
+      "loss": 0.6834,
+      "step": 14290
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002209479781664253,
+      "loss": 0.7013,
+      "step": 14300
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00022089228027180572,
+      "loss": 0.6318,
+      "step": 14310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00022083658237718613,
+      "loss": 0.6377,
+      "step": 14320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00022078088448256654,
+      "loss": 0.6101,
+      "step": 14330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00022072518658794696,
+      "loss": 0.5967,
+      "step": 14340
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00022066948869332737,
+      "loss": 0.5797,
+      "step": 14350
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002206137907987078,
+      "loss": 0.7185,
+      "step": 14360
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00022055809290408822,
+      "loss": 0.6422,
+      "step": 14370
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00022050239500946863,
+      "loss": 0.5973,
+      "step": 14380
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00022044669711484904,
+      "loss": 0.7087,
+      "step": 14390
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00022039099922022945,
+      "loss": 0.7315,
+      "step": 14400
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00022033530132560986,
+      "loss": 0.5749,
+      "step": 14410
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002202796034309903,
+      "loss": 0.6827,
+      "step": 14420
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002202239055363707,
+      "loss": 0.6614,
+      "step": 14430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00022016820764175112,
+      "loss": 0.7314,
+      "step": 14440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00022011250974713153,
+      "loss": 0.7085,
+      "step": 14450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00022005681185251194,
+      "loss": 0.7075,
+      "step": 14460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00022000111395789235,
+      "loss": 0.659,
+      "step": 14470
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021994541606327277,
+      "loss": 0.753,
+      "step": 14480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021988971816865323,
+      "loss": 0.7349,
+      "step": 14490
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021983402027403364,
+      "loss": 0.7157,
+      "step": 14500
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021977832237941405,
+      "loss": 0.6433,
+      "step": 14510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021972262448479446,
+      "loss": 0.825,
+      "step": 14520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021966692659017488,
+      "loss": 0.7108,
+      "step": 14530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021961122869555529,
+      "loss": 0.7532,
+      "step": 14540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021955553080093572,
+      "loss": 0.5937,
+      "step": 14550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021949983290631614,
+      "loss": 0.7208,
+      "step": 14560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021944413501169655,
+      "loss": 0.6708,
+      "step": 14570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021938843711707696,
+      "loss": 0.7305,
+      "step": 14580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021933273922245737,
+      "loss": 0.6432,
+      "step": 14590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021927704132783778,
+      "loss": 0.7485,
+      "step": 14600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021922134343321822,
+      "loss": 0.6242,
+      "step": 14610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021916564553859863,
+      "loss": 0.6528,
+      "step": 14620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021910994764397904,
+      "loss": 0.693,
+      "step": 14630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021905424974935945,
+      "loss": 0.645,
+      "step": 14640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021899855185473986,
+      "loss": 0.8306,
+      "step": 14650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021894285396012027,
+      "loss": 0.657,
+      "step": 14660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021888715606550069,
+      "loss": 0.6826,
+      "step": 14670
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021883145817088112,
+      "loss": 0.6073,
+      "step": 14680
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021877576027626153,
+      "loss": 0.7809,
+      "step": 14690
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00021872006238164195,
+      "loss": 0.6255,
+      "step": 14700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00021866436448702236,
+      "loss": 0.4938,
+      "step": 14710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00021860866659240277,
+      "loss": 0.8137,
+      "step": 14720
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002185529686977832,
+      "loss": 0.7493,
+      "step": 14730
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00021849727080316364,
+      "loss": 0.789,
+      "step": 14740
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00021844157290854406,
+      "loss": 0.7218,
+      "step": 14750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00021838587501392447,
+      "loss": 0.7999,
+      "step": 14760
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00021833017711930488,
+      "loss": 0.6433,
+      "step": 14770
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002182744792246853,
+      "loss": 0.692,
+      "step": 14780
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002182187813300657,
+      "loss": 0.6873,
+      "step": 14790
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002181630834354461,
+      "loss": 0.7035,
+      "step": 14800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00021810738554082655,
+      "loss": 0.641,
+      "step": 14810
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00021805168764620696,
+      "loss": 0.6117,
+      "step": 14820
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00021799598975158737,
+      "loss": 0.6816,
+      "step": 14830
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00021794029185696778,
+      "loss": 0.643,
+      "step": 14840
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002178845939623482,
+      "loss": 0.6706,
+      "step": 14850
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002178288960677286,
+      "loss": 0.629,
+      "step": 14860
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00021777319817310904,
+      "loss": 0.6919,
+      "step": 14870
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00021771750027848945,
+      "loss": 0.6551,
+      "step": 14880
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00021766180238386987,
+      "loss": 0.8308,
+      "step": 14890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00021760610448925028,
+      "loss": 0.5696,
+      "step": 14900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002175504065946307,
+      "loss": 0.663,
+      "step": 14910
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002174947087000111,
+      "loss": 0.5983,
+      "step": 14920
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002174390108053915,
+      "loss": 0.6473,
+      "step": 14930
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00021738331291077198,
+      "loss": 0.6577,
+      "step": 14940
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002173276150161524,
+      "loss": 0.7108,
+      "step": 14950
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002172719171215328,
+      "loss": 0.6457,
+      "step": 14960
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002172162192269132,
+      "loss": 0.5915,
+      "step": 14970
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00021716052133229362,
+      "loss": 0.8125,
+      "step": 14980
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00021710482343767403,
+      "loss": 0.6549,
+      "step": 14990
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00021704912554305447,
+      "loss": 0.6892,
+      "step": 15000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00021699342764843488,
+      "loss": 0.5753,
+      "step": 15010
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002169377297538153,
+      "loss": 0.6894,
+      "step": 15020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002168820318591957,
+      "loss": 0.7092,
+      "step": 15030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00021682633396457611,
+      "loss": 0.6203,
+      "step": 15040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00021677063606995653,
+      "loss": 0.7342,
+      "step": 15050
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00021671493817533694,
+      "loss": 0.7339,
+      "step": 15060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00021665924028071737,
+      "loss": 0.6012,
+      "step": 15070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00021660354238609779,
+      "loss": 0.6904,
+      "step": 15080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002165478444914782,
+      "loss": 0.6205,
+      "step": 15090
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002164921465968586,
+      "loss": 0.5442,
+      "step": 15100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00021643644870223902,
+      "loss": 0.7029,
+      "step": 15110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00021638075080761943,
+      "loss": 0.7041,
+      "step": 15120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002163250529129999,
+      "loss": 0.7117,
+      "step": 15130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002162693550183803,
+      "loss": 0.6666,
+      "step": 15140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00021621365712376072,
+      "loss": 0.672,
+      "step": 15150
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00021615795922914113,
+      "loss": 0.6775,
+      "step": 15160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00021610226133452154,
+      "loss": 0.6723,
+      "step": 15170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00021604656343990195,
+      "loss": 0.6951,
+      "step": 15180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00021599086554528236,
+      "loss": 0.6959,
+      "step": 15190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002159351676506628,
+      "loss": 0.7746,
+      "step": 15200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002158794697560432,
+      "loss": 0.6634,
+      "step": 15210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00021582377186142362,
+      "loss": 0.6506,
+      "step": 15220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00021576807396680403,
+      "loss": 0.7388,
+      "step": 15230
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00021571237607218445,
+      "loss": 0.581,
+      "step": 15240
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021565667817756486,
+      "loss": 0.6209,
+      "step": 15250
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0002156009802829453,
+      "loss": 0.6746,
+      "step": 15260
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0002155452823883257,
+      "loss": 0.6239,
+      "step": 15270
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021548958449370612,
+      "loss": 0.7832,
+      "step": 15280
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021543388659908653,
+      "loss": 0.6331,
+      "step": 15290
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021537818870446694,
+      "loss": 0.6882,
+      "step": 15300
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021532249080984735,
+      "loss": 0.7722,
+      "step": 15310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021526679291522776,
+      "loss": 0.6875,
+      "step": 15320
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021521109502060823,
+      "loss": 0.5636,
+      "step": 15330
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021515539712598864,
+      "loss": 0.7001,
+      "step": 15340
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021509969923136905,
+      "loss": 0.5869,
+      "step": 15350
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021504400133674946,
+      "loss": 0.6365,
+      "step": 15360
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021499387323159185,
+      "loss": 0.7286,
+      "step": 15370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021493817533697226,
+      "loss": 0.6721,
+      "step": 15380
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021488247744235267,
+      "loss": 0.5865,
+      "step": 15390
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021482677954773308,
+      "loss": 0.6176,
+      "step": 15400
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0002147710816531135,
+      "loss": 0.756,
+      "step": 15410
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0002147153837584939,
+      "loss": 0.7346,
+      "step": 15420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021465968586387434,
+      "loss": 0.5988,
+      "step": 15430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021460398796925475,
+      "loss": 0.7183,
+      "step": 15440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021454829007463516,
+      "loss": 0.7797,
+      "step": 15450
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021449259218001557,
+      "loss": 0.593,
+      "step": 15460
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021443689428539598,
+      "loss": 0.6311,
+      "step": 15470
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0002143811963907764,
+      "loss": 0.5723,
+      "step": 15480
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0002143254984961568,
+      "loss": 0.7624,
+      "step": 15490
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021426980060153725,
+      "loss": 0.6886,
+      "step": 15500
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00021421410270691766,
+      "loss": 0.7613,
+      "step": 15510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00021415840481229807,
+      "loss": 0.7022,
+      "step": 15520
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00021410270691767848,
+      "loss": 0.6362,
+      "step": 15530
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002140470090230589,
+      "loss": 0.6311,
+      "step": 15540
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002139913111284393,
+      "loss": 0.6845,
+      "step": 15550
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00021393561323381977,
+      "loss": 0.5831,
+      "step": 15560
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00021387991533920018,
+      "loss": 0.6478,
+      "step": 15570
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002138242174445806,
+      "loss": 0.7938,
+      "step": 15580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000213768519549961,
+      "loss": 0.6994,
+      "step": 15590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002137128216553414,
+      "loss": 0.6496,
+      "step": 15600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00021365712376072182,
+      "loss": 0.7755,
+      "step": 15610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00021360142586610223,
+      "loss": 0.6232,
+      "step": 15620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00021354572797148267,
+      "loss": 0.7156,
+      "step": 15630
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00021349003007686308,
+      "loss": 0.8258,
+      "step": 15640
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002134343321822435,
+      "loss": 0.6521,
+      "step": 15650
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002133786342876239,
+      "loss": 0.6671,
+      "step": 15660
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00021332293639300432,
+      "loss": 0.6472,
+      "step": 15670
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00021326723849838473,
+      "loss": 0.7124,
+      "step": 15680
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00021321154060376517,
+      "loss": 0.4767,
+      "step": 15690
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00021315584270914558,
+      "loss": 0.5941,
+      "step": 15700
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000213100144814526,
+      "loss": 0.584,
+      "step": 15710
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002130444469199064,
+      "loss": 0.7129,
+      "step": 15720
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002129887490252868,
+      "loss": 0.6647,
+      "step": 15730
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00021293305113066722,
+      "loss": 0.6782,
+      "step": 15740
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00021287735323604763,
+      "loss": 0.7348,
+      "step": 15750
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002128216553414281,
+      "loss": 0.5956,
+      "step": 15760
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002127659574468085,
+      "loss": 0.7793,
+      "step": 15770
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00021271025955218892,
+      "loss": 0.7324,
+      "step": 15780
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021265456165756933,
+      "loss": 0.6776,
+      "step": 15790
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021259886376294974,
+      "loss": 0.6559,
+      "step": 15800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021254316586833015,
+      "loss": 0.6334,
+      "step": 15810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0002124874679737106,
+      "loss": 0.6819,
+      "step": 15820
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.000212431770079091,
+      "loss": 0.6798,
+      "step": 15830
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021237607218447141,
+      "loss": 0.5819,
+      "step": 15840
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021232037428985182,
+      "loss": 0.6284,
+      "step": 15850
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021226467639523224,
+      "loss": 0.8205,
+      "step": 15860
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021220897850061265,
+      "loss": 0.6568,
+      "step": 15870
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021215328060599306,
+      "loss": 0.6719,
+      "step": 15880
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0002120975827113735,
+      "loss": 0.6046,
+      "step": 15890
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0002120418848167539,
+      "loss": 0.6394,
+      "step": 15900
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021198618692213432,
+      "loss": 0.6688,
+      "step": 15910
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021193048902751473,
+      "loss": 0.5539,
+      "step": 15920
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021187479113289514,
+      "loss": 0.5964,
+      "step": 15930
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021181909323827555,
+      "loss": 0.6626,
+      "step": 15940
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021176339534365602,
+      "loss": 0.6718,
+      "step": 15950
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021170769744903643,
+      "loss": 0.6036,
+      "step": 15960
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021165199955441684,
+      "loss": 0.673,
+      "step": 15970
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021159630165979725,
+      "loss": 0.5856,
+      "step": 15980
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021154060376517766,
+      "loss": 0.6437,
+      "step": 15990
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021148490587055807,
+      "loss": 0.5944,
+      "step": 16000
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021142920797593848,
+      "loss": 0.7816,
+      "step": 16010
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021137351008131892,
+      "loss": 0.5892,
+      "step": 16020
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021131781218669933,
+      "loss": 0.591,
+      "step": 16030
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021126211429207974,
+      "loss": 0.6353,
+      "step": 16040
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00021120641639746016,
+      "loss": 0.6361,
+      "step": 16050
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00021115071850284057,
+      "loss": 0.7992,
+      "step": 16060
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00021109502060822098,
+      "loss": 0.6903,
+      "step": 16070
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00021103932271360142,
+      "loss": 0.6197,
+      "step": 16080
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00021098362481898183,
+      "loss": 0.7122,
+      "step": 16090
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00021092792692436224,
+      "loss": 0.5584,
+      "step": 16100
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00021087222902974265,
+      "loss": 0.637,
+      "step": 16110
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00021081653113512306,
+      "loss": 0.7129,
+      "step": 16120
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00021076083324050347,
+      "loss": 0.5697,
+      "step": 16130
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00021070513534588388,
+      "loss": 0.7463,
+      "step": 16140
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00021064943745126435,
+      "loss": 0.6301,
+      "step": 16150
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00021059373955664476,
+      "loss": 0.8484,
+      "step": 16160
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00021053804166202517,
+      "loss": 0.5577,
+      "step": 16170
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00021048234376740558,
+      "loss": 0.7574,
+      "step": 16180
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000210426645872786,
+      "loss": 0.6606,
+      "step": 16190
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0002103709479781664,
+      "loss": 0.7742,
+      "step": 16200
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00021031525008354684,
+      "loss": 0.7174,
+      "step": 16210
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00021025955218892725,
+      "loss": 0.6275,
+      "step": 16220
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00021020385429430766,
+      "loss": 0.7237,
+      "step": 16230
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00021014815639968808,
+      "loss": 0.7215,
+      "step": 16240
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0002100924585050685,
+      "loss": 0.5755,
+      "step": 16250
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0002100367606104489,
+      "loss": 0.5667,
+      "step": 16260
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0002099810627158293,
+      "loss": 0.6133,
+      "step": 16270
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00020992536482120975,
+      "loss": 0.7582,
+      "step": 16280
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00020986966692659016,
+      "loss": 0.6751,
+      "step": 16290
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00020981396903197057,
+      "loss": 0.6342,
+      "step": 16300
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00020975827113735098,
+      "loss": 0.6421,
+      "step": 16310
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0002097025732427314,
+      "loss": 0.6766,
+      "step": 16320
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002096468753481118,
+      "loss": 0.6065,
+      "step": 16330
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020959117745349224,
+      "loss": 0.7122,
+      "step": 16340
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020953547955887265,
+      "loss": 0.6652,
+      "step": 16350
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020947978166425306,
+      "loss": 0.7515,
+      "step": 16360
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020942408376963348,
+      "loss": 0.7357,
+      "step": 16370
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002093683858750139,
+      "loss": 0.7068,
+      "step": 16380
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020931268798039432,
+      "loss": 0.6725,
+      "step": 16390
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020925699008577474,
+      "loss": 0.6435,
+      "step": 16400
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020920129219115517,
+      "loss": 0.8076,
+      "step": 16410
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020914559429653558,
+      "loss": 0.6132,
+      "step": 16420
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.000209089896401916,
+      "loss": 0.5879,
+      "step": 16430
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002090341985072964,
+      "loss": 0.6298,
+      "step": 16440
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020897850061267682,
+      "loss": 0.6888,
+      "step": 16450
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020892280271805723,
+      "loss": 0.7021,
+      "step": 16460
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020886710482343767,
+      "loss": 0.7169,
+      "step": 16470
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020881140692881808,
+      "loss": 0.6699,
+      "step": 16480
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002087557090341985,
+      "loss": 0.5833,
+      "step": 16490
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002087000111395789,
+      "loss": 0.7712,
+      "step": 16500
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002086443132449593,
+      "loss": 0.6582,
+      "step": 16510
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020858861535033972,
+      "loss": 0.7592,
+      "step": 16520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020853291745572016,
+      "loss": 0.7448,
+      "step": 16530
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020847721956110057,
+      "loss": 0.7102,
+      "step": 16540
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020842152166648098,
+      "loss": 0.5721,
+      "step": 16550
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002083658237718614,
+      "loss": 0.6425,
+      "step": 16560
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002083101258772418,
+      "loss": 0.7187,
+      "step": 16570
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020825442798262222,
+      "loss": 0.5357,
+      "step": 16580
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020819873008800263,
+      "loss": 0.5917,
+      "step": 16590
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002081430321933831,
+      "loss": 0.6529,
+      "step": 16600
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002080873342987635,
+      "loss": 0.6148,
+      "step": 16610
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020803163640414392,
+      "loss": 0.7299,
+      "step": 16620
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020797593850952433,
+      "loss": 0.575,
+      "step": 16630
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020792024061490474,
+      "loss": 0.6312,
+      "step": 16640
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020786454272028515,
+      "loss": 0.7628,
+      "step": 16650
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002078088448256656,
+      "loss": 0.7716,
+      "step": 16660
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.000207753146931046,
+      "loss": 0.7695,
+      "step": 16670
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002076974490364264,
+      "loss": 0.5351,
+      "step": 16680
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020764175114180682,
+      "loss": 0.7956,
+      "step": 16690
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020758605324718723,
+      "loss": 0.7127,
+      "step": 16700
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020753035535256764,
+      "loss": 0.6495,
+      "step": 16710
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020747465745794805,
+      "loss": 0.6567,
+      "step": 16720
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002074189595633285,
+      "loss": 0.5979,
+      "step": 16730
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002073632616687089,
+      "loss": 0.6473,
+      "step": 16740
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020730756377408932,
+      "loss": 0.7699,
+      "step": 16750
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020725186587946973,
+      "loss": 0.6821,
+      "step": 16760
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020719616798485014,
+      "loss": 0.6146,
+      "step": 16770
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020714047009023055,
+      "loss": 0.6764,
+      "step": 16780
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020708477219561101,
+      "loss": 0.688,
+      "step": 16790
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020702907430099142,
+      "loss": 0.8138,
+      "step": 16800
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020697337640637184,
+      "loss": 0.747,
+      "step": 16810
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020691767851175225,
+      "loss": 0.4782,
+      "step": 16820
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020686198061713266,
+      "loss": 0.5576,
+      "step": 16830
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020680628272251307,
+      "loss": 0.6364,
+      "step": 16840
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020675058482789348,
+      "loss": 0.6803,
+      "step": 16850
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020669488693327392,
+      "loss": 0.7732,
+      "step": 16860
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020663918903865433,
+      "loss": 0.7448,
+      "step": 16870
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020658349114403474,
+      "loss": 0.6297,
+      "step": 16880
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020652779324941515,
+      "loss": 0.6941,
+      "step": 16890
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020647209535479556,
+      "loss": 0.6296,
+      "step": 16900
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020641639746017597,
+      "loss": 0.609,
+      "step": 16910
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0002063606995655564,
+      "loss": 0.6334,
+      "step": 16920
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020630500167093682,
+      "loss": 0.6894,
+      "step": 16930
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020624930377631724,
+      "loss": 0.666,
+      "step": 16940
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020619360588169765,
+      "loss": 0.6468,
+      "step": 16950
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020613790798707806,
+      "loss": 0.5107,
+      "step": 16960
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020608221009245847,
+      "loss": 0.698,
+      "step": 16970
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020602651219783888,
+      "loss": 0.7222,
+      "step": 16980
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020597081430321934,
+      "loss": 0.6228,
+      "step": 16990
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020591511640859976,
+      "loss": 0.604,
+      "step": 17000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020585941851398017,
+      "loss": 0.6368,
+      "step": 17010
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020580372061936058,
+      "loss": 0.7241,
+      "step": 17020
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.000205748022724741,
+      "loss": 0.66,
+      "step": 17030
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0002056923248301214,
+      "loss": 0.662,
+      "step": 17040
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020563662693550184,
+      "loss": 0.6612,
+      "step": 17050
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020558092904088225,
+      "loss": 0.7032,
+      "step": 17060
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020552523114626266,
+      "loss": 0.6303,
+      "step": 17070
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020546953325164307,
+      "loss": 0.6672,
+      "step": 17080
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020541383535702348,
+      "loss": 0.5975,
+      "step": 17090
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0002053581374624039,
+      "loss": 0.6077,
+      "step": 17100
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0002053024395677843,
+      "loss": 0.6935,
+      "step": 17110
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020524674167316474,
+      "loss": 0.5895,
+      "step": 17120
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020519104377854516,
+      "loss": 0.5868,
+      "step": 17130
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00020513534588392557,
+      "loss": 0.5879,
+      "step": 17140
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00020507964798930598,
+      "loss": 0.5866,
+      "step": 17150
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002050239500946864,
+      "loss": 0.6544,
+      "step": 17160
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002049682522000668,
+      "loss": 0.6674,
+      "step": 17170
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00020491255430544726,
+      "loss": 0.5829,
+      "step": 17180
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00020485685641082768,
+      "loss": 0.7002,
+      "step": 17190
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002048011585162081,
+      "loss": 0.6267,
+      "step": 17200
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002047454606215885,
+      "loss": 0.6676,
+      "step": 17210
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002046897627269689,
+      "loss": 0.6414,
+      "step": 17220
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00020463406483234932,
+      "loss": 0.6435,
+      "step": 17230
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00020457836693772973,
+      "loss": 0.6674,
+      "step": 17240
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00020452266904311017,
+      "loss": 0.7893,
+      "step": 17250
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00020446697114849058,
+      "loss": 0.5701,
+      "step": 17260
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.000204411273253871,
+      "loss": 0.537,
+      "step": 17270
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002043555753592514,
+      "loss": 0.4973,
+      "step": 17280
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00020429987746463181,
+      "loss": 0.7679,
+      "step": 17290
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00020424417957001223,
+      "loss": 0.6712,
+      "step": 17300
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00020418848167539266,
+      "loss": 0.7045,
+      "step": 17310
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00020413278378077308,
+      "loss": 0.6852,
+      "step": 17320
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00020407708588615349,
+      "loss": 0.6592,
+      "step": 17330
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002040213879915339,
+      "loss": 0.6403,
+      "step": 17340
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002039656900969143,
+      "loss": 0.7452,
+      "step": 17350
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00020390999220229472,
+      "loss": 0.6997,
+      "step": 17360
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00020385429430767513,
+      "loss": 0.6126,
+      "step": 17370
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00020379859641305557,
+      "loss": 0.7993,
+      "step": 17380
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00020374289851843598,
+      "loss": 0.7044,
+      "step": 17390
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002036872006238164,
+      "loss": 0.609,
+      "step": 17400
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020363150272919683,
+      "loss": 0.643,
+      "step": 17410
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020357580483457724,
+      "loss": 0.7644,
+      "step": 17420
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020352010693995765,
+      "loss": 0.7087,
+      "step": 17430
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002034644090453381,
+      "loss": 0.6256,
+      "step": 17440
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002034087111507185,
+      "loss": 0.5963,
+      "step": 17450
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002033530132560989,
+      "loss": 0.6397,
+      "step": 17460
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020329731536147932,
+      "loss": 0.682,
+      "step": 17470
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020324161746685973,
+      "loss": 0.7407,
+      "step": 17480
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020318591957224015,
+      "loss": 0.7645,
+      "step": 17490
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020313022167762056,
+      "loss": 0.6624,
+      "step": 17500
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.000203074523783001,
+      "loss": 0.5608,
+      "step": 17510
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002030188258883814,
+      "loss": 0.7884,
+      "step": 17520
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020296312799376182,
+      "loss": 0.6766,
+      "step": 17530
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020290743009914223,
+      "loss": 0.6252,
+      "step": 17540
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020285173220452264,
+      "loss": 0.7199,
+      "step": 17550
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020279603430990305,
+      "loss": 0.7059,
+      "step": 17560
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002027403364152835,
+      "loss": 0.6424,
+      "step": 17570
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002026846385206639,
+      "loss": 0.7056,
+      "step": 17580
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002026289406260443,
+      "loss": 0.5768,
+      "step": 17590
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020257324273142472,
+      "loss": 0.5203,
+      "step": 17600
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020251754483680513,
+      "loss": 0.6972,
+      "step": 17610
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020246184694218554,
+      "loss": 0.652,
+      "step": 17620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020240614904756596,
+      "loss": 0.5759,
+      "step": 17630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020235045115294642,
+      "loss": 0.6558,
+      "step": 17640
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020229475325832683,
+      "loss": 0.6341,
+      "step": 17650
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020223905536370724,
+      "loss": 0.7283,
+      "step": 17660
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00020218335746908765,
+      "loss": 0.7642,
+      "step": 17670
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020212765957446807,
+      "loss": 0.6286,
+      "step": 17680
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020207196167984848,
+      "loss": 0.6005,
+      "step": 17690
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020201626378522892,
+      "loss": 0.6777,
+      "step": 17700
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020196056589060933,
+      "loss": 0.5649,
+      "step": 17710
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020190486799598974,
+      "loss": 0.68,
+      "step": 17720
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020184917010137015,
+      "loss": 0.6041,
+      "step": 17730
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020179347220675056,
+      "loss": 0.6246,
+      "step": 17740
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020173777431213097,
+      "loss": 0.7433,
+      "step": 17750
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020168207641751138,
+      "loss": 0.5697,
+      "step": 17760
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020162637852289182,
+      "loss": 0.5997,
+      "step": 17770
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020157068062827223,
+      "loss": 0.6056,
+      "step": 17780
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020151498273365264,
+      "loss": 0.7026,
+      "step": 17790
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020145928483903305,
+      "loss": 0.6083,
+      "step": 17800
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020140358694441346,
+      "loss": 0.7213,
+      "step": 17810
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020134788904979388,
+      "loss": 0.629,
+      "step": 17820
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020129219115517434,
+      "loss": 0.7491,
+      "step": 17830
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020123649326055475,
+      "loss": 0.5696,
+      "step": 17840
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020118079536593516,
+      "loss": 0.7119,
+      "step": 17850
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020112509747131557,
+      "loss": 0.7255,
+      "step": 17860
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020106939957669599,
+      "loss": 0.5669,
+      "step": 17870
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002010137016820764,
+      "loss": 0.6848,
+      "step": 17880
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002009580037874568,
+      "loss": 0.5886,
+      "step": 17890
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020090230589283725,
+      "loss": 0.5726,
+      "step": 17900
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020084660799821766,
+      "loss": 0.6728,
+      "step": 17910
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020079091010359807,
+      "loss": 0.6245,
+      "step": 17920
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00020073521220897848,
+      "loss": 0.6743,
+      "step": 17930
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002006795143143589,
+      "loss": 0.5713,
+      "step": 17940
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002006238164197393,
+      "loss": 0.6733,
+      "step": 17950
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00020056811852511974,
+      "loss": 0.661,
+      "step": 17960
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00020051242063050015,
+      "loss": 0.6789,
+      "step": 17970
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00020045672273588056,
+      "loss": 0.5861,
+      "step": 17980
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00020040102484126097,
+      "loss": 0.7447,
+      "step": 17990
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00020034532694664138,
+      "loss": 0.7067,
+      "step": 18000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002002896290520218,
+      "loss": 0.6465,
+      "step": 18010
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002002339311574022,
+      "loss": 0.6069,
+      "step": 18020
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00020017823326278267,
+      "loss": 0.7092,
+      "step": 18030
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00020012253536816308,
+      "loss": 0.6169,
+      "step": 18040
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002000668374735435,
+      "loss": 0.5533,
+      "step": 18050
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002000111395789239,
+      "loss": 0.8475,
+      "step": 18060
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00019995544168430432,
+      "loss": 0.6603,
+      "step": 18070
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00019989974378968473,
+      "loss": 0.6336,
+      "step": 18080
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00019984404589506517,
+      "loss": 0.5824,
+      "step": 18090
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00019978834800044558,
+      "loss": 0.6861,
+      "step": 18100
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.000199732650105826,
+      "loss": 0.5162,
+      "step": 18110
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001996769522112064,
+      "loss": 0.6371,
+      "step": 18120
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001996212543165868,
+      "loss": 0.6672,
+      "step": 18130
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00019956555642196722,
+      "loss": 0.6438,
+      "step": 18140
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00019950985852734763,
+      "loss": 0.6769,
+      "step": 18150
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00019945416063272807,
+      "loss": 0.8908,
+      "step": 18160
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00019939846273810848,
+      "loss": 0.6725,
+      "step": 18170
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001993427648434889,
+      "loss": 0.6551,
+      "step": 18180
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001992870669488693,
+      "loss": 0.6015,
+      "step": 18190
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00019923136905424972,
+      "loss": 0.5302,
+      "step": 18200
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00019917567115963013,
+      "loss": 0.5743,
+      "step": 18210
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001991199732650106,
+      "loss": 0.6065,
+      "step": 18220
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.000199064275370391,
+      "loss": 0.7092,
+      "step": 18230
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00019900857747577141,
+      "loss": 0.7828,
+      "step": 18240
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00019895287958115183,
+      "loss": 0.6355,
+      "step": 18250
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00019889718168653224,
+      "loss": 0.6551,
+      "step": 18260
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00019884148379191265,
+      "loss": 0.5744,
+      "step": 18270
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00019878578589729306,
+      "loss": 0.653,
+      "step": 18280
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001987300880026735,
+      "loss": 0.7121,
+      "step": 18290
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001986743901080539,
+      "loss": 0.6119,
+      "step": 18300
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00019861869221343432,
+      "loss": 0.6687,
+      "step": 18310
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00019856299431881473,
+      "loss": 0.6533,
+      "step": 18320
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00019850729642419514,
+      "loss": 0.78,
+      "step": 18330
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00019845159852957555,
+      "loss": 0.6069,
+      "step": 18340
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.000198395900634956,
+      "loss": 0.7008,
+      "step": 18350
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001983402027403364,
+      "loss": 0.7084,
+      "step": 18360
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001982845048457168,
+      "loss": 0.63,
+      "step": 18370
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00019822880695109722,
+      "loss": 0.4814,
+      "step": 18380
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00019817310905647764,
+      "loss": 0.7997,
+      "step": 18390
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00019811741116185805,
+      "loss": 0.7286,
+      "step": 18400
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00019806171326723849,
+      "loss": 0.6538,
+      "step": 18410
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001980060153726189,
+      "loss": 0.582,
+      "step": 18420
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0001979503174779993,
+      "loss": 0.6606,
+      "step": 18430
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00019789461958337972,
+      "loss": 0.9235,
+      "step": 18440
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00019783892168876016,
+      "loss": 0.6075,
+      "step": 18450
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00019778322379414057,
+      "loss": 0.6675,
+      "step": 18460
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00019772752589952098,
+      "loss": 0.6643,
+      "step": 18470
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00019767182800490142,
+      "loss": 0.6605,
+      "step": 18480
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019761613011028183,
+      "loss": 0.5987,
+      "step": 18490
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019756043221566224,
+      "loss": 0.6143,
+      "step": 18500
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019750473432104265,
+      "loss": 0.6291,
+      "step": 18510
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019744903642642306,
+      "loss": 0.6206,
+      "step": 18520
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019739333853180347,
+      "loss": 0.6683,
+      "step": 18530
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001973376406371839,
+      "loss": 0.6322,
+      "step": 18540
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019728194274256432,
+      "loss": 0.6859,
+      "step": 18550
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019722624484794473,
+      "loss": 0.5951,
+      "step": 18560
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019717054695332514,
+      "loss": 0.6626,
+      "step": 18570
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019711484905870556,
+      "loss": 0.593,
+      "step": 18580
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019705915116408597,
+      "loss": 0.7337,
+      "step": 18590
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019700345326946638,
+      "loss": 0.6291,
+      "step": 18600
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019694775537484682,
+      "loss": 0.7644,
+      "step": 18610
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019689205748022723,
+      "loss": 0.6554,
+      "step": 18620
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019683635958560764,
+      "loss": 0.6219,
+      "step": 18630
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019678066169098805,
+      "loss": 0.621,
+      "step": 18640
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019672496379636846,
+      "loss": 0.7215,
+      "step": 18650
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019666926590174887,
+      "loss": 0.6493,
+      "step": 18660
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019661356800712934,
+      "loss": 0.625,
+      "step": 18670
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019655787011250975,
+      "loss": 0.6352,
+      "step": 18680
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019650217221789016,
+      "loss": 0.7213,
+      "step": 18690
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019644647432327057,
+      "loss": 0.7981,
+      "step": 18700
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019639077642865098,
+      "loss": 0.5422,
+      "step": 18710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001963350785340314,
+      "loss": 0.6416,
+      "step": 18720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001962793806394118,
+      "loss": 0.5892,
+      "step": 18730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019622368274479224,
+      "loss": 0.7087,
+      "step": 18740
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019616798485017265,
+      "loss": 0.7,
+      "step": 18750
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019611228695555306,
+      "loss": 0.6006,
+      "step": 18760
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019605658906093348,
+      "loss": 0.5555,
+      "step": 18770
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001960008911663139,
+      "loss": 0.7352,
+      "step": 18780
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001959451932716943,
+      "loss": 0.6628,
+      "step": 18790
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019588949537707474,
+      "loss": 0.7434,
+      "step": 18800
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019583379748245515,
+      "loss": 0.7914,
+      "step": 18810
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019577809958783556,
+      "loss": 0.6911,
+      "step": 18820
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019572240169321597,
+      "loss": 0.6312,
+      "step": 18830
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019566670379859638,
+      "loss": 0.6617,
+      "step": 18840
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001956110059039768,
+      "loss": 0.8041,
+      "step": 18850
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001955553080093572,
+      "loss": 0.5595,
+      "step": 18860
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019549961011473767,
+      "loss": 0.6595,
+      "step": 18870
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019544391222011808,
+      "loss": 0.6615,
+      "step": 18880
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001953882143254985,
+      "loss": 0.5824,
+      "step": 18890
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001953325164308789,
+      "loss": 0.847,
+      "step": 18900
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001952768185362593,
+      "loss": 0.7381,
+      "step": 18910
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019522112064163972,
+      "loss": 0.5843,
+      "step": 18920
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019516542274702016,
+      "loss": 0.5933,
+      "step": 18930
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019510972485240057,
+      "loss": 0.531,
+      "step": 18940
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019505402695778098,
+      "loss": 0.6608,
+      "step": 18950
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001949983290631614,
+      "loss": 0.5853,
+      "step": 18960
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001949426311685418,
+      "loss": 0.6104,
+      "step": 18970
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019488693327392222,
+      "loss": 0.586,
+      "step": 18980
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019483123537930263,
+      "loss": 0.8872,
+      "step": 18990
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019477553748468307,
+      "loss": 0.5051,
+      "step": 19000
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019471983959006348,
+      "loss": 0.583,
+      "step": 19010
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001946641416954439,
+      "loss": 0.627,
+      "step": 19020
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001946084438008243,
+      "loss": 0.6354,
+      "step": 19030
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001945527459062047,
+      "loss": 0.6424,
+      "step": 19040
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019449704801158512,
+      "loss": 0.6479,
+      "step": 19050
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001944413501169656,
+      "loss": 0.6363,
+      "step": 19060
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.000194385652222346,
+      "loss": 0.5707,
+      "step": 19070
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001943299543277264,
+      "loss": 0.6264,
+      "step": 19080
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019427425643310682,
+      "loss": 0.741,
+      "step": 19090
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019421855853848723,
+      "loss": 0.7087,
+      "step": 19100
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019416286064386764,
+      "loss": 0.6304,
+      "step": 19110
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019410716274924806,
+      "loss": 0.6409,
+      "step": 19120
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001940514648546285,
+      "loss": 0.5855,
+      "step": 19130
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001939957669600089,
+      "loss": 0.5732,
+      "step": 19140
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019394006906538932,
+      "loss": 0.5669,
+      "step": 19150
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019388437117076973,
+      "loss": 0.6845,
+      "step": 19160
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019382867327615014,
+      "loss": 0.6808,
+      "step": 19170
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019377297538153055,
+      "loss": 0.7629,
+      "step": 19180
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.000193717277486911,
+      "loss": 0.6542,
+      "step": 19190
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001936615795922914,
+      "loss": 0.7577,
+      "step": 19200
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0001936058816976718,
+      "loss": 0.5944,
+      "step": 19210
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019355018380305222,
+      "loss": 0.7429,
+      "step": 19220
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019349448590843263,
+      "loss": 0.604,
+      "step": 19230
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019343878801381304,
+      "loss": 0.5441,
+      "step": 19240
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019338309011919345,
+      "loss": 0.6282,
+      "step": 19250
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019332739222457392,
+      "loss": 0.651,
+      "step": 19260
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019327169432995433,
+      "loss": 0.8442,
+      "step": 19270
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019321599643533474,
+      "loss": 0.6684,
+      "step": 19280
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00019316029854071515,
+      "loss": 0.7902,
+      "step": 19290
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019310460064609556,
+      "loss": 0.672,
+      "step": 19300
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019304890275147598,
+      "loss": 0.6835,
+      "step": 19310
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001929932048568564,
+      "loss": 0.5991,
+      "step": 19320
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019293750696223682,
+      "loss": 0.6007,
+      "step": 19330
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019288180906761724,
+      "loss": 0.7035,
+      "step": 19340
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019282611117299765,
+      "loss": 0.7652,
+      "step": 19350
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019277041327837806,
+      "loss": 0.6993,
+      "step": 19360
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019271471538375847,
+      "loss": 0.6339,
+      "step": 19370
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019265901748913888,
+      "loss": 0.6306,
+      "step": 19380
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019260331959451932,
+      "loss": 0.7027,
+      "step": 19390
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019254762169989973,
+      "loss": 0.6746,
+      "step": 19400
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019249192380528014,
+      "loss": 0.6782,
+      "step": 19410
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019243622591066055,
+      "loss": 0.6163,
+      "step": 19420
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019238052801604096,
+      "loss": 0.6623,
+      "step": 19430
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019232483012142137,
+      "loss": 0.6736,
+      "step": 19440
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001922691322268018,
+      "loss": 0.7566,
+      "step": 19450
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019221343433218222,
+      "loss": 0.6878,
+      "step": 19460
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019215773643756263,
+      "loss": 0.605,
+      "step": 19470
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019210203854294307,
+      "loss": 0.5416,
+      "step": 19480
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019204634064832348,
+      "loss": 0.738,
+      "step": 19490
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001919906427537039,
+      "loss": 0.6201,
+      "step": 19500
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019194051464854628,
+      "loss": 0.7031,
+      "step": 19510
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001918848167539267,
+      "loss": 0.5377,
+      "step": 19520
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001918291188593071,
+      "loss": 0.755,
+      "step": 19530
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019177342096468751,
+      "loss": 0.77,
+      "step": 19540
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019171772307006793,
+      "loss": 0.7279,
+      "step": 19550
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00019166202517544836,
+      "loss": 0.7072,
+      "step": 19560
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019160632728082878,
+      "loss": 0.6377,
+      "step": 19570
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001915506293862092,
+      "loss": 0.5932,
+      "step": 19580
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001914949314915896,
+      "loss": 0.6916,
+      "step": 19590
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019143923359697,
+      "loss": 0.6103,
+      "step": 19600
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019138353570235042,
+      "loss": 0.7542,
+      "step": 19610
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019132783780773086,
+      "loss": 0.6895,
+      "step": 19620
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019127213991311127,
+      "loss": 0.6957,
+      "step": 19630
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019121644201849168,
+      "loss": 0.7112,
+      "step": 19640
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001911607441238721,
+      "loss": 0.6982,
+      "step": 19650
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001911050462292525,
+      "loss": 0.6008,
+      "step": 19660
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019104934833463291,
+      "loss": 0.5489,
+      "step": 19670
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019099365044001333,
+      "loss": 0.6644,
+      "step": 19680
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001909379525453938,
+      "loss": 0.6523,
+      "step": 19690
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001908822546507742,
+      "loss": 0.6632,
+      "step": 19700
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001908265567561546,
+      "loss": 0.7229,
+      "step": 19710
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019077085886153502,
+      "loss": 0.6739,
+      "step": 19720
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019071516096691543,
+      "loss": 0.6678,
+      "step": 19730
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019065946307229585,
+      "loss": 0.6929,
+      "step": 19740
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019060376517767628,
+      "loss": 0.6331,
+      "step": 19750
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001905480672830567,
+      "loss": 0.7414,
+      "step": 19760
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001904923693884371,
+      "loss": 0.7257,
+      "step": 19770
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019043667149381752,
+      "loss": 0.7176,
+      "step": 19780
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019038097359919793,
+      "loss": 0.6459,
+      "step": 19790
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019032527570457834,
+      "loss": 0.8041,
+      "step": 19800
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00019026957780995875,
+      "loss": 0.6608,
+      "step": 19810
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001902138799153392,
+      "loss": 0.6923,
+      "step": 19820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001901581820207196,
+      "loss": 0.6175,
+      "step": 19830
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001901024841261,
+      "loss": 0.6667,
+      "step": 19840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00019004678623148042,
+      "loss": 0.7444,
+      "step": 19850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018999108833686083,
+      "loss": 0.6638,
+      "step": 19860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018993539044224125,
+      "loss": 0.6069,
+      "step": 19870
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001898796925476217,
+      "loss": 0.6109,
+      "step": 19880
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018982399465300212,
+      "loss": 0.6163,
+      "step": 19890
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018976829675838253,
+      "loss": 0.6371,
+      "step": 19900
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018971259886376294,
+      "loss": 0.7007,
+      "step": 19910
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018965690096914335,
+      "loss": 0.6362,
+      "step": 19920
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018960120307452377,
+      "loss": 0.6207,
+      "step": 19930
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018954550517990418,
+      "loss": 0.7749,
+      "step": 19940
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018948980728528462,
+      "loss": 0.6498,
+      "step": 19950
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018943410939066503,
+      "loss": 0.605,
+      "step": 19960
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018937841149604544,
+      "loss": 0.702,
+      "step": 19970
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018932271360142585,
+      "loss": 0.5961,
+      "step": 19980
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018926701570680626,
+      "loss": 0.6399,
+      "step": 19990
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018921131781218667,
+      "loss": 0.5963,
+      "step": 20000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0001891556199175671,
+      "loss": 0.6984,
+      "step": 20010
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018909992202294752,
+      "loss": 0.6384,
+      "step": 20020
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018904422412832793,
+      "loss": 0.6585,
+      "step": 20030
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018898852623370834,
+      "loss": 0.6981,
+      "step": 20040
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018893282833908875,
+      "loss": 0.7814,
+      "step": 20050
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018887713044446917,
+      "loss": 0.6603,
+      "step": 20060
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018882143254984958,
+      "loss": 0.7115,
+      "step": 20070
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018876573465523001,
+      "loss": 0.6559,
+      "step": 20080
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018871003676061043,
+      "loss": 0.6861,
+      "step": 20090
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018865433886599086,
+      "loss": 0.6876,
+      "step": 20100
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018859864097137127,
+      "loss": 0.5808,
+      "step": 20110
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018854294307675169,
+      "loss": 0.6236,
+      "step": 20120
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001884872451821321,
+      "loss": 0.6091,
+      "step": 20130
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018843154728751254,
+      "loss": 0.655,
+      "step": 20140
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018837584939289295,
+      "loss": 0.7106,
+      "step": 20150
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018832015149827336,
+      "loss": 0.7765,
+      "step": 20160
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018826445360365377,
+      "loss": 0.5648,
+      "step": 20170
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018820875570903418,
+      "loss": 0.5975,
+      "step": 20180
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001881530578144146,
+      "loss": 0.6609,
+      "step": 20190
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.000188097359919795,
+      "loss": 0.7302,
+      "step": 20200
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018804166202517544,
+      "loss": 0.7204,
+      "step": 20210
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018798596413055585,
+      "loss": 0.8433,
+      "step": 20220
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018793026623593626,
+      "loss": 0.7414,
+      "step": 20230
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018787456834131667,
+      "loss": 0.6493,
+      "step": 20240
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018781887044669709,
+      "loss": 0.7527,
+      "step": 20250
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001877631725520775,
+      "loss": 0.5422,
+      "step": 20260
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018770747465745793,
+      "loss": 0.699,
+      "step": 20270
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018765177676283835,
+      "loss": 0.6425,
+      "step": 20280
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018759607886821876,
+      "loss": 0.6404,
+      "step": 20290
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018754038097359917,
+      "loss": 0.6666,
+      "step": 20300
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018748468307897958,
+      "loss": 0.6949,
+      "step": 20310
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018742898518436,
+      "loss": 0.5536,
+      "step": 20320
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018737328728974046,
+      "loss": 0.6066,
+      "step": 20330
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018731758939512087,
+      "loss": 0.6386,
+      "step": 20340
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018726189150050128,
+      "loss": 0.5856,
+      "step": 20350
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001872061936058817,
+      "loss": 0.6494,
+      "step": 20360
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001871504957112621,
+      "loss": 0.7225,
+      "step": 20370
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001870947978166425,
+      "loss": 0.6422,
+      "step": 20380
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018703909992202292,
+      "loss": 0.6051,
+      "step": 20390
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018698340202740336,
+      "loss": 0.7431,
+      "step": 20400
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018692770413278377,
+      "loss": 0.57,
+      "step": 20410
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018687200623816418,
+      "loss": 0.5635,
+      "step": 20420
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001868163083435446,
+      "loss": 0.5906,
+      "step": 20430
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.000186760610448925,
+      "loss": 0.6621,
+      "step": 20440
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018670491255430542,
+      "loss": 0.6276,
+      "step": 20450
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018664921465968585,
+      "loss": 0.6124,
+      "step": 20460
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018659351676506627,
+      "loss": 0.6845,
+      "step": 20470
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018653781887044668,
+      "loss": 0.7457,
+      "step": 20480
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001864821209758271,
+      "loss": 0.7103,
+      "step": 20490
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001864264230812075,
+      "loss": 0.7599,
+      "step": 20500
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001863707251865879,
+      "loss": 0.6047,
+      "step": 20510
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018631502729196832,
+      "loss": 0.7335,
+      "step": 20520
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001862593293973488,
+      "loss": 0.6653,
+      "step": 20530
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001862036315027292,
+      "loss": 0.7796,
+      "step": 20540
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001861479336081096,
+      "loss": 0.7679,
+      "step": 20550
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018609223571349002,
+      "loss": 0.5963,
+      "step": 20560
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018603653781887043,
+      "loss": 0.6968,
+      "step": 20570
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018598083992425084,
+      "loss": 0.6086,
+      "step": 20580
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018592514202963128,
+      "loss": 0.6239,
+      "step": 20590
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001858694441350117,
+      "loss": 0.6947,
+      "step": 20600
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0001858137462403921,
+      "loss": 0.6271,
+      "step": 20610
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018575804834577251,
+      "loss": 0.7061,
+      "step": 20620
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018570235045115293,
+      "loss": 0.6634,
+      "step": 20630
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018564665255653334,
+      "loss": 0.6456,
+      "step": 20640
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018559095466191375,
+      "loss": 0.8379,
+      "step": 20650
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018553525676729419,
+      "loss": 0.6626,
+      "step": 20660
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001854795588726746,
+      "loss": 0.6343,
+      "step": 20670
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.000185423860978055,
+      "loss": 0.6303,
+      "step": 20680
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018536816308343542,
+      "loss": 0.5683,
+      "step": 20690
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018531246518881583,
+      "loss": 0.6122,
+      "step": 20700
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018525676729419624,
+      "loss": 0.6551,
+      "step": 20710
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001852010693995767,
+      "loss": 0.6795,
+      "step": 20720
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018514537150495712,
+      "loss": 0.7068,
+      "step": 20730
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018508967361033753,
+      "loss": 0.5266,
+      "step": 20740
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018503397571571794,
+      "loss": 0.651,
+      "step": 20750
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018497827782109835,
+      "loss": 0.6797,
+      "step": 20760
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018492257992647876,
+      "loss": 0.6936,
+      "step": 20770
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018486688203185917,
+      "loss": 0.6737,
+      "step": 20780
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001848111841372396,
+      "loss": 0.7991,
+      "step": 20790
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018475548624262002,
+      "loss": 0.6172,
+      "step": 20800
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018469978834800043,
+      "loss": 0.676,
+      "step": 20810
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018464409045338085,
+      "loss": 0.6675,
+      "step": 20820
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018458839255876126,
+      "loss": 0.5879,
+      "step": 20830
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018453269466414167,
+      "loss": 0.6728,
+      "step": 20840
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001844769967695221,
+      "loss": 0.6116,
+      "step": 20850
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018442129887490252,
+      "loss": 0.5568,
+      "step": 20860
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018436560098028293,
+      "loss": 0.6798,
+      "step": 20870
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018430990308566334,
+      "loss": 0.662,
+      "step": 20880
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018425420519104375,
+      "loss": 0.6456,
+      "step": 20890
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018419850729642416,
+      "loss": 0.8552,
+      "step": 20900
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018414280940180457,
+      "loss": 0.7782,
+      "step": 20910
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018408711150718504,
+      "loss": 0.6797,
+      "step": 20920
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018403141361256545,
+      "loss": 0.6356,
+      "step": 20930
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018397571571794586,
+      "loss": 0.6465,
+      "step": 20940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018392001782332627,
+      "loss": 0.6044,
+      "step": 20950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018386431992870668,
+      "loss": 0.5418,
+      "step": 20960
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001838086220340871,
+      "loss": 0.6875,
+      "step": 20970
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018375292413946753,
+      "loss": 0.7343,
+      "step": 20980
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018369722624484794,
+      "loss": 0.5542,
+      "step": 20990
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018364152835022835,
+      "loss": 0.6504,
+      "step": 21000
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018358583045560877,
+      "loss": 0.6411,
+      "step": 21010
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018353013256098918,
+      "loss": 0.6451,
+      "step": 21020
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001834744346663696,
+      "loss": 0.6064,
+      "step": 21030
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018341873677175,
+      "loss": 0.6803,
+      "step": 21040
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018336303887713044,
+      "loss": 0.7642,
+      "step": 21050
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018330734098251085,
+      "loss": 0.6216,
+      "step": 21060
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018325164308789126,
+      "loss": 0.6669,
+      "step": 21070
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018319594519327167,
+      "loss": 0.5003,
+      "step": 21080
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018314024729865208,
+      "loss": 0.5968,
+      "step": 21090
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001830845494040325,
+      "loss": 0.7301,
+      "step": 21100
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018302885150941293,
+      "loss": 0.5535,
+      "step": 21110
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018297315361479334,
+      "loss": 0.6121,
+      "step": 21120
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018291745572017378,
+      "loss": 0.6282,
+      "step": 21130
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001828617578255542,
+      "loss": 0.576,
+      "step": 21140
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0001828060599309346,
+      "loss": 0.7058,
+      "step": 21150
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.000182750362036315,
+      "loss": 0.6843,
+      "step": 21160
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018269466414169542,
+      "loss": 0.638,
+      "step": 21170
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018263896624707586,
+      "loss": 0.5182,
+      "step": 21180
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018258326835245627,
+      "loss": 0.6797,
+      "step": 21190
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018252757045783669,
+      "loss": 0.752,
+      "step": 21200
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001824718725632171,
+      "loss": 0.6137,
+      "step": 21210
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001824161746685975,
+      "loss": 0.7475,
+      "step": 21220
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018236047677397792,
+      "loss": 0.8034,
+      "step": 21230
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018230477887935836,
+      "loss": 0.691,
+      "step": 21240
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018224908098473877,
+      "loss": 0.5071,
+      "step": 21250
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018219338309011918,
+      "loss": 0.6671,
+      "step": 21260
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001821376851954996,
+      "loss": 0.7274,
+      "step": 21270
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018208198730088,
+      "loss": 0.7533,
+      "step": 21280
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001820262894062604,
+      "loss": 0.6973,
+      "step": 21290
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018197059151164082,
+      "loss": 0.7688,
+      "step": 21300
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018191489361702126,
+      "loss": 0.693,
+      "step": 21310
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018185919572240167,
+      "loss": 0.6405,
+      "step": 21320
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018180349782778208,
+      "loss": 0.6902,
+      "step": 21330
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001817477999331625,
+      "loss": 0.661,
+      "step": 21340
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001816921020385429,
+      "loss": 0.6779,
+      "step": 21350
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018163640414392332,
+      "loss": 0.591,
+      "step": 21360
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018158070624930378,
+      "loss": 0.8125,
+      "step": 21370
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001815250083546842,
+      "loss": 0.6631,
+      "step": 21380
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001814693104600646,
+      "loss": 0.6675,
+      "step": 21390
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018141361256544502,
+      "loss": 0.7044,
+      "step": 21400
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018135791467082543,
+      "loss": 0.6833,
+      "step": 21410
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018130221677620584,
+      "loss": 0.626,
+      "step": 21420
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018124651888158625,
+      "loss": 0.6701,
+      "step": 21430
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001811908209869667,
+      "loss": 0.6441,
+      "step": 21440
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001811351230923471,
+      "loss": 0.5859,
+      "step": 21450
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001810794251977275,
+      "loss": 0.6164,
+      "step": 21460
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018102372730310792,
+      "loss": 0.6975,
+      "step": 21470
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018096802940848833,
+      "loss": 0.6003,
+      "step": 21480
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018091233151386874,
+      "loss": 0.6407,
+      "step": 21490
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018085663361924918,
+      "loss": 0.6791,
+      "step": 21500
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018080650551409154,
+      "loss": 0.8427,
+      "step": 21510
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018075080761947195,
+      "loss": 0.5963,
+      "step": 21520
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018069510972485236,
+      "loss": 0.611,
+      "step": 21530
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018063941183023283,
+      "loss": 0.6788,
+      "step": 21540
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018058371393561324,
+      "loss": 0.6973,
+      "step": 21550
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018052801604099365,
+      "loss": 0.6802,
+      "step": 21560
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018047231814637406,
+      "loss": 0.7667,
+      "step": 21570
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018041662025175447,
+      "loss": 0.6205,
+      "step": 21580
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018036092235713488,
+      "loss": 0.6302,
+      "step": 21590
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001803052244625153,
+      "loss": 0.5154,
+      "step": 21600
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018024952656789573,
+      "loss": 0.7663,
+      "step": 21610
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018019382867327614,
+      "loss": 0.6918,
+      "step": 21620
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018013813077865656,
+      "loss": 0.8157,
+      "step": 21630
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018008243288403697,
+      "loss": 0.761,
+      "step": 21640
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018002673498941738,
+      "loss": 0.6386,
+      "step": 21650
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001799710370947978,
+      "loss": 0.6069,
+      "step": 21660
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00017991533920017823,
+      "loss": 0.6493,
+      "step": 21670
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00017985964130555864,
+      "loss": 0.564,
+      "step": 21680
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00017980394341093905,
+      "loss": 0.5672,
+      "step": 21690
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00017974824551631946,
+      "loss": 0.5828,
+      "step": 21700
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00017969254762169987,
+      "loss": 0.6515,
+      "step": 21710
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00017963684972708028,
+      "loss": 0.5852,
+      "step": 21720
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001795811518324607,
+      "loss": 0.7617,
+      "step": 21730
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00017952545393784113,
+      "loss": 0.7882,
+      "step": 21740
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00017946975604322157,
+      "loss": 0.7354,
+      "step": 21750
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00017941405814860198,
+      "loss": 0.7405,
+      "step": 21760
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001793583602539824,
+      "loss": 0.6197,
+      "step": 21770
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001793026623593628,
+      "loss": 0.7285,
+      "step": 21780
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00017924696446474322,
+      "loss": 0.6673,
+      "step": 21790
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00017919126657012365,
+      "loss": 0.6731,
+      "step": 21800
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00017913556867550406,
+      "loss": 0.6656,
+      "step": 21810
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00017907987078088448,
+      "loss": 0.6673,
+      "step": 21820
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001790241728862649,
+      "loss": 0.7103,
+      "step": 21830
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001789684749916453,
+      "loss": 0.6246,
+      "step": 21840
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001789127770970257,
+      "loss": 0.7215,
+      "step": 21850
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00017885707920240612,
+      "loss": 0.6166,
+      "step": 21860
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00017880138130778656,
+      "loss": 0.6467,
+      "step": 21870
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00017874568341316697,
+      "loss": 0.6146,
+      "step": 21880
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00017868998551854738,
+      "loss": 0.6723,
+      "step": 21890
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001786342876239278,
+      "loss": 0.6375,
+      "step": 21900
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001785785897293082,
+      "loss": 0.6821,
+      "step": 21910
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00017852289183468861,
+      "loss": 0.6494,
+      "step": 21920
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00017846719394006905,
+      "loss": 0.6948,
+      "step": 21930
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00017841149604544946,
+      "loss": 0.5204,
+      "step": 21940
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00017835579815082987,
+      "loss": 0.6715,
+      "step": 21950
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00017830010025621029,
+      "loss": 0.6334,
+      "step": 21960
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001782444023615907,
+      "loss": 0.6761,
+      "step": 21970
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0001781887044669711,
+      "loss": 0.6914,
+      "step": 21980
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017813300657235155,
+      "loss": 0.6114,
+      "step": 21990
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017807730867773198,
+      "loss": 0.684,
+      "step": 22000
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001780216107831124,
+      "loss": 0.6062,
+      "step": 22010
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001779659128884928,
+      "loss": 0.6965,
+      "step": 22020
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017791021499387322,
+      "loss": 0.7334,
+      "step": 22030
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017785451709925363,
+      "loss": 0.7175,
+      "step": 22040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017779881920463404,
+      "loss": 0.7129,
+      "step": 22050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017774312131001448,
+      "loss": 0.625,
+      "step": 22060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001776874234153949,
+      "loss": 0.5337,
+      "step": 22070
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001776317255207753,
+      "loss": 0.7045,
+      "step": 22080
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001775760276261557,
+      "loss": 0.61,
+      "step": 22090
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017752032973153612,
+      "loss": 0.6923,
+      "step": 22100
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017746463183691653,
+      "loss": 0.6564,
+      "step": 22110
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017740893394229695,
+      "loss": 0.6906,
+      "step": 22120
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017735323604767738,
+      "loss": 0.6841,
+      "step": 22130
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001772975381530578,
+      "loss": 0.6599,
+      "step": 22140
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001772418402584382,
+      "loss": 0.7184,
+      "step": 22150
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017718614236381862,
+      "loss": 0.6033,
+      "step": 22160
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017713044446919903,
+      "loss": 0.638,
+      "step": 22170
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017707474657457944,
+      "loss": 0.6752,
+      "step": 22180
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001770190486799599,
+      "loss": 0.6036,
+      "step": 22190
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017696335078534032,
+      "loss": 0.6212,
+      "step": 22200
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017690765289072073,
+      "loss": 0.6125,
+      "step": 22210
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017685195499610114,
+      "loss": 0.7277,
+      "step": 22220
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017679625710148155,
+      "loss": 0.6884,
+      "step": 22230
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017674055920686196,
+      "loss": 0.6407,
+      "step": 22240
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0001766848613122424,
+      "loss": 0.5581,
+      "step": 22250
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001766291634176228,
+      "loss": 0.6857,
+      "step": 22260
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017657346552300322,
+      "loss": 0.6205,
+      "step": 22270
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017651776762838363,
+      "loss": 0.7017,
+      "step": 22280
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017646206973376404,
+      "loss": 0.6955,
+      "step": 22290
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017640637183914445,
+      "loss": 0.7789,
+      "step": 22300
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017635067394452487,
+      "loss": 0.7204,
+      "step": 22310
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001762949760499053,
+      "loss": 0.6703,
+      "step": 22320
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017623927815528571,
+      "loss": 0.6914,
+      "step": 22330
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017618358026066613,
+      "loss": 0.702,
+      "step": 22340
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017612788236604654,
+      "loss": 0.6555,
+      "step": 22350
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017607218447142695,
+      "loss": 0.6718,
+      "step": 22360
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017601648657680736,
+      "loss": 0.569,
+      "step": 22370
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017596078868218782,
+      "loss": 0.6394,
+      "step": 22380
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017590509078756824,
+      "loss": 0.8083,
+      "step": 22390
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017584939289294865,
+      "loss": 0.5963,
+      "step": 22400
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017579369499832906,
+      "loss": 0.6576,
+      "step": 22410
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017573799710370947,
+      "loss": 0.5232,
+      "step": 22420
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017568229920908988,
+      "loss": 0.6411,
+      "step": 22430
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0001756266013144703,
+      "loss": 0.6751,
+      "step": 22440
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017557090341985073,
+      "loss": 0.6556,
+      "step": 22450
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017551520552523114,
+      "loss": 0.6646,
+      "step": 22460
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017545950763061155,
+      "loss": 0.7813,
+      "step": 22470
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017540380973599196,
+      "loss": 0.5784,
+      "step": 22480
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017534811184137237,
+      "loss": 0.6503,
+      "step": 22490
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017529241394675279,
+      "loss": 0.5332,
+      "step": 22500
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017523671605213322,
+      "loss": 0.6944,
+      "step": 22510
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017518101815751363,
+      "loss": 0.6788,
+      "step": 22520
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017512532026289405,
+      "loss": 0.657,
+      "step": 22530
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017506962236827446,
+      "loss": 0.686,
+      "step": 22540
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017501392447365487,
+      "loss": 0.7483,
+      "step": 22550
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017495822657903528,
+      "loss": 0.7148,
+      "step": 22560
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001749025286844157,
+      "loss": 0.7035,
+      "step": 22570
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017484683078979616,
+      "loss": 0.6277,
+      "step": 22580
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017479113289517657,
+      "loss": 0.7663,
+      "step": 22590
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017473543500055698,
+      "loss": 0.6972,
+      "step": 22600
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001746797371059374,
+      "loss": 0.7526,
+      "step": 22610
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001746240392113178,
+      "loss": 0.5441,
+      "step": 22620
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001745683413166982,
+      "loss": 0.6928,
+      "step": 22630
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017451264342207865,
+      "loss": 0.6568,
+      "step": 22640
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017445694552745906,
+      "loss": 0.6503,
+      "step": 22650
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017440124763283947,
+      "loss": 0.6712,
+      "step": 22660
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017434554973821988,
+      "loss": 0.7726,
+      "step": 22670
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001742898518436003,
+      "loss": 0.7783,
+      "step": 22680
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001742341539489807,
+      "loss": 0.5867,
+      "step": 22690
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017417845605436112,
+      "loss": 0.652,
+      "step": 22700
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017412275815974155,
+      "loss": 0.6259,
+      "step": 22710
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017406706026512197,
+      "loss": 0.6372,
+      "step": 22720
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017401136237050238,
+      "loss": 0.7315,
+      "step": 22730
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001739556644758828,
+      "loss": 0.7145,
+      "step": 22740
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001738999665812632,
+      "loss": 0.692,
+      "step": 22750
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001738442686866436,
+      "loss": 0.5947,
+      "step": 22760
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017378857079202405,
+      "loss": 0.5777,
+      "step": 22770
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001737328728974045,
+      "loss": 0.5751,
+      "step": 22780
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0001736771750027849,
+      "loss": 0.6465,
+      "step": 22790
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001736214771081653,
+      "loss": 0.6318,
+      "step": 22800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017356577921354572,
+      "loss": 0.6551,
+      "step": 22810
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017351008131892613,
+      "loss": 0.5258,
+      "step": 22820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017345438342430654,
+      "loss": 0.6216,
+      "step": 22830
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017339868552968698,
+      "loss": 0.6446,
+      "step": 22840
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001733429876350674,
+      "loss": 0.6706,
+      "step": 22850
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001732872897404478,
+      "loss": 0.6963,
+      "step": 22860
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017323159184582821,
+      "loss": 0.5914,
+      "step": 22870
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017317589395120863,
+      "loss": 0.5853,
+      "step": 22880
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017312019605658904,
+      "loss": 0.6259,
+      "step": 22890
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017306449816196947,
+      "loss": 0.6524,
+      "step": 22900
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017300880026734989,
+      "loss": 0.8211,
+      "step": 22910
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001729531023727303,
+      "loss": 0.6335,
+      "step": 22920
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001728974044781107,
+      "loss": 0.679,
+      "step": 22930
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017284170658349112,
+      "loss": 0.6377,
+      "step": 22940
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017278600868887153,
+      "loss": 0.6331,
+      "step": 22950
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017273031079425194,
+      "loss": 0.6661,
+      "step": 22960
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017267461289963238,
+      "loss": 0.6344,
+      "step": 22970
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001726189150050128,
+      "loss": 0.522,
+      "step": 22980
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001725632171103932,
+      "loss": 0.6078,
+      "step": 22990
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001725075192157736,
+      "loss": 0.6613,
+      "step": 23000
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017245182132115402,
+      "loss": 0.7214,
+      "step": 23010
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017239612342653446,
+      "loss": 0.6207,
+      "step": 23020
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001723404255319149,
+      "loss": 0.587,
+      "step": 23030
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001722847276372953,
+      "loss": 0.6524,
+      "step": 23040
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017222902974267572,
+      "loss": 0.582,
+      "step": 23050
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00017217333184805613,
+      "loss": 0.6086,
+      "step": 23060
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017211763395343655,
+      "loss": 0.7461,
+      "step": 23070
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017206193605881696,
+      "loss": 0.6574,
+      "step": 23080
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017200623816419737,
+      "loss": 0.5291,
+      "step": 23090
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001719505402695778,
+      "loss": 0.622,
+      "step": 23100
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017189484237495822,
+      "loss": 0.7323,
+      "step": 23110
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017183914448033863,
+      "loss": 0.7087,
+      "step": 23120
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017178344658571904,
+      "loss": 0.8162,
+      "step": 23130
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017172774869109945,
+      "loss": 0.6663,
+      "step": 23140
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017167205079647986,
+      "loss": 0.6863,
+      "step": 23150
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001716163529018603,
+      "loss": 0.7078,
+      "step": 23160
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001715606550072407,
+      "loss": 0.5336,
+      "step": 23170
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017150495711262112,
+      "loss": 0.7072,
+      "step": 23180
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017144925921800153,
+      "loss": 0.6792,
+      "step": 23190
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017139356132338194,
+      "loss": 0.5563,
+      "step": 23200
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017133786342876236,
+      "loss": 0.7307,
+      "step": 23210
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017128216553414277,
+      "loss": 0.5978,
+      "step": 23220
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017122646763952323,
+      "loss": 0.7075,
+      "step": 23230
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017117076974490364,
+      "loss": 0.6717,
+      "step": 23240
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017111507185028405,
+      "loss": 0.783,
+      "step": 23250
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017105937395566447,
+      "loss": 0.6898,
+      "step": 23260
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017100367606104488,
+      "loss": 0.6023,
+      "step": 23270
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0001709479781664253,
+      "loss": 0.6497,
+      "step": 23280
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017089228027180573,
+      "loss": 0.6195,
+      "step": 23290
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017083658237718614,
+      "loss": 0.7395,
+      "step": 23300
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017078088448256655,
+      "loss": 0.6426,
+      "step": 23310
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017072518658794696,
+      "loss": 0.6224,
+      "step": 23320
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017066948869332737,
+      "loss": 0.6004,
+      "step": 23330
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017061379079870778,
+      "loss": 0.6231,
+      "step": 23340
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001705580929040882,
+      "loss": 0.6916,
+      "step": 23350
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017050239500946863,
+      "loss": 0.6701,
+      "step": 23360
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017044669711484904,
+      "loss": 0.7606,
+      "step": 23370
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017039099922022945,
+      "loss": 0.6962,
+      "step": 23380
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017033530132560986,
+      "loss": 0.5405,
+      "step": 23390
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017027960343099028,
+      "loss": 0.5927,
+      "step": 23400
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001702239055363707,
+      "loss": 0.7452,
+      "step": 23410
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017016820764175115,
+      "loss": 0.7691,
+      "step": 23420
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017011250974713156,
+      "loss": 0.7378,
+      "step": 23430
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017005681185251197,
+      "loss": 0.7659,
+      "step": 23440
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017000111395789239,
+      "loss": 0.5904,
+      "step": 23450
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001699454160632728,
+      "loss": 0.8135,
+      "step": 23460
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001698897181686532,
+      "loss": 0.5632,
+      "step": 23470
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00016983402027403362,
+      "loss": 0.6855,
+      "step": 23480
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00016977832237941406,
+      "loss": 0.6673,
+      "step": 23490
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00016972262448479447,
+      "loss": 0.572,
+      "step": 23500
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00016966692659017488,
+      "loss": 0.5296,
+      "step": 23510
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00016961679848501724,
+      "loss": 0.5319,
+      "step": 23520
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00016956110059039768,
+      "loss": 0.7523,
+      "step": 23530
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001695054026957781,
+      "loss": 0.685,
+      "step": 23540
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001694497048011585,
+      "loss": 0.6187,
+      "step": 23550
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001693940069065389,
+      "loss": 0.6582,
+      "step": 23560
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00016933830901191932,
+      "loss": 0.6512,
+      "step": 23570
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00016928261111729973,
+      "loss": 0.7045,
+      "step": 23580
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00016922691322268017,
+      "loss": 0.745,
+      "step": 23590
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00016917121532806058,
+      "loss": 0.6822,
+      "step": 23600
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.000169115517433441,
+      "loss": 0.5626,
+      "step": 23610
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001690598195388214,
+      "loss": 0.4967,
+      "step": 23620
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016900412164420182,
+      "loss": 0.7235,
+      "step": 23630
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016894842374958225,
+      "loss": 0.5501,
+      "step": 23640
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016889272585496266,
+      "loss": 0.7744,
+      "step": 23650
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001688370279603431,
+      "loss": 0.6494,
+      "step": 23660
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016878133006572351,
+      "loss": 0.6036,
+      "step": 23670
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016872563217110393,
+      "loss": 0.612,
+      "step": 23680
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016866993427648434,
+      "loss": 0.7653,
+      "step": 23690
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016861423638186475,
+      "loss": 0.7363,
+      "step": 23700
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016855853848724516,
+      "loss": 0.5293,
+      "step": 23710
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001685028405926256,
+      "loss": 0.5533,
+      "step": 23720
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.000168447142698006,
+      "loss": 0.7485,
+      "step": 23730
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016839144480338642,
+      "loss": 0.7341,
+      "step": 23740
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016833574690876683,
+      "loss": 0.5985,
+      "step": 23750
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016828004901414724,
+      "loss": 0.5604,
+      "step": 23760
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016822435111952765,
+      "loss": 0.5703,
+      "step": 23770
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016816865322490806,
+      "loss": 0.6245,
+      "step": 23780
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001681129553302885,
+      "loss": 0.6768,
+      "step": 23790
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001680572574356689,
+      "loss": 0.6663,
+      "step": 23800
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016800155954104932,
+      "loss": 0.6341,
+      "step": 23810
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016794586164642974,
+      "loss": 0.7778,
+      "step": 23820
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016789016375181015,
+      "loss": 0.612,
+      "step": 23830
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016783446585719056,
+      "loss": 0.652,
+      "step": 23840
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016777876796257102,
+      "loss": 0.5799,
+      "step": 23850
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016772307006795143,
+      "loss": 0.776,
+      "step": 23860
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00016766737217333185,
+      "loss": 0.7102,
+      "step": 23870
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016761167427871226,
+      "loss": 0.619,
+      "step": 23880
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016755597638409267,
+      "loss": 0.706,
+      "step": 23890
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016750027848947308,
+      "loss": 0.5741,
+      "step": 23900
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001674445805948535,
+      "loss": 0.6037,
+      "step": 23910
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016738888270023393,
+      "loss": 0.6974,
+      "step": 23920
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016733318480561434,
+      "loss": 0.6473,
+      "step": 23930
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016727748691099475,
+      "loss": 0.6176,
+      "step": 23940
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016722178901637516,
+      "loss": 0.7261,
+      "step": 23950
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016716609112175557,
+      "loss": 0.6298,
+      "step": 23960
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016711039322713598,
+      "loss": 0.685,
+      "step": 23970
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016705469533251642,
+      "loss": 0.6892,
+      "step": 23980
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016699899743789683,
+      "loss": 0.7442,
+      "step": 23990
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016694329954327724,
+      "loss": 0.52,
+      "step": 24000
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016688760164865766,
+      "loss": 0.6932,
+      "step": 24010
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016683190375403807,
+      "loss": 0.6591,
+      "step": 24020
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016677620585941848,
+      "loss": 0.6619,
+      "step": 24030
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001667205079647989,
+      "loss": 0.5335,
+      "step": 24040
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016666481007017935,
+      "loss": 0.6035,
+      "step": 24050
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016660911217555977,
+      "loss": 0.656,
+      "step": 24060
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016655341428094018,
+      "loss": 0.6138,
+      "step": 24070
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001664977163863206,
+      "loss": 0.6546,
+      "step": 24080
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.000166442018491701,
+      "loss": 0.698,
+      "step": 24090
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0001663863205970814,
+      "loss": 0.6264,
+      "step": 24100
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016633062270246185,
+      "loss": 0.626,
+      "step": 24110
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016627492480784226,
+      "loss": 0.6154,
+      "step": 24120
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016621922691322267,
+      "loss": 0.6858,
+      "step": 24130
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00016616352901860308,
+      "loss": 0.5447,
+      "step": 24140
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001661078311239835,
+      "loss": 0.7641,
+      "step": 24150
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001660521332293639,
+      "loss": 0.8559,
+      "step": 24160
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00016599643533474434,
+      "loss": 0.5974,
+      "step": 24170
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00016594073744012475,
+      "loss": 0.6061,
+      "step": 24180
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00016588503954550516,
+      "loss": 0.646,
+      "step": 24190
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00016582934165088558,
+      "loss": 0.7893,
+      "step": 24200
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.000165773643756266,
+      "loss": 0.6778,
+      "step": 24210
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001657179458616464,
+      "loss": 0.6634,
+      "step": 24220
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001656622479670268,
+      "loss": 0.6203,
+      "step": 24230
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00016560655007240727,
+      "loss": 0.542,
+      "step": 24240
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00016555085217778769,
+      "loss": 0.5076,
+      "step": 24250
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001654951542831681,
+      "loss": 0.7691,
+      "step": 24260
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001654394563885485,
+      "loss": 0.5768,
+      "step": 24270
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00016538375849392892,
+      "loss": 0.8354,
+      "step": 24280
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00016532806059930933,
+      "loss": 0.6833,
+      "step": 24290
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00016527236270468977,
+      "loss": 0.6914,
+      "step": 24300
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00016521666481007018,
+      "loss": 0.7365,
+      "step": 24310
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001651609669154506,
+      "loss": 0.581,
+      "step": 24320
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.000165105269020831,
+      "loss": 0.52,
+      "step": 24330
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001650495711262114,
+      "loss": 0.7986,
+      "step": 24340
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00016499387323159182,
+      "loss": 0.6155,
+      "step": 24350
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00016493817533697223,
+      "loss": 0.8769,
+      "step": 24360
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00016488247744235267,
+      "loss": 0.5053,
+      "step": 24370
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00016482677954773308,
+      "loss": 0.6034,
+      "step": 24380
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001647710816531135,
+      "loss": 0.6345,
+      "step": 24390
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0001647153837584939,
+      "loss": 0.4832,
+      "step": 24400
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00016465968586387432,
+      "loss": 0.634,
+      "step": 24410
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00016460398796925473,
+      "loss": 0.5884,
+      "step": 24420
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001645482900746352,
+      "loss": 0.6038,
+      "step": 24430
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001644925921800156,
+      "loss": 0.6318,
+      "step": 24440
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00016443689428539602,
+      "loss": 0.6705,
+      "step": 24450
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00016438119639077643,
+      "loss": 0.7112,
+      "step": 24460
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00016432549849615684,
+      "loss": 0.5746,
+      "step": 24470
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00016426980060153725,
+      "loss": 0.6422,
+      "step": 24480
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00016421410270691766,
+      "loss": 0.5374,
+      "step": 24490
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001641584048122981,
+      "loss": 0.7395,
+      "step": 24500
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001641027069176785,
+      "loss": 0.6743,
+      "step": 24510
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00016404700902305892,
+      "loss": 0.6367,
+      "step": 24520
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00016399131112843933,
+      "loss": 0.6077,
+      "step": 24530
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00016393561323381974,
+      "loss": 0.6965,
+      "step": 24540
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00016387991533920015,
+      "loss": 0.8226,
+      "step": 24550
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001638242174445806,
+      "loss": 0.685,
+      "step": 24560
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.000163768519549961,
+      "loss": 0.6093,
+      "step": 24570
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00016371282165534142,
+      "loss": 0.7029,
+      "step": 24580
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00016365712376072183,
+      "loss": 0.7415,
+      "step": 24590
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00016360142586610224,
+      "loss": 0.7158,
+      "step": 24600
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00016354572797148265,
+      "loss": 0.7229,
+      "step": 24610
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00016349003007686306,
+      "loss": 0.6929,
+      "step": 24620
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001634343321822435,
+      "loss": 0.7338,
+      "step": 24630
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001633786342876239,
+      "loss": 0.7115,
+      "step": 24640
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00016332293639300432,
+      "loss": 0.5188,
+      "step": 24650
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00016326723849838473,
+      "loss": 0.6324,
+      "step": 24660
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00016321154060376517,
+      "loss": 0.5535,
+      "step": 24670
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00016315584270914558,
+      "loss": 0.591,
+      "step": 24680
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016310014481452602,
+      "loss": 0.6096,
+      "step": 24690
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016304444691990643,
+      "loss": 0.6553,
+      "step": 24700
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016298874902528684,
+      "loss": 0.6191,
+      "step": 24710
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016293305113066725,
+      "loss": 0.5887,
+      "step": 24720
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016287735323604766,
+      "loss": 0.672,
+      "step": 24730
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016282165534142807,
+      "loss": 0.5524,
+      "step": 24740
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016276595744680849,
+      "loss": 0.5515,
+      "step": 24750
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016271025955218892,
+      "loss": 0.6581,
+      "step": 24760
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016265456165756934,
+      "loss": 0.7206,
+      "step": 24770
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016259886376294975,
+      "loss": 0.554,
+      "step": 24780
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016254316586833016,
+      "loss": 0.7198,
+      "step": 24790
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016248746797371057,
+      "loss": 0.6821,
+      "step": 24800
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016243177007909098,
+      "loss": 0.7164,
+      "step": 24810
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016237607218447142,
+      "loss": 0.6463,
+      "step": 24820
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016232037428985183,
+      "loss": 0.6428,
+      "step": 24830
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016226467639523224,
+      "loss": 0.5694,
+      "step": 24840
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016220897850061265,
+      "loss": 0.5452,
+      "step": 24850
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016215328060599306,
+      "loss": 0.6166,
+      "step": 24860
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016209758271137347,
+      "loss": 0.5327,
+      "step": 24870
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016204188481675388,
+      "loss": 0.6611,
+      "step": 24880
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016198618692213435,
+      "loss": 0.5479,
+      "step": 24890
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016193048902751476,
+      "loss": 0.5646,
+      "step": 24900
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016187479113289517,
+      "loss": 0.6692,
+      "step": 24910
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016181909323827558,
+      "loss": 0.7033,
+      "step": 24920
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.000161763395343656,
+      "loss": 0.7868,
+      "step": 24930
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001617076974490364,
+      "loss": 0.6833,
+      "step": 24940
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00016165199955441684,
+      "loss": 0.7623,
+      "step": 24950
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00016159630165979726,
+      "loss": 0.662,
+      "step": 24960
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00016154060376517767,
+      "loss": 0.5462,
+      "step": 24970
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00016148490587055808,
+      "loss": 0.717,
+      "step": 24980
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001614292079759385,
+      "loss": 0.5175,
+      "step": 24990
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001613735100813189,
+      "loss": 0.7492,
+      "step": 25000
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001613178121866993,
+      "loss": 0.6851,
+      "step": 25010
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00016126211429207975,
+      "loss": 0.643,
+      "step": 25020
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00016120641639746016,
+      "loss": 0.7349,
+      "step": 25030
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00016115071850284057,
+      "loss": 0.6396,
+      "step": 25040
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00016109502060822098,
+      "loss": 0.6984,
+      "step": 25050
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001610393227136014,
+      "loss": 0.5591,
+      "step": 25060
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001609836248189818,
+      "loss": 0.6068,
+      "step": 25070
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00016092792692436227,
+      "loss": 0.7693,
+      "step": 25080
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00016087222902974268,
+      "loss": 0.7127,
+      "step": 25090
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001608165311351231,
+      "loss": 0.7093,
+      "step": 25100
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001607608332405035,
+      "loss": 0.6731,
+      "step": 25110
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00016070513534588391,
+      "loss": 0.678,
+      "step": 25120
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00016064943745126433,
+      "loss": 0.7756,
+      "step": 25130
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00016059373955664474,
+      "loss": 0.7264,
+      "step": 25140
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00016053804166202518,
+      "loss": 0.8083,
+      "step": 25150
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001604823437674056,
+      "loss": 0.6405,
+      "step": 25160
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.000160426645872786,
+      "loss": 0.6652,
+      "step": 25170
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0001603709479781664,
+      "loss": 0.6089,
+      "step": 25180
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00016031525008354682,
+      "loss": 0.5786,
+      "step": 25190
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00016025955218892723,
+      "loss": 0.6221,
+      "step": 25200
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00016020385429430767,
+      "loss": 0.6511,
+      "step": 25210
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00016014815639968808,
+      "loss": 0.6086,
+      "step": 25220
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001600924585050685,
+      "loss": 0.584,
+      "step": 25230
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001600367606104489,
+      "loss": 0.631,
+      "step": 25240
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00015998106271582931,
+      "loss": 0.6757,
+      "step": 25250
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00015992536482120972,
+      "loss": 0.6999,
+      "step": 25260
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00015986966692659014,
+      "loss": 0.5917,
+      "step": 25270
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001598139690319706,
+      "loss": 0.5864,
+      "step": 25280
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.000159758271137351,
+      "loss": 0.6321,
+      "step": 25290
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00015970257324273142,
+      "loss": 0.7112,
+      "step": 25300
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00015964687534811183,
+      "loss": 0.5841,
+      "step": 25310
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00015959117745349225,
+      "loss": 0.6637,
+      "step": 25320
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00015953547955887266,
+      "loss": 0.5794,
+      "step": 25330
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001594797816642531,
+      "loss": 0.712,
+      "step": 25340
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001594240837696335,
+      "loss": 0.5649,
+      "step": 25350
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00015936838587501392,
+      "loss": 0.6051,
+      "step": 25360
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00015931268798039433,
+      "loss": 0.6621,
+      "step": 25370
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00015925699008577474,
+      "loss": 0.6029,
+      "step": 25380
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00015920129219115515,
+      "loss": 0.6278,
+      "step": 25390
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00015914559429653556,
+      "loss": 0.5972,
+      "step": 25400
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.000159089896401916,
+      "loss": 0.5629,
+      "step": 25410
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001590341985072964,
+      "loss": 0.5566,
+      "step": 25420
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00015897850061267682,
+      "loss": 0.7261,
+      "step": 25430
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00015892280271805723,
+      "loss": 0.6248,
+      "step": 25440
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00015886710482343764,
+      "loss": 0.7223,
+      "step": 25450
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00015881140692881806,
+      "loss": 0.7283,
+      "step": 25460
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00015875570903419852,
+      "loss": 0.6315,
+      "step": 25470
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00015870001113957893,
+      "loss": 0.5885,
+      "step": 25480
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00015864431324495934,
+      "loss": 0.665,
+      "step": 25490
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015858861535033975,
+      "loss": 0.4955,
+      "step": 25500
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015853291745572017,
+      "loss": 0.6782,
+      "step": 25510
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015847721956110058,
+      "loss": 0.5667,
+      "step": 25520
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.000158421521666481,
+      "loss": 0.6292,
+      "step": 25530
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015836582377186143,
+      "loss": 0.664,
+      "step": 25540
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015831012587724184,
+      "loss": 0.6065,
+      "step": 25550
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015825442798262225,
+      "loss": 0.723,
+      "step": 25560
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015819873008800266,
+      "loss": 0.6907,
+      "step": 25570
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015814303219338307,
+      "loss": 0.7137,
+      "step": 25580
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015808733429876348,
+      "loss": 0.7402,
+      "step": 25590
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015803163640414392,
+      "loss": 0.593,
+      "step": 25600
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015797593850952433,
+      "loss": 0.6021,
+      "step": 25610
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015792024061490474,
+      "loss": 0.6045,
+      "step": 25620
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015786454272028515,
+      "loss": 0.6405,
+      "step": 25630
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015780884482566556,
+      "loss": 0.7044,
+      "step": 25640
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015775314693104598,
+      "loss": 0.5941,
+      "step": 25650
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001576974490364264,
+      "loss": 0.5414,
+      "step": 25660
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015764175114180683,
+      "loss": 0.648,
+      "step": 25670
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015758605324718724,
+      "loss": 0.5308,
+      "step": 25680
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001575359251420296,
+      "loss": 0.8367,
+      "step": 25690
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015748022724741,
+      "loss": 0.6401,
+      "step": 25700
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015742452935279047,
+      "loss": 0.6349,
+      "step": 25710
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015736883145817088,
+      "loss": 0.8038,
+      "step": 25720
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001573131335635513,
+      "loss": 0.541,
+      "step": 25730
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0001572574356689317,
+      "loss": 0.4583,
+      "step": 25740
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015720173777431212,
+      "loss": 0.7132,
+      "step": 25750
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00015714603987969253,
+      "loss": 0.71,
+      "step": 25760
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00015709034198507297,
+      "loss": 0.7718,
+      "step": 25770
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00015703464409045338,
+      "loss": 0.7438,
+      "step": 25780
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001569789461958338,
+      "loss": 0.6256,
+      "step": 25790
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001569232483012142,
+      "loss": 0.7839,
+      "step": 25800
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001568675504065946,
+      "loss": 0.6743,
+      "step": 25810
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00015681185251197502,
+      "loss": 0.5459,
+      "step": 25820
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00015675615461735543,
+      "loss": 0.6577,
+      "step": 25830
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00015670045672273587,
+      "loss": 0.7282,
+      "step": 25840
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00015664475882811628,
+      "loss": 0.6497,
+      "step": 25850
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001565890609334967,
+      "loss": 0.5238,
+      "step": 25860
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001565333630388771,
+      "loss": 0.7108,
+      "step": 25870
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00015647766514425752,
+      "loss": 0.6008,
+      "step": 25880
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00015642196724963793,
+      "loss": 0.6701,
+      "step": 25890
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001563662693550184,
+      "loss": 0.6457,
+      "step": 25900
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001563105714603988,
+      "loss": 0.5937,
+      "step": 25910
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00015625487356577921,
+      "loss": 0.6476,
+      "step": 25920
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00015619917567115963,
+      "loss": 0.7736,
+      "step": 25930
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00015614347777654004,
+      "loss": 0.5294,
+      "step": 25940
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00015608777988192045,
+      "loss": 0.7819,
+      "step": 25950
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00015603208198730086,
+      "loss": 0.7433,
+      "step": 25960
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001559763840926813,
+      "loss": 0.541,
+      "step": 25970
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001559206861980617,
+      "loss": 0.6122,
+      "step": 25980
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00015586498830344212,
+      "loss": 0.7083,
+      "step": 25990
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00015580929040882253,
+      "loss": 0.6321,
+      "step": 26000
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00015575359251420294,
+      "loss": 0.6895,
+      "step": 26010
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00015569789461958335,
+      "loss": 0.6287,
+      "step": 26020
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001556421967249638,
+      "loss": 0.6067,
+      "step": 26030
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001555864988303442,
+      "loss": 0.7166,
+      "step": 26040
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001555308009357246,
+      "loss": 0.6167,
+      "step": 26050
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015547510304110502,
+      "loss": 0.7379,
+      "step": 26060
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015541940514648544,
+      "loss": 0.5657,
+      "step": 26070
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015536370725186585,
+      "loss": 0.6784,
+      "step": 26080
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001553080093572463,
+      "loss": 0.6877,
+      "step": 26090
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015525231146262672,
+      "loss": 0.9396,
+      "step": 26100
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015519661356800713,
+      "loss": 0.7001,
+      "step": 26110
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015514091567338755,
+      "loss": 0.6222,
+      "step": 26120
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015508521777876796,
+      "loss": 0.5846,
+      "step": 26130
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015502951988414837,
+      "loss": 0.5977,
+      "step": 26140
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015497382198952878,
+      "loss": 0.6845,
+      "step": 26150
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015491812409490922,
+      "loss": 0.623,
+      "step": 26160
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015486242620028963,
+      "loss": 0.6241,
+      "step": 26170
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015480672830567004,
+      "loss": 0.7736,
+      "step": 26180
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015475103041105045,
+      "loss": 0.6809,
+      "step": 26190
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015469533251643086,
+      "loss": 0.6846,
+      "step": 26200
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015463963462181127,
+      "loss": 0.562,
+      "step": 26210
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0001545839367271917,
+      "loss": 0.6656,
+      "step": 26220
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015452823883257212,
+      "loss": 0.5681,
+      "step": 26230
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015447254093795253,
+      "loss": 0.749,
+      "step": 26240
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015441684304333294,
+      "loss": 0.72,
+      "step": 26250
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015436114514871336,
+      "loss": 0.7233,
+      "step": 26260
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015430544725409377,
+      "loss": 0.658,
+      "step": 26270
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015424974935947418,
+      "loss": 0.6725,
+      "step": 26280
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015419405146485462,
+      "loss": 0.6724,
+      "step": 26290
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00015413835357023503,
+      "loss": 0.6132,
+      "step": 26300
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00015408265567561544,
+      "loss": 0.7725,
+      "step": 26310
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00015402695778099585,
+      "loss": 0.6893,
+      "step": 26320
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001539712598863763,
+      "loss": 0.5694,
+      "step": 26330
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001539155619917567,
+      "loss": 0.6053,
+      "step": 26340
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00015385986409713714,
+      "loss": 0.79,
+      "step": 26350
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00015380416620251755,
+      "loss": 0.6865,
+      "step": 26360
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00015374846830789796,
+      "loss": 0.5849,
+      "step": 26370
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00015369277041327837,
+      "loss": 0.5446,
+      "step": 26380
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00015363707251865878,
+      "loss": 0.6265,
+      "step": 26390
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001535813746240392,
+      "loss": 0.5822,
+      "step": 26400
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001535256767294196,
+      "loss": 0.6995,
+      "step": 26410
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00015346997883480004,
+      "loss": 0.6874,
+      "step": 26420
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00015341428094018045,
+      "loss": 0.5584,
+      "step": 26430
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00015335858304556086,
+      "loss": 0.6684,
+      "step": 26440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00015330288515094128,
+      "loss": 0.7603,
+      "step": 26450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001532471872563217,
+      "loss": 0.6375,
+      "step": 26460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001531914893617021,
+      "loss": 0.7927,
+      "step": 26470
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00015313579146708254,
+      "loss": 0.7341,
+      "step": 26480
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00015308009357246295,
+      "loss": 0.6145,
+      "step": 26490
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00015302439567784336,
+      "loss": 0.7039,
+      "step": 26500
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00015296869778322377,
+      "loss": 0.565,
+      "step": 26510
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00015291299988860418,
+      "loss": 0.7221,
+      "step": 26520
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001528573019939846,
+      "loss": 0.5973,
+      "step": 26530
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.000152801604099365,
+      "loss": 0.6814,
+      "step": 26540
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00015274590620474547,
+      "loss": 0.6685,
+      "step": 26550
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00015269020831012588,
+      "loss": 0.5877,
+      "step": 26560
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0001526345104155063,
+      "loss": 0.7141,
+      "step": 26570
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001525788125208867,
+      "loss": 0.6058,
+      "step": 26580
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001525231146262671,
+      "loss": 0.7058,
+      "step": 26590
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00015246741673164752,
+      "loss": 0.654,
+      "step": 26600
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00015241171883702796,
+      "loss": 0.615,
+      "step": 26610
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00015235602094240837,
+      "loss": 0.6208,
+      "step": 26620
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00015230032304778878,
+      "loss": 0.7558,
+      "step": 26630
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001522446251531692,
+      "loss": 0.584,
+      "step": 26640
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001521889272585496,
+      "loss": 0.571,
+      "step": 26650
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00015213322936393002,
+      "loss": 0.6147,
+      "step": 26660
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00015207753146931043,
+      "loss": 0.5617,
+      "step": 26670
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00015202183357469087,
+      "loss": 0.6879,
+      "step": 26680
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00015196613568007128,
+      "loss": 0.5883,
+      "step": 26690
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001519104377854517,
+      "loss": 0.6871,
+      "step": 26700
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001518547398908321,
+      "loss": 0.6884,
+      "step": 26710
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001517990419962125,
+      "loss": 0.6958,
+      "step": 26720
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00015174334410159292,
+      "loss": 0.7593,
+      "step": 26730
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001516876462069734,
+      "loss": 0.6588,
+      "step": 26740
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001516319483123538,
+      "loss": 0.6906,
+      "step": 26750
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001515762504177342,
+      "loss": 0.5798,
+      "step": 26760
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00015152055252311462,
+      "loss": 0.5316,
+      "step": 26770
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00015146485462849503,
+      "loss": 0.7071,
+      "step": 26780
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00015140915673387544,
+      "loss": 0.6039,
+      "step": 26790
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00015135345883925586,
+      "loss": 0.6719,
+      "step": 26800
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001512977609446363,
+      "loss": 0.5624,
+      "step": 26810
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0001512420630500167,
+      "loss": 0.6667,
+      "step": 26820
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00015118636515539712,
+      "loss": 0.6293,
+      "step": 26830
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00015113066726077753,
+      "loss": 0.564,
+      "step": 26840
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00015107496936615794,
+      "loss": 0.6243,
+      "step": 26850
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00015101927147153835,
+      "loss": 0.707,
+      "step": 26860
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001509635735769188,
+      "loss": 0.6507,
+      "step": 26870
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001509078756822992,
+      "loss": 0.5755,
+      "step": 26880
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001508521777876796,
+      "loss": 0.658,
+      "step": 26890
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00015079647989306002,
+      "loss": 0.7153,
+      "step": 26900
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00015074078199844043,
+      "loss": 0.6214,
+      "step": 26910
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00015068508410382084,
+      "loss": 0.6069,
+      "step": 26920
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00015062938620920125,
+      "loss": 0.5345,
+      "step": 26930
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00015057368831458172,
+      "loss": 0.7243,
+      "step": 26940
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00015051799041996213,
+      "loss": 0.7971,
+      "step": 26950
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00015046229252534254,
+      "loss": 0.5465,
+      "step": 26960
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00015040659463072295,
+      "loss": 0.5908,
+      "step": 26970
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00015035089673610336,
+      "loss": 0.668,
+      "step": 26980
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00015029519884148378,
+      "loss": 0.4634,
+      "step": 26990
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00015023950094686421,
+      "loss": 0.7426,
+      "step": 27000
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00015018380305224462,
+      "loss": 0.5737,
+      "step": 27010
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00015012810515762504,
+      "loss": 0.6468,
+      "step": 27020
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00015007240726300545,
+      "loss": 0.5475,
+      "step": 27030
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00015001670936838586,
+      "loss": 0.7392,
+      "step": 27040
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001499610114737663,
+      "loss": 0.7629,
+      "step": 27050
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001499053135791467,
+      "loss": 0.6523,
+      "step": 27060
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00014984961568452712,
+      "loss": 0.6695,
+      "step": 27070
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00014979391778990753,
+      "loss": 0.447,
+      "step": 27080
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00014973821989528794,
+      "loss": 0.6267,
+      "step": 27090
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00014968252200066835,
+      "loss": 0.6354,
+      "step": 27100
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00014962682410604876,
+      "loss": 0.5283,
+      "step": 27110
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001495711262114292,
+      "loss": 0.551,
+      "step": 27120
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001495154283168096,
+      "loss": 0.7446,
+      "step": 27130
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014945973042219002,
+      "loss": 0.6156,
+      "step": 27140
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014940403252757046,
+      "loss": 0.6953,
+      "step": 27150
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014934833463295087,
+      "loss": 0.5751,
+      "step": 27160
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014929263673833128,
+      "loss": 0.52,
+      "step": 27170
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001492369388437117,
+      "loss": 0.5435,
+      "step": 27180
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001491812409490921,
+      "loss": 0.6627,
+      "step": 27190
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014912554305447252,
+      "loss": 0.6555,
+      "step": 27200
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014906984515985296,
+      "loss": 0.6808,
+      "step": 27210
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014901414726523337,
+      "loss": 0.5402,
+      "step": 27220
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014895844937061378,
+      "loss": 0.6724,
+      "step": 27230
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001489027514759942,
+      "loss": 0.6722,
+      "step": 27240
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014884705358137463,
+      "loss": 0.5147,
+      "step": 27250
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014879135568675504,
+      "loss": 0.8395,
+      "step": 27260
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014873565779213545,
+      "loss": 0.605,
+      "step": 27270
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014867995989751586,
+      "loss": 0.6245,
+      "step": 27280
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014862426200289627,
+      "loss": 0.5898,
+      "step": 27290
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014856856410827668,
+      "loss": 0.557,
+      "step": 27300
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014851286621365712,
+      "loss": 0.704,
+      "step": 27310
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014845716831903753,
+      "loss": 0.4358,
+      "step": 27320
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014840147042441794,
+      "loss": 0.5128,
+      "step": 27330
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014834577252979835,
+      "loss": 0.5276,
+      "step": 27340
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014829007463517877,
+      "loss": 0.6491,
+      "step": 27350
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001482343767405592,
+      "loss": 0.6062,
+      "step": 27360
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014817867884593962,
+      "loss": 0.759,
+      "step": 27370
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00014812298095132003,
+      "loss": 0.5563,
+      "step": 27380
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00014806728305670044,
+      "loss": 0.764,
+      "step": 27390
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00014801158516208085,
+      "loss": 0.4679,
+      "step": 27400
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001479558872674613,
+      "loss": 0.5921,
+      "step": 27410
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001479001893728417,
+      "loss": 0.6356,
+      "step": 27420
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001478444914782221,
+      "loss": 0.5951,
+      "step": 27430
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00014778879358360252,
+      "loss": 0.5872,
+      "step": 27440
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00014773309568898293,
+      "loss": 0.5969,
+      "step": 27450
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00014767739779436334,
+      "loss": 0.5302,
+      "step": 27460
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00014762169989974378,
+      "loss": 0.6272,
+      "step": 27470
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001475660020051242,
+      "loss": 0.563,
+      "step": 27480
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001475103041105046,
+      "loss": 0.6368,
+      "step": 27490
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00014745460621588501,
+      "loss": 0.5762,
+      "step": 27500
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00014739890832126545,
+      "loss": 0.6564,
+      "step": 27510
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00014734321042664586,
+      "loss": 0.747,
+      "step": 27520
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00014728751253202627,
+      "loss": 0.4904,
+      "step": 27530
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00014723181463740669,
+      "loss": 0.5357,
+      "step": 27540
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001471761167427871,
+      "loss": 0.6118,
+      "step": 27550
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001471204188481675,
+      "loss": 0.6067,
+      "step": 27560
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00014706472095354795,
+      "loss": 0.5896,
+      "step": 27570
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00014700902305892836,
+      "loss": 0.5464,
+      "step": 27580
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00014695332516430877,
+      "loss": 0.6674,
+      "step": 27590
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0001468976272696892,
+      "loss": 0.6726,
+      "step": 27600
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00014684192937506962,
+      "loss": 0.5646,
+      "step": 27610
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00014678623148045003,
+      "loss": 0.5054,
+      "step": 27620
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00014673053358583044,
+      "loss": 0.53,
+      "step": 27630
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00014667483569121085,
+      "loss": 0.6189,
+      "step": 27640
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00014661913779659126,
+      "loss": 0.6531,
+      "step": 27650
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014656343990197167,
+      "loss": 0.6671,
+      "step": 27660
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001465077420073521,
+      "loss": 0.5816,
+      "step": 27670
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014645204411273252,
+      "loss": 0.6263,
+      "step": 27680
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014639634621811293,
+      "loss": 0.6456,
+      "step": 27690
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014634064832349337,
+      "loss": 0.5815,
+      "step": 27700
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014628495042887378,
+      "loss": 0.6021,
+      "step": 27710
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001462292525342542,
+      "loss": 0.6195,
+      "step": 27720
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001461735546396346,
+      "loss": 0.576,
+      "step": 27730
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014611785674501502,
+      "loss": 0.6761,
+      "step": 27740
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014606215885039543,
+      "loss": 0.4831,
+      "step": 27750
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014600646095577587,
+      "loss": 0.5613,
+      "step": 27760
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014595076306115628,
+      "loss": 0.5572,
+      "step": 27770
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001458950651665367,
+      "loss": 0.5268,
+      "step": 27780
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014584493706137908,
+      "loss": 0.6035,
+      "step": 27790
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014578923916675949,
+      "loss": 0.5981,
+      "step": 27800
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001457335412721399,
+      "loss": 0.6424,
+      "step": 27810
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001456778433775203,
+      "loss": 0.4824,
+      "step": 27820
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014562214548290072,
+      "loss": 0.6069,
+      "step": 27830
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014556644758828116,
+      "loss": 0.497,
+      "step": 27840
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014551074969366157,
+      "loss": 0.5539,
+      "step": 27850
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014545505179904198,
+      "loss": 0.6126,
+      "step": 27860
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014539935390442242,
+      "loss": 0.5176,
+      "step": 27870
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014534365600980283,
+      "loss": 0.6168,
+      "step": 27880
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014528795811518324,
+      "loss": 0.5872,
+      "step": 27890
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014523226022056365,
+      "loss": 0.6517,
+      "step": 27900
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014517656232594406,
+      "loss": 0.5029,
+      "step": 27910
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00014512086443132447,
+      "loss": 0.5582,
+      "step": 27920
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001450651665367049,
+      "loss": 0.515,
+      "step": 27930
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00014500946864208532,
+      "loss": 0.5168,
+      "step": 27940
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00014495377074746573,
+      "loss": 0.5861,
+      "step": 27950
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00014489807285284615,
+      "loss": 0.6858,
+      "step": 27960
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00014484237495822656,
+      "loss": 0.5829,
+      "step": 27970
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.000144786677063607,
+      "loss": 0.5791,
+      "step": 27980
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001447309791689874,
+      "loss": 0.6624,
+      "step": 27990
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00014467528127436782,
+      "loss": 0.6267,
+      "step": 28000
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00014461958337974823,
+      "loss": 0.5421,
+      "step": 28010
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00014456388548512864,
+      "loss": 0.5864,
+      "step": 28020
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00014450818759050908,
+      "loss": 0.5509,
+      "step": 28030
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001444524896958895,
+      "loss": 0.6254,
+      "step": 28040
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001443967918012699,
+      "loss": 0.6194,
+      "step": 28050
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001443410939066503,
+      "loss": 0.5668,
+      "step": 28060
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00014428539601203072,
+      "loss": 0.6672,
+      "step": 28070
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00014422969811741113,
+      "loss": 0.5642,
+      "step": 28080
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00014417400022279154,
+      "loss": 0.639,
+      "step": 28090
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00014411830232817198,
+      "loss": 0.5227,
+      "step": 28100
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001440626044335524,
+      "loss": 0.6036,
+      "step": 28110
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001440069065389328,
+      "loss": 0.5755,
+      "step": 28120
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00014395120864431324,
+      "loss": 0.6136,
+      "step": 28130
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00014389551074969365,
+      "loss": 0.6739,
+      "step": 28140
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00014383981285507407,
+      "loss": 0.7447,
+      "step": 28150
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00014378411496045448,
+      "loss": 0.5471,
+      "step": 28160
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001437284170658349,
+      "loss": 0.5419,
+      "step": 28170
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0001436727191712153,
+      "loss": 0.59,
+      "step": 28180
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00014361702127659574,
+      "loss": 0.6926,
+      "step": 28190
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014356132338197615,
+      "loss": 0.5994,
+      "step": 28200
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014350562548735656,
+      "loss": 0.5359,
+      "step": 28210
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014344992759273697,
+      "loss": 0.6353,
+      "step": 28220
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0001433942296981174,
+      "loss": 0.5661,
+      "step": 28230
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014333853180349782,
+      "loss": 0.5923,
+      "step": 28240
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014328283390887823,
+      "loss": 0.6818,
+      "step": 28250
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014322713601425864,
+      "loss": 0.6298,
+      "step": 28260
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014317143811963905,
+      "loss": 0.653,
+      "step": 28270
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014311574022501946,
+      "loss": 0.5953,
+      "step": 28280
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0001430600423303999,
+      "loss": 0.5873,
+      "step": 28290
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014300434443578031,
+      "loss": 0.5688,
+      "step": 28300
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014294864654116073,
+      "loss": 0.6153,
+      "step": 28310
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014289294864654116,
+      "loss": 0.622,
+      "step": 28320
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014283725075192157,
+      "loss": 0.4433,
+      "step": 28330
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014278155285730199,
+      "loss": 0.6258,
+      "step": 28340
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0001427258549626824,
+      "loss": 0.6222,
+      "step": 28350
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0001426701570680628,
+      "loss": 0.7373,
+      "step": 28360
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014261445917344322,
+      "loss": 0.6533,
+      "step": 28370
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014255876127882363,
+      "loss": 0.6338,
+      "step": 28380
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014250306338420407,
+      "loss": 0.6207,
+      "step": 28390
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014244736548958448,
+      "loss": 0.5709,
+      "step": 28400
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0001423916675949649,
+      "loss": 0.5064,
+      "step": 28410
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014233596970034533,
+      "loss": 0.5504,
+      "step": 28420
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014228027180572574,
+      "loss": 0.5052,
+      "step": 28430
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014222457391110615,
+      "loss": 0.5614,
+      "step": 28440
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014216887601648656,
+      "loss": 0.751,
+      "step": 28450
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014211317812186697,
+      "loss": 0.5501,
+      "step": 28460
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00014205748022724738,
+      "loss": 0.5157,
+      "step": 28470
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0001420017823326278,
+      "loss": 0.6579,
+      "step": 28480
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00014194608443800823,
+      "loss": 0.5991,
+      "step": 28490
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00014189038654338865,
+      "loss": 0.621,
+      "step": 28500
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00014183468864876906,
+      "loss": 0.6897,
+      "step": 28510
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0001417789907541495,
+      "loss": 0.6427,
+      "step": 28520
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0001417232928595299,
+      "loss": 0.6338,
+      "step": 28530
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00014166759496491032,
+      "loss": 0.6423,
+      "step": 28540
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00014161189707029073,
+      "loss": 0.6832,
+      "step": 28550
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00014155619917567114,
+      "loss": 0.6847,
+      "step": 28560
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00014150050128105155,
+      "loss": 0.6373,
+      "step": 28570
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.000141444803386432,
+      "loss": 0.6729,
+      "step": 28580
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0001413891054918124,
+      "loss": 0.6302,
+      "step": 28590
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0001413334075971928,
+      "loss": 0.5747,
+      "step": 28600
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00014127770970257325,
+      "loss": 0.5689,
+      "step": 28610
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00014122201180795366,
+      "loss": 0.7226,
+      "step": 28620
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00014116631391333407,
+      "loss": 0.6597,
+      "step": 28630
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00014111061601871448,
+      "loss": 0.6318,
+      "step": 28640
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0001410549181240949,
+      "loss": 0.7549,
+      "step": 28650
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0001409992202294753,
+      "loss": 0.6374,
+      "step": 28660
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00014094352233485572,
+      "loss": 0.6357,
+      "step": 28670
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00014088782444023615,
+      "loss": 0.6386,
+      "step": 28680
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00014083212654561657,
+      "loss": 0.7495,
+      "step": 28690
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00014077642865099698,
+      "loss": 0.5263,
+      "step": 28700
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00014072073075637741,
+      "loss": 0.5395,
+      "step": 28710
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00014066503286175783,
+      "loss": 0.6153,
+      "step": 28720
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00014060933496713824,
+      "loss": 0.5783,
+      "step": 28730
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00014055363707251865,
+      "loss": 0.5446,
+      "step": 28740
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00014049793917789906,
+      "loss": 0.6922,
+      "step": 28750
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00014044224128327947,
+      "loss": 0.4609,
+      "step": 28760
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00014038654338865988,
+      "loss": 0.6839,
+      "step": 28770
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00014033084549404032,
+      "loss": 0.6182,
+      "step": 28780
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00014027514759942073,
+      "loss": 0.7257,
+      "step": 28790
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00014021944970480114,
+      "loss": 0.6834,
+      "step": 28800
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00014016375181018158,
+      "loss": 0.5532,
+      "step": 28810
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.000140108053915562,
+      "loss": 0.6574,
+      "step": 28820
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001400523560209424,
+      "loss": 0.58,
+      "step": 28830
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001399966581263228,
+      "loss": 0.5084,
+      "step": 28840
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00013994096023170322,
+      "loss": 0.6035,
+      "step": 28850
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00013988526233708364,
+      "loss": 0.5978,
+      "step": 28860
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00013982956444246407,
+      "loss": 0.6123,
+      "step": 28870
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00013977386654784449,
+      "loss": 0.5973,
+      "step": 28880
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001397181686532249,
+      "loss": 0.545,
+      "step": 28890
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001396624707586053,
+      "loss": 0.6065,
+      "step": 28900
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00013960677286398575,
+      "loss": 0.5697,
+      "step": 28910
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00013955107496936616,
+      "loss": 0.6563,
+      "step": 28920
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00013949537707474657,
+      "loss": 0.5535,
+      "step": 28930
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00013943967918012698,
+      "loss": 0.5575,
+      "step": 28940
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001393839812855074,
+      "loss": 0.5724,
+      "step": 28950
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001393282833908878,
+      "loss": 0.5215,
+      "step": 28960
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001392781552857302,
+      "loss": 0.6067,
+      "step": 28970
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001392224573911106,
+      "loss": 0.6617,
+      "step": 28980
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.000139166759496491,
+      "loss": 0.6187,
+      "step": 28990
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00013911106160187142,
+      "loss": 0.5893,
+      "step": 29000
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013905536370725186,
+      "loss": 0.6454,
+      "step": 29010
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013899966581263227,
+      "loss": 0.5571,
+      "step": 29020
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013894396791801268,
+      "loss": 0.7921,
+      "step": 29030
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013888827002339312,
+      "loss": 0.5683,
+      "step": 29040
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013883257212877353,
+      "loss": 0.5369,
+      "step": 29050
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013877687423415394,
+      "loss": 0.5291,
+      "step": 29060
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013872117633953435,
+      "loss": 0.6136,
+      "step": 29070
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013866547844491476,
+      "loss": 0.569,
+      "step": 29080
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013860978055029518,
+      "loss": 0.4386,
+      "step": 29090
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0001385540826556756,
+      "loss": 0.5527,
+      "step": 29100
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013849838476105602,
+      "loss": 0.5459,
+      "step": 29110
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013844268686643644,
+      "loss": 0.6416,
+      "step": 29120
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013838698897181685,
+      "loss": 0.586,
+      "step": 29130
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013833129107719729,
+      "loss": 0.4898,
+      "step": 29140
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0001382755931825777,
+      "loss": 0.6606,
+      "step": 29150
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0001382198952879581,
+      "loss": 0.7579,
+      "step": 29160
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013816419739333852,
+      "loss": 0.7036,
+      "step": 29170
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013810849949871893,
+      "loss": 0.5601,
+      "step": 29180
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013805280160409934,
+      "loss": 0.4931,
+      "step": 29190
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013799710370947975,
+      "loss": 0.6479,
+      "step": 29200
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0001379414058148602,
+      "loss": 0.61,
+      "step": 29210
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0001378857079202406,
+      "loss": 0.6225,
+      "step": 29220
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.000137830010025621,
+      "loss": 0.5489,
+      "step": 29230
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013777431213100145,
+      "loss": 0.5307,
+      "step": 29240
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013771861423638186,
+      "loss": 0.6051,
+      "step": 29250
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013766291634176227,
+      "loss": 0.604,
+      "step": 29260
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00013760721844714268,
+      "loss": 0.5571,
+      "step": 29270
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001375515205525231,
+      "loss": 0.4529,
+      "step": 29280
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001374958226579035,
+      "loss": 0.4945,
+      "step": 29290
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013744012476328394,
+      "loss": 0.7455,
+      "step": 29300
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013738442686866436,
+      "loss": 0.7226,
+      "step": 29310
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013732872897404477,
+      "loss": 0.6678,
+      "step": 29320
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013727303107942518,
+      "loss": 0.6338,
+      "step": 29330
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013721733318480562,
+      "loss": 0.5416,
+      "step": 29340
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013716163529018603,
+      "loss": 0.5477,
+      "step": 29350
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013710593739556644,
+      "loss": 0.6874,
+      "step": 29360
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013705023950094685,
+      "loss": 0.5968,
+      "step": 29370
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013699454160632726,
+      "loss": 0.4746,
+      "step": 29380
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013693884371170767,
+      "loss": 0.5656,
+      "step": 29390
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001368831458170881,
+      "loss": 0.6235,
+      "step": 29400
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013682744792246852,
+      "loss": 0.5645,
+      "step": 29410
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013677175002784893,
+      "loss": 0.5845,
+      "step": 29420
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013671605213322937,
+      "loss": 0.5449,
+      "step": 29430
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013666035423860978,
+      "loss": 0.6532,
+      "step": 29440
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001366046563439902,
+      "loss": 0.6428,
+      "step": 29450
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001365489584493706,
+      "loss": 0.5571,
+      "step": 29460
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013649326055475102,
+      "loss": 0.624,
+      "step": 29470
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013643756266013143,
+      "loss": 0.5874,
+      "step": 29480
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013638186476551184,
+      "loss": 0.5713,
+      "step": 29490
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013632616687089228,
+      "loss": 0.6077,
+      "step": 29500
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001362704689762727,
+      "loss": 0.5108,
+      "step": 29510
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0001362147710816531,
+      "loss": 0.6291,
+      "step": 29520
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013615907318703354,
+      "loss": 0.6297,
+      "step": 29530
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00013610337529241395,
+      "loss": 0.5852,
+      "step": 29540
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00013604767739779436,
+      "loss": 0.5487,
+      "step": 29550
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00013599197950317477,
+      "loss": 0.64,
+      "step": 29560
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00013593628160855518,
+      "loss": 0.559,
+      "step": 29570
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0001358805837139356,
+      "loss": 0.7788,
+      "step": 29580
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00013582488581931603,
+      "loss": 0.6607,
+      "step": 29590
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00013576918792469644,
+      "loss": 0.6098,
+      "step": 29600
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00013571349003007685,
+      "loss": 0.6405,
+      "step": 29610
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00013565779213545726,
+      "loss": 0.5299,
+      "step": 29620
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0001356020942408377,
+      "loss": 0.5513,
+      "step": 29630
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0001355463963462181,
+      "loss": 0.7151,
+      "step": 29640
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00013549069845159852,
+      "loss": 0.6107,
+      "step": 29650
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00013543500055697894,
+      "loss": 0.6224,
+      "step": 29660
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00013537930266235935,
+      "loss": 0.4986,
+      "step": 29670
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00013532360476773976,
+      "loss": 0.5785,
+      "step": 29680
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0001352679068731202,
+      "loss": 0.7123,
+      "step": 29690
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0001352122089785006,
+      "loss": 0.6334,
+      "step": 29700
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00013515651108388102,
+      "loss": 0.5445,
+      "step": 29710
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00013510081318926143,
+      "loss": 0.5722,
+      "step": 29720
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00013504511529464184,
+      "loss": 0.6186,
+      "step": 29730
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00013498941740002225,
+      "loss": 0.6591,
+      "step": 29740
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0001349337195054027,
+      "loss": 0.6334,
+      "step": 29750
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0001348780216107831,
+      "loss": 0.6241,
+      "step": 29760
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0001348223237161635,
+      "loss": 0.6335,
+      "step": 29770
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00013476662582154392,
+      "loss": 0.576,
+      "step": 29780
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00013471092792692436,
+      "loss": 0.6302,
+      "step": 29790
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00013465523003230477,
+      "loss": 0.5341,
+      "step": 29800
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00013459953213768518,
+      "loss": 0.5089,
+      "step": 29810
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.0001345438342430656,
+      "loss": 0.5115,
+      "step": 29820
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.000134488136348446,
+      "loss": 0.5861,
+      "step": 29830
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00013443243845382642,
+      "loss": 0.5592,
+      "step": 29840
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00013437674055920686,
+      "loss": 0.6329,
+      "step": 29850
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00013432104266458727,
+      "loss": 0.6253,
+      "step": 29860
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00013426534476996768,
+      "loss": 0.6283,
+      "step": 29870
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.0001342096468753481,
+      "loss": 0.548,
+      "step": 29880
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00013415394898072853,
+      "loss": 0.6634,
+      "step": 29890
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00013409825108610894,
+      "loss": 0.6905,
+      "step": 29900
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00013404255319148935,
+      "loss": 0.6338,
+      "step": 29910
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00013398685529686976,
+      "loss": 0.6552,
+      "step": 29920
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00013393115740225017,
+      "loss": 0.5042,
+      "step": 29930
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00013387545950763058,
+      "loss": 0.5048,
+      "step": 29940
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00013381976161301102,
+      "loss": 0.5588,
+      "step": 29950
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00013376406371839143,
+      "loss": 0.6686,
+      "step": 29960
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00013370836582377184,
+      "loss": 0.6719,
+      "step": 29970
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00013365266792915228,
+      "loss": 0.4942,
+      "step": 29980
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.0001335969700345327,
+      "loss": 0.5265,
+      "step": 29990
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.0001335412721399131,
+      "loss": 0.499,
+      "step": 30000
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00013348557424529351,
+      "loss": 0.6948,
+      "step": 30010
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00013342987635067393,
+      "loss": 0.6221,
+      "step": 30020
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00013337417845605434,
+      "loss": 0.6036,
+      "step": 30030
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00013331848056143475,
+      "loss": 0.6215,
+      "step": 30040
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.0001332627826668152,
+      "loss": 0.6551,
+      "step": 30050
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.0001332070847721956,
+      "loss": 0.6517,
+      "step": 30060
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.000133151386877576,
+      "loss": 0.6602,
+      "step": 30070
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00013309568898295645,
+      "loss": 0.589,
+      "step": 30080
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013303999108833686,
+      "loss": 0.4997,
+      "step": 30090
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013298429319371727,
+      "loss": 0.6696,
+      "step": 30100
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013292859529909768,
+      "loss": 0.6318,
+      "step": 30110
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0001328728974044781,
+      "loss": 0.5285,
+      "step": 30120
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0001328171995098585,
+      "loss": 0.6166,
+      "step": 30130
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013276150161523891,
+      "loss": 0.6143,
+      "step": 30140
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013270580372061935,
+      "loss": 0.7933,
+      "step": 30150
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013265010582599976,
+      "loss": 0.5831,
+      "step": 30160
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013259440793138017,
+      "loss": 0.5665,
+      "step": 30170
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0001325387100367606,
+      "loss": 0.5789,
+      "step": 30180
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013248301214214102,
+      "loss": 0.6259,
+      "step": 30190
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013242731424752143,
+      "loss": 0.5455,
+      "step": 30200
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013237161635290185,
+      "loss": 0.5899,
+      "step": 30210
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013231591845828226,
+      "loss": 0.532,
+      "step": 30220
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013226022056366267,
+      "loss": 0.5555,
+      "step": 30230
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0001322045226690431,
+      "loss": 0.5188,
+      "step": 30240
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013214882477442352,
+      "loss": 0.6565,
+      "step": 30250
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013209312687980393,
+      "loss": 0.6032,
+      "step": 30260
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013203742898518434,
+      "loss": 0.5055,
+      "step": 30270
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013198173109056478,
+      "loss": 0.6185,
+      "step": 30280
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0001319260331959452,
+      "loss": 0.5907,
+      "step": 30290
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0001318703353013256,
+      "loss": 0.5669,
+      "step": 30300
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.000131814637406706,
+      "loss": 0.6931,
+      "step": 30310
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013175893951208642,
+      "loss": 0.6343,
+      "step": 30320
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013170324161746683,
+      "loss": 0.6885,
+      "step": 30330
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013164754372284727,
+      "loss": 0.5956,
+      "step": 30340
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00013159184582822768,
+      "loss": 0.5588,
+      "step": 30350
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0001315361479336081,
+      "loss": 0.5711,
+      "step": 30360
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00013148045003898853,
+      "loss": 0.6897,
+      "step": 30370
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00013142475214436894,
+      "loss": 0.7197,
+      "step": 30380
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00013136905424974935,
+      "loss": 0.6109,
+      "step": 30390
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00013131335635512977,
+      "loss": 0.648,
+      "step": 30400
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00013125765846051018,
+      "loss": 0.5518,
+      "step": 30410
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0001312019605658906,
+      "loss": 0.8319,
+      "step": 30420
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.000131146262671271,
+      "loss": 0.5594,
+      "step": 30430
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00013109056477665144,
+      "loss": 0.583,
+      "step": 30440
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00013103486688203185,
+      "loss": 0.6148,
+      "step": 30450
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00013097916898741226,
+      "loss": 0.5644,
+      "step": 30460
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0001309234710927927,
+      "loss": 0.4781,
+      "step": 30470
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0001308677731981731,
+      "loss": 0.6261,
+      "step": 30480
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00013081207530355352,
+      "loss": 0.6081,
+      "step": 30490
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00013075637740893393,
+      "loss": 0.7234,
+      "step": 30500
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00013070067951431434,
+      "loss": 0.5754,
+      "step": 30510
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00013064498161969475,
+      "loss": 0.7789,
+      "step": 30520
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0001305892837250752,
+      "loss": 0.6051,
+      "step": 30530
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0001305335858304556,
+      "loss": 0.5934,
+      "step": 30540
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00013047788793583601,
+      "loss": 0.6367,
+      "step": 30550
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00013042219004121643,
+      "loss": 0.6543,
+      "step": 30560
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00013036649214659686,
+      "loss": 0.6487,
+      "step": 30570
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00013031079425197727,
+      "loss": 0.6244,
+      "step": 30580
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00013025509635735769,
+      "loss": 0.516,
+      "step": 30590
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0001301993984627381,
+      "loss": 0.506,
+      "step": 30600
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0001301437005681185,
+      "loss": 0.7163,
+      "step": 30610
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00013008800267349892,
+      "loss": 0.5847,
+      "step": 30620
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00013003230477887936,
+      "loss": 0.5561,
+      "step": 30630
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012997660688425977,
+      "loss": 0.5224,
+      "step": 30640
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012992090898964018,
+      "loss": 0.5045,
+      "step": 30650
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012986521109502062,
+      "loss": 0.6612,
+      "step": 30660
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012980951320040103,
+      "loss": 0.6424,
+      "step": 30670
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012975381530578144,
+      "loss": 0.5195,
+      "step": 30680
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012969811741116185,
+      "loss": 0.5494,
+      "step": 30690
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012964241951654226,
+      "loss": 0.6533,
+      "step": 30700
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012958672162192267,
+      "loss": 0.6205,
+      "step": 30710
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012953102372730309,
+      "loss": 0.5592,
+      "step": 30720
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012947532583268352,
+      "loss": 0.5923,
+      "step": 30730
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012941962793806393,
+      "loss": 0.5933,
+      "step": 30740
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012936393004344435,
+      "loss": 0.6032,
+      "step": 30750
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012930823214882476,
+      "loss": 0.5956,
+      "step": 30760
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012925253425420517,
+      "loss": 0.6083,
+      "step": 30770
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0001291968363595856,
+      "loss": 0.5612,
+      "step": 30780
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012914113846496602,
+      "loss": 0.6401,
+      "step": 30790
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012908544057034643,
+      "loss": 0.5297,
+      "step": 30800
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012902974267572684,
+      "loss": 0.6286,
+      "step": 30810
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012897404478110725,
+      "loss": 0.5737,
+      "step": 30820
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0001289183468864877,
+      "loss": 0.5317,
+      "step": 30830
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0001288626489918681,
+      "loss": 0.5812,
+      "step": 30840
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0001288069510972485,
+      "loss": 0.6965,
+      "step": 30850
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012875125320262892,
+      "loss": 0.5091,
+      "step": 30860
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012869555530800933,
+      "loss": 0.3956,
+      "step": 30870
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012863985741338974,
+      "loss": 0.6235,
+      "step": 30880
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00012858415951877018,
+      "loss": 0.7077,
+      "step": 30890
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0001285284616241506,
+      "loss": 0.6143,
+      "step": 30900
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.000128472763729531,
+      "loss": 0.5636,
+      "step": 30910
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00012841706583491144,
+      "loss": 0.6609,
+      "step": 30920
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00012836136794029185,
+      "loss": 0.4568,
+      "step": 30930
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00012830567004567227,
+      "loss": 0.5697,
+      "step": 30940
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00012824997215105268,
+      "loss": 0.557,
+      "step": 30950
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0001281942742564331,
+      "loss": 0.5489,
+      "step": 30960
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0001281385763618135,
+      "loss": 0.6089,
+      "step": 30970
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0001280828784671939,
+      "loss": 0.6368,
+      "step": 30980
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00012802718057257435,
+      "loss": 0.6284,
+      "step": 30990
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00012797148267795476,
+      "loss": 0.6526,
+      "step": 31000
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00012791578478333517,
+      "loss": 0.5384,
+      "step": 31010
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0001278600868887156,
+      "loss": 0.5488,
+      "step": 31020
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00012780438899409602,
+      "loss": 0.5408,
+      "step": 31030
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00012774869109947643,
+      "loss": 0.6549,
+      "step": 31040
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00012769299320485684,
+      "loss": 0.6361,
+      "step": 31050
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00012763729531023725,
+      "loss": 0.6188,
+      "step": 31060
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00012758159741561766,
+      "loss": 0.6195,
+      "step": 31070
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00012752589952099808,
+      "loss": 0.6442,
+      "step": 31080
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00012747020162637851,
+      "loss": 0.6078,
+      "step": 31090
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00012741450373175893,
+      "loss": 0.6378,
+      "step": 31100
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00012735880583713934,
+      "loss": 0.6,
+      "step": 31110
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00012730310794251977,
+      "loss": 0.4613,
+      "step": 31120
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00012724741004790019,
+      "loss": 0.5122,
+      "step": 31130
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0001271917121532806,
+      "loss": 0.7254,
+      "step": 31140
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.000127136014258661,
+      "loss": 0.5964,
+      "step": 31150
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00012708031636404142,
+      "loss": 0.6135,
+      "step": 31160
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012702461846942183,
+      "loss": 0.5749,
+      "step": 31170
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012696892057480227,
+      "loss": 0.6659,
+      "step": 31180
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012691322268018268,
+      "loss": 0.7007,
+      "step": 31190
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0001268575247855631,
+      "loss": 0.6446,
+      "step": 31200
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0001268018268909435,
+      "loss": 0.6617,
+      "step": 31210
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012674612899632394,
+      "loss": 0.657,
+      "step": 31220
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012669043110170435,
+      "loss": 0.5117,
+      "step": 31230
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012663473320708476,
+      "loss": 0.5951,
+      "step": 31240
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012657903531246517,
+      "loss": 0.5559,
+      "step": 31250
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012652333741784558,
+      "loss": 0.6683,
+      "step": 31260
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.000126467639523226,
+      "loss": 0.5747,
+      "step": 31270
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012641194162860643,
+      "loss": 0.6786,
+      "step": 31280
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012635624373398685,
+      "loss": 0.5221,
+      "step": 31290
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012630054583936726,
+      "loss": 0.5608,
+      "step": 31300
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0001262448479447477,
+      "loss": 0.4458,
+      "step": 31310
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0001261891500501281,
+      "loss": 0.6528,
+      "step": 31320
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012613345215550852,
+      "loss": 0.5739,
+      "step": 31330
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012607775426088893,
+      "loss": 0.7336,
+      "step": 31340
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012602205636626934,
+      "loss": 0.5657,
+      "step": 31350
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012596635847164975,
+      "loss": 0.6638,
+      "step": 31360
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012591066057703016,
+      "loss": 0.6651,
+      "step": 31370
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0001258549626824106,
+      "loss": 0.4942,
+      "step": 31380
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.000125799264787791,
+      "loss": 0.4958,
+      "step": 31390
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012574356689317142,
+      "loss": 0.6147,
+      "step": 31400
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012568786899855186,
+      "loss": 0.5585,
+      "step": 31410
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012563217110393227,
+      "loss": 0.5992,
+      "step": 31420
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00012557647320931268,
+      "loss": 0.4611,
+      "step": 31430
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0001255207753146931,
+      "loss": 0.5153,
+      "step": 31440
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0001254650774200735,
+      "loss": 0.5479,
+      "step": 31450
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00012540937952545392,
+      "loss": 0.6117,
+      "step": 31460
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00012535368163083435,
+      "loss": 0.5487,
+      "step": 31470
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00012529798373621477,
+      "loss": 0.675,
+      "step": 31480
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00012524228584159518,
+      "loss": 0.6062,
+      "step": 31490
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0001251865879469756,
+      "loss": 0.5955,
+      "step": 31500
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00012513089005235603,
+      "loss": 0.6006,
+      "step": 31510
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00012507519215773644,
+      "loss": 0.5306,
+      "step": 31520
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00012501949426311685,
+      "loss": 0.621,
+      "step": 31530
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00012496379636849726,
+      "loss": 0.5824,
+      "step": 31540
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00012490809847387767,
+      "loss": 0.5662,
+      "step": 31550
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00012485240057925808,
+      "loss": 0.6175,
+      "step": 31560
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00012479670268463852,
+      "loss": 0.6209,
+      "step": 31570
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00012474100479001893,
+      "loss": 0.465,
+      "step": 31580
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00012468530689539934,
+      "loss": 0.6752,
+      "step": 31590
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00012462960900077978,
+      "loss": 0.6325,
+      "step": 31600
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0001245739111061602,
+      "loss": 0.6166,
+      "step": 31610
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0001245182132115406,
+      "loss": 0.6125,
+      "step": 31620
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.000124462515316921,
+      "loss": 0.6336,
+      "step": 31630
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00012440681742230142,
+      "loss": 0.6797,
+      "step": 31640
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00012435111952768184,
+      "loss": 0.5689,
+      "step": 31650
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00012429542163306225,
+      "loss": 0.7314,
+      "step": 31660
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00012423972373844269,
+      "loss": 0.5623,
+      "step": 31670
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0001241840258438231,
+      "loss": 0.599,
+      "step": 31680
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0001241283279492035,
+      "loss": 0.5516,
+      "step": 31690
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00012407263005458395,
+      "loss": 0.7278,
+      "step": 31700
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012401693215996436,
+      "loss": 0.6036,
+      "step": 31710
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012396123426534477,
+      "loss": 0.6737,
+      "step": 31720
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012390553637072518,
+      "loss": 0.6582,
+      "step": 31730
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0001238498384761056,
+      "loss": 0.615,
+      "step": 31740
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.000123794140581486,
+      "loss": 0.5693,
+      "step": 31750
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0001237384426868664,
+      "loss": 0.6386,
+      "step": 31760
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012368274479224685,
+      "loss": 0.7385,
+      "step": 31770
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012362704689762726,
+      "loss": 0.4999,
+      "step": 31780
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012357134900300767,
+      "loss": 0.5856,
+      "step": 31790
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012351565110838808,
+      "loss": 0.5285,
+      "step": 31800
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012345995321376852,
+      "loss": 0.5889,
+      "step": 31810
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012340425531914893,
+      "loss": 0.5239,
+      "step": 31820
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012334855742452934,
+      "loss": 0.5242,
+      "step": 31830
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012329285952990976,
+      "loss": 0.5063,
+      "step": 31840
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012323716163529017,
+      "loss": 0.568,
+      "step": 31850
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0001231814637406706,
+      "loss": 0.5067,
+      "step": 31860
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012312576584605102,
+      "loss": 0.6699,
+      "step": 31870
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012307006795143143,
+      "loss": 0.6278,
+      "step": 31880
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012301437005681184,
+      "loss": 0.6087,
+      "step": 31890
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012295867216219225,
+      "loss": 0.5501,
+      "step": 31900
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012290297426757266,
+      "loss": 0.5187,
+      "step": 31910
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012284727637295307,
+      "loss": 0.6098,
+      "step": 31920
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0001227915784783335,
+      "loss": 0.5598,
+      "step": 31930
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012273588058371392,
+      "loss": 0.6283,
+      "step": 31940
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012268018268909433,
+      "loss": 0.5931,
+      "step": 31950
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012262448479447477,
+      "loss": 0.6998,
+      "step": 31960
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00012256878689985518,
+      "loss": 0.6068,
+      "step": 31970
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0001225130890052356,
+      "loss": 0.6025,
+      "step": 31980
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.000122457391110616,
+      "loss": 0.4986,
+      "step": 31990
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00012240169321599642,
+      "loss": 0.5281,
+      "step": 32000
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00012234599532137683,
+      "loss": 0.4619,
+      "step": 32010
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00012229029742675724,
+      "loss": 0.5052,
+      "step": 32020
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00012223459953213768,
+      "loss": 0.6326,
+      "step": 32030
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0001221789016375181,
+      "loss": 0.6529,
+      "step": 32040
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0001221232037428985,
+      "loss": 0.5777,
+      "step": 32050
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00012206750584827892,
+      "loss": 0.4897,
+      "step": 32060
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00012201180795365935,
+      "loss": 0.6097,
+      "step": 32070
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00012195611005903976,
+      "loss": 0.6802,
+      "step": 32080
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00012190041216442018,
+      "loss": 0.6042,
+      "step": 32090
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0001218447142698006,
+      "loss": 0.4821,
+      "step": 32100
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.000121789016375181,
+      "loss": 0.7123,
+      "step": 32110
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00012173331848056143,
+      "loss": 0.6211,
+      "step": 32120
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00012167762058594184,
+      "loss": 0.5428,
+      "step": 32130
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00012162192269132225,
+      "loss": 0.5707,
+      "step": 32140
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00012156622479670266,
+      "loss": 0.585,
+      "step": 32150
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00012151052690208309,
+      "loss": 0.5349,
+      "step": 32160
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0001214548290074635,
+      "loss": 0.5231,
+      "step": 32170
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00012139913111284391,
+      "loss": 0.6829,
+      "step": 32180
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00012134343321822435,
+      "loss": 0.5766,
+      "step": 32190
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00012128773532360476,
+      "loss": 0.5331,
+      "step": 32200
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00012123203742898517,
+      "loss": 0.6083,
+      "step": 32210
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0001211763395343656,
+      "loss": 0.6622,
+      "step": 32220
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00012112064163974601,
+      "loss": 0.657,
+      "step": 32230
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00012106494374512642,
+      "loss": 0.556,
+      "step": 32240
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00012100924585050684,
+      "loss": 0.6711,
+      "step": 32250
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00012095354795588725,
+      "loss": 0.6491,
+      "step": 32260
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00012089785006126766,
+      "loss": 0.6109,
+      "step": 32270
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00012084215216664808,
+      "loss": 0.6375,
+      "step": 32280
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00012078645427202851,
+      "loss": 0.5367,
+      "step": 32290
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00012073075637740893,
+      "loss": 0.4988,
+      "step": 32300
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00012067505848278934,
+      "loss": 0.5967,
+      "step": 32310
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00012061936058816976,
+      "loss": 0.8054,
+      "step": 32320
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00012056366269355017,
+      "loss": 0.6418,
+      "step": 32330
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00012050796479893058,
+      "loss": 0.5394,
+      "step": 32340
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00012045226690431101,
+      "loss": 0.532,
+      "step": 32350
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00012039656900969142,
+      "loss": 0.5286,
+      "step": 32360
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00012034087111507183,
+      "loss": 0.4502,
+      "step": 32370
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00012028517322045227,
+      "loss": 0.5606,
+      "step": 32380
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00012022947532583268,
+      "loss": 0.6236,
+      "step": 32390
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00012017377743121309,
+      "loss": 0.5886,
+      "step": 32400
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00012011807953659352,
+      "loss": 0.6355,
+      "step": 32410
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00012006238164197393,
+      "loss": 0.7605,
+      "step": 32420
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00012000668374735434,
+      "loss": 0.6216,
+      "step": 32430
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00011995098585273475,
+      "loss": 0.5012,
+      "step": 32440
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00011989528795811517,
+      "loss": 0.5292,
+      "step": 32450
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00011983959006349558,
+      "loss": 0.5439,
+      "step": 32460
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.000119783892168876,
+      "loss": 0.5853,
+      "step": 32470
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00011972819427425643,
+      "loss": 0.5336,
+      "step": 32480
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00011967249637963685,
+      "loss": 0.6061,
+      "step": 32490
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00011961679848501726,
+      "loss": 0.6228,
+      "step": 32500
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00011956667037985963,
+      "loss": 0.6121,
+      "step": 32510
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011951097248524005,
+      "loss": 0.8102,
+      "step": 32520
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011945527459062047,
+      "loss": 0.6016,
+      "step": 32530
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011939957669600088,
+      "loss": 0.6931,
+      "step": 32540
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0001193438788013813,
+      "loss": 0.603,
+      "step": 32550
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011928818090676171,
+      "loss": 0.5655,
+      "step": 32560
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011923248301214212,
+      "loss": 0.7121,
+      "step": 32570
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011917678511752256,
+      "loss": 0.6758,
+      "step": 32580
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011912108722290297,
+      "loss": 0.5436,
+      "step": 32590
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011906538932828338,
+      "loss": 0.6406,
+      "step": 32600
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0001190096914336638,
+      "loss": 0.4377,
+      "step": 32610
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011895399353904422,
+      "loss": 0.5245,
+      "step": 32620
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011889829564442463,
+      "loss": 0.5301,
+      "step": 32630
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011884259774980504,
+      "loss": 0.5823,
+      "step": 32640
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011878689985518547,
+      "loss": 0.5917,
+      "step": 32650
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011873120196056588,
+      "loss": 0.5153,
+      "step": 32660
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011867550406594629,
+      "loss": 0.5509,
+      "step": 32670
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011861980617132671,
+      "loss": 0.5931,
+      "step": 32680
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011856410827670714,
+      "loss": 0.6181,
+      "step": 32690
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011850841038208755,
+      "loss": 0.4056,
+      "step": 32700
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011845271248746797,
+      "loss": 0.5899,
+      "step": 32710
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011839701459284839,
+      "loss": 0.5242,
+      "step": 32720
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0001183413166982288,
+      "loss": 0.7591,
+      "step": 32730
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011828561880360921,
+      "loss": 0.7375,
+      "step": 32740
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011822992090898963,
+      "loss": 0.5444,
+      "step": 32750
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011817422301437004,
+      "loss": 0.5866,
+      "step": 32760
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011811852511975045,
+      "loss": 0.6804,
+      "step": 32770
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00011806282722513088,
+      "loss": 0.5704,
+      "step": 32780
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011800712933051129,
+      "loss": 0.7149,
+      "step": 32790
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0001179514314358917,
+      "loss": 0.565,
+      "step": 32800
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011789573354127214,
+      "loss": 0.5827,
+      "step": 32810
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011784003564665255,
+      "loss": 0.483,
+      "step": 32820
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011778433775203296,
+      "loss": 0.5689,
+      "step": 32830
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011772863985741339,
+      "loss": 0.5791,
+      "step": 32840
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0001176729419627938,
+      "loss": 0.5717,
+      "step": 32850
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011761724406817421,
+      "loss": 0.5968,
+      "step": 32860
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011756154617355462,
+      "loss": 0.6956,
+      "step": 32870
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011750584827893504,
+      "loss": 0.7132,
+      "step": 32880
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011745015038431546,
+      "loss": 0.5122,
+      "step": 32890
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011739445248969587,
+      "loss": 0.5693,
+      "step": 32900
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0001173387545950763,
+      "loss": 0.5706,
+      "step": 32910
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011728305670045672,
+      "loss": 0.5681,
+      "step": 32920
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011722735880583713,
+      "loss": 0.649,
+      "step": 32930
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011717166091121755,
+      "loss": 0.745,
+      "step": 32940
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011711596301659796,
+      "loss": 0.5371,
+      "step": 32950
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011706026512197837,
+      "loss": 0.6556,
+      "step": 32960
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0001170045672273588,
+      "loss": 0.6298,
+      "step": 32970
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011694886933273921,
+      "loss": 0.6426,
+      "step": 32980
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011689317143811962,
+      "loss": 0.5628,
+      "step": 32990
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011683747354350003,
+      "loss": 0.5488,
+      "step": 33000
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011678177564888047,
+      "loss": 0.7095,
+      "step": 33010
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011672607775426088,
+      "loss": 0.5199,
+      "step": 33020
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011667037985964129,
+      "loss": 0.5917,
+      "step": 33030
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011661468196502172,
+      "loss": 0.5515,
+      "step": 33040
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00011655898407040213,
+      "loss": 0.5361,
+      "step": 33050
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011650328617578254,
+      "loss": 0.6972,
+      "step": 33060
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011644758828116296,
+      "loss": 0.5692,
+      "step": 33070
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011639189038654338,
+      "loss": 0.6353,
+      "step": 33080
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011633619249192379,
+      "loss": 0.6797,
+      "step": 33090
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011628049459730423,
+      "loss": 0.5158,
+      "step": 33100
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011622479670268464,
+      "loss": 0.7578,
+      "step": 33110
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011616909880806505,
+      "loss": 0.6854,
+      "step": 33120
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011611340091344546,
+      "loss": 0.5312,
+      "step": 33130
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011605770301882588,
+      "loss": 0.602,
+      "step": 33140
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0001160020051242063,
+      "loss": 0.5676,
+      "step": 33150
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0001159463072295867,
+      "loss": 0.5214,
+      "step": 33160
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011589060933496713,
+      "loss": 0.5496,
+      "step": 33170
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011583491144034754,
+      "loss": 0.5627,
+      "step": 33180
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011577921354572795,
+      "loss": 0.4726,
+      "step": 33190
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011572351565110838,
+      "loss": 0.5614,
+      "step": 33200
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0001156678177564888,
+      "loss": 0.7228,
+      "step": 33210
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011561211986186921,
+      "loss": 0.7056,
+      "step": 33220
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011555642196724964,
+      "loss": 0.6883,
+      "step": 33230
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011550072407263005,
+      "loss": 0.7162,
+      "step": 33240
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011544502617801046,
+      "loss": 0.5727,
+      "step": 33250
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011538932828339087,
+      "loss": 0.5708,
+      "step": 33260
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0001153336303887713,
+      "loss": 0.53,
+      "step": 33270
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011527793249415171,
+      "loss": 0.5894,
+      "step": 33280
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011522223459953212,
+      "loss": 0.5931,
+      "step": 33290
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011516653670491254,
+      "loss": 0.5561,
+      "step": 33300
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011511083881029295,
+      "loss": 0.5732,
+      "step": 33310
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00011505514091567337,
+      "loss": 0.7054,
+      "step": 33320
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0001149994430210538,
+      "loss": 0.6209,
+      "step": 33330
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011494374512643421,
+      "loss": 0.5694,
+      "step": 33340
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011488804723181463,
+      "loss": 0.49,
+      "step": 33350
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011483234933719505,
+      "loss": 0.6434,
+      "step": 33360
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011477665144257546,
+      "loss": 0.5169,
+      "step": 33370
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011472095354795587,
+      "loss": 0.6233,
+      "step": 33380
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0001146652556533363,
+      "loss": 0.5409,
+      "step": 33390
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011460955775871671,
+      "loss": 0.5444,
+      "step": 33400
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011455385986409712,
+      "loss": 0.5678,
+      "step": 33410
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011449816196947753,
+      "loss": 0.6457,
+      "step": 33420
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011444246407485797,
+      "loss": 0.507,
+      "step": 33430
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011438676618023838,
+      "loss": 0.5409,
+      "step": 33440
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011433106828561879,
+      "loss": 0.5237,
+      "step": 33450
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011427537039099922,
+      "loss": 0.5237,
+      "step": 33460
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011421967249637963,
+      "loss": 0.3855,
+      "step": 33470
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011416397460176004,
+      "loss": 0.6786,
+      "step": 33480
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011410827670714046,
+      "loss": 0.7491,
+      "step": 33490
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011405257881252087,
+      "loss": 0.5322,
+      "step": 33500
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011399688091790129,
+      "loss": 0.4671,
+      "step": 33510
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011394118302328172,
+      "loss": 0.6247,
+      "step": 33520
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011388548512866213,
+      "loss": 0.5359,
+      "step": 33530
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011382978723404255,
+      "loss": 0.5205,
+      "step": 33540
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011377408933942296,
+      "loss": 0.5354,
+      "step": 33550
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011371839144480338,
+      "loss": 0.4421,
+      "step": 33560
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011366269355018379,
+      "loss": 0.5371,
+      "step": 33570
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0001136069956555642,
+      "loss": 0.6531,
+      "step": 33580
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00011355129776094463,
+      "loss": 0.5245,
+      "step": 33590
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011349559986632504,
+      "loss": 0.5616,
+      "step": 33600
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011343990197170545,
+      "loss": 0.5483,
+      "step": 33610
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011338420407708589,
+      "loss": 0.5441,
+      "step": 33620
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0001133285061824663,
+      "loss": 0.6472,
+      "step": 33630
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011327280828784671,
+      "loss": 0.6369,
+      "step": 33640
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011321711039322714,
+      "loss": 0.6239,
+      "step": 33650
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011316141249860755,
+      "loss": 0.554,
+      "step": 33660
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011310571460398796,
+      "loss": 0.5735,
+      "step": 33670
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011305001670936837,
+      "loss": 0.6482,
+      "step": 33680
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0001129943188147488,
+      "loss": 0.6359,
+      "step": 33690
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0001129386209201292,
+      "loss": 0.5881,
+      "step": 33700
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011288292302550962,
+      "loss": 0.7353,
+      "step": 33710
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011282722513089005,
+      "loss": 0.555,
+      "step": 33720
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011277152723627047,
+      "loss": 0.6089,
+      "step": 33730
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011271582934165088,
+      "loss": 0.6418,
+      "step": 33740
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0001126601314470313,
+      "loss": 0.6514,
+      "step": 33750
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011260443355241171,
+      "loss": 0.5712,
+      "step": 33760
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011254873565779212,
+      "loss": 0.4586,
+      "step": 33770
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011249303776317255,
+      "loss": 0.554,
+      "step": 33780
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011243733986855296,
+      "loss": 0.5715,
+      "step": 33790
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011238164197393337,
+      "loss": 0.5781,
+      "step": 33800
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011232594407931378,
+      "loss": 0.6765,
+      "step": 33810
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0001122702461846942,
+      "loss": 0.5847,
+      "step": 33820
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011221454829007462,
+      "loss": 0.6121,
+      "step": 33830
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011215885039545504,
+      "loss": 0.4199,
+      "step": 33840
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011210315250083547,
+      "loss": 0.6374,
+      "step": 33850
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00011204745460621588,
+      "loss": 0.5438,
+      "step": 33860
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011199175671159629,
+      "loss": 0.5745,
+      "step": 33870
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011193605881697671,
+      "loss": 0.5881,
+      "step": 33880
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011188036092235713,
+      "loss": 0.7153,
+      "step": 33890
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011182466302773754,
+      "loss": 0.5935,
+      "step": 33900
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011176896513311796,
+      "loss": 0.5815,
+      "step": 33910
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011171326723849837,
+      "loss": 0.607,
+      "step": 33920
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011165756934387878,
+      "loss": 0.6031,
+      "step": 33930
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0001116018714492592,
+      "loss": 0.4774,
+      "step": 33940
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011154617355463963,
+      "loss": 0.6115,
+      "step": 33950
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011149047566002004,
+      "loss": 0.6182,
+      "step": 33960
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011143477776540045,
+      "loss": 0.6672,
+      "step": 33970
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011137907987078088,
+      "loss": 0.4804,
+      "step": 33980
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011132338197616129,
+      "loss": 0.6426,
+      "step": 33990
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0001112676840815417,
+      "loss": 0.6869,
+      "step": 34000
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011121198618692213,
+      "loss": 0.5712,
+      "step": 34010
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011115628829230254,
+      "loss": 0.5906,
+      "step": 34020
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011110059039768295,
+      "loss": 0.6791,
+      "step": 34030
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011104489250306339,
+      "loss": 0.548,
+      "step": 34040
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0001109891946084438,
+      "loss": 0.4906,
+      "step": 34050
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011093349671382421,
+      "loss": 0.5314,
+      "step": 34060
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011087779881920462,
+      "loss": 0.649,
+      "step": 34070
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011082210092458505,
+      "loss": 0.6003,
+      "step": 34080
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011076640302996546,
+      "loss": 0.5759,
+      "step": 34090
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011071070513534587,
+      "loss": 0.6643,
+      "step": 34100
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011065500724072629,
+      "loss": 0.7104,
+      "step": 34110
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0001105993093461067,
+      "loss": 0.5609,
+      "step": 34120
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00011054361145148711,
+      "loss": 0.5502,
+      "step": 34130
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00011048791355686755,
+      "loss": 0.4918,
+      "step": 34140
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00011043221566224796,
+      "loss": 0.6105,
+      "step": 34150
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00011037651776762837,
+      "loss": 0.5179,
+      "step": 34160
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0001103208198730088,
+      "loss": 0.5122,
+      "step": 34170
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00011026512197838921,
+      "loss": 0.6892,
+      "step": 34180
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00011020942408376962,
+      "loss": 0.7243,
+      "step": 34190
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00011015372618915003,
+      "loss": 0.5811,
+      "step": 34200
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00011009802829453046,
+      "loss": 0.5849,
+      "step": 34210
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00011004233039991087,
+      "loss": 0.4927,
+      "step": 34220
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00010998663250529128,
+      "loss": 0.5549,
+      "step": 34230
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00010993093461067172,
+      "loss": 0.5977,
+      "step": 34240
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00010987523671605213,
+      "loss": 0.6311,
+      "step": 34250
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00010981953882143254,
+      "loss": 0.5196,
+      "step": 34260
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00010976384092681297,
+      "loss": 0.5122,
+      "step": 34270
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00010970814303219338,
+      "loss": 0.5519,
+      "step": 34280
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00010965244513757379,
+      "loss": 0.6774,
+      "step": 34290
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00010959674724295421,
+      "loss": 0.553,
+      "step": 34300
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00010954104934833462,
+      "loss": 0.6607,
+      "step": 34310
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00010948535145371503,
+      "loss": 0.5861,
+      "step": 34320
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00010942965355909546,
+      "loss": 0.685,
+      "step": 34330
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00010937395566447587,
+      "loss": 0.6033,
+      "step": 34340
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00010931825776985628,
+      "loss": 0.6781,
+      "step": 34350
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0001092625598752367,
+      "loss": 0.7013,
+      "step": 34360
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00010920686198061713,
+      "loss": 0.5961,
+      "step": 34370
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00010915116408599754,
+      "loss": 0.489,
+      "step": 34380
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00010909546619137795,
+      "loss": 0.5715,
+      "step": 34390
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00010903976829675838,
+      "loss": 0.757,
+      "step": 34400
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010898407040213879,
+      "loss": 0.6089,
+      "step": 34410
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0001089283725075192,
+      "loss": 0.5625,
+      "step": 34420
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010887267461289962,
+      "loss": 0.6743,
+      "step": 34430
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010881697671828004,
+      "loss": 0.5399,
+      "step": 34440
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010876127882366045,
+      "loss": 0.4402,
+      "step": 34450
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010870558092904089,
+      "loss": 0.5848,
+      "step": 34460
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0001086498830344213,
+      "loss": 0.7084,
+      "step": 34470
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010859418513980171,
+      "loss": 0.5794,
+      "step": 34480
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010853848724518212,
+      "loss": 0.6234,
+      "step": 34490
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010848278935056254,
+      "loss": 0.607,
+      "step": 34500
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010842709145594295,
+      "loss": 0.6204,
+      "step": 34510
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010837139356132337,
+      "loss": 0.5963,
+      "step": 34520
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010831569566670379,
+      "loss": 0.5794,
+      "step": 34530
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0001082599977720842,
+      "loss": 0.4941,
+      "step": 34540
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010820429987746461,
+      "loss": 0.5723,
+      "step": 34550
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010814860198284505,
+      "loss": 0.5818,
+      "step": 34560
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010809290408822546,
+      "loss": 0.6557,
+      "step": 34570
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010803720619360587,
+      "loss": 0.6478,
+      "step": 34580
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0001079815082989863,
+      "loss": 0.4837,
+      "step": 34590
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010792581040436671,
+      "loss": 0.5239,
+      "step": 34600
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010787011250974712,
+      "loss": 0.503,
+      "step": 34610
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010781441461512753,
+      "loss": 0.6065,
+      "step": 34620
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010775871672050796,
+      "loss": 0.5867,
+      "step": 34630
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010770301882588837,
+      "loss": 0.5584,
+      "step": 34640
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010764732093126878,
+      "loss": 0.5843,
+      "step": 34650
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010759162303664922,
+      "loss": 0.6004,
+      "step": 34660
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00010753592514202963,
+      "loss": 0.6147,
+      "step": 34670
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010748022724741004,
+      "loss": 0.6261,
+      "step": 34680
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010742452935279046,
+      "loss": 0.69,
+      "step": 34690
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010736883145817087,
+      "loss": 0.6349,
+      "step": 34700
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010731313356355129,
+      "loss": 0.6343,
+      "step": 34710
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010725743566893171,
+      "loss": 0.5058,
+      "step": 34720
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010720173777431212,
+      "loss": 0.5787,
+      "step": 34730
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010714603987969253,
+      "loss": 0.5448,
+      "step": 34740
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010709034198507294,
+      "loss": 0.6968,
+      "step": 34750
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010703464409045338,
+      "loss": 0.787,
+      "step": 34760
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010697894619583379,
+      "loss": 0.5372,
+      "step": 34770
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0001069232483012142,
+      "loss": 0.5287,
+      "step": 34780
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010686755040659463,
+      "loss": 0.6151,
+      "step": 34790
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010681185251197504,
+      "loss": 0.7136,
+      "step": 34800
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010675615461735545,
+      "loss": 0.6034,
+      "step": 34810
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010670045672273588,
+      "loss": 0.5197,
+      "step": 34820
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010664475882811629,
+      "loss": 0.5162,
+      "step": 34830
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0001065890609334967,
+      "loss": 0.5379,
+      "step": 34840
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010653336303887712,
+      "loss": 0.5716,
+      "step": 34850
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010647766514425753,
+      "loss": 0.7303,
+      "step": 34860
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010642196724963794,
+      "loss": 0.6452,
+      "step": 34870
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010636626935501837,
+      "loss": 0.5527,
+      "step": 34880
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0001063105714603988,
+      "loss": 0.5434,
+      "step": 34890
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0001062548735657792,
+      "loss": 0.6569,
+      "step": 34900
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010619917567115962,
+      "loss": 0.6112,
+      "step": 34910
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010614347777654004,
+      "loss": 0.5214,
+      "step": 34920
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010608777988192045,
+      "loss": 0.5814,
+      "step": 34930
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00010603208198730086,
+      "loss": 0.5672,
+      "step": 34940
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010597638409268129,
+      "loss": 0.673,
+      "step": 34950
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0001059206861980617,
+      "loss": 0.6287,
+      "step": 34960
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010586498830344211,
+      "loss": 0.6001,
+      "step": 34970
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010580929040882255,
+      "loss": 0.6409,
+      "step": 34980
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010575359251420296,
+      "loss": 0.4455,
+      "step": 34990
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010569789461958337,
+      "loss": 0.5373,
+      "step": 35000
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010564219672496378,
+      "loss": 0.6713,
+      "step": 35010
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010558649883034421,
+      "loss": 0.6506,
+      "step": 35020
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010553080093572462,
+      "loss": 0.6333,
+      "step": 35030
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010547510304110503,
+      "loss": 0.572,
+      "step": 35040
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010541940514648545,
+      "loss": 0.4745,
+      "step": 35050
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010536370725186586,
+      "loss": 0.6048,
+      "step": 35060
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010530800935724628,
+      "loss": 0.6027,
+      "step": 35070
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010525231146262671,
+      "loss": 0.7227,
+      "step": 35080
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010519661356800713,
+      "loss": 0.7286,
+      "step": 35090
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010514091567338754,
+      "loss": 0.5431,
+      "step": 35100
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010508521777876796,
+      "loss": 0.6007,
+      "step": 35110
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010502951988414837,
+      "loss": 0.6274,
+      "step": 35120
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010497382198952878,
+      "loss": 0.7099,
+      "step": 35130
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0001049181240949092,
+      "loss": 0.6034,
+      "step": 35140
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010486242620028962,
+      "loss": 0.5529,
+      "step": 35150
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010480672830567003,
+      "loss": 0.6614,
+      "step": 35160
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010475103041105044,
+      "loss": 0.7438,
+      "step": 35170
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010469533251643088,
+      "loss": 0.7015,
+      "step": 35180
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010463963462181129,
+      "loss": 0.5895,
+      "step": 35190
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0001045839367271917,
+      "loss": 0.6446,
+      "step": 35200
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00010452823883257213,
+      "loss": 0.7804,
+      "step": 35210
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010447254093795254,
+      "loss": 0.6018,
+      "step": 35220
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010441684304333295,
+      "loss": 0.5574,
+      "step": 35230
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010436114514871337,
+      "loss": 0.7188,
+      "step": 35240
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010430544725409378,
+      "loss": 0.5537,
+      "step": 35250
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0001042497493594742,
+      "loss": 0.6568,
+      "step": 35260
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010419405146485463,
+      "loss": 0.6737,
+      "step": 35270
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010413835357023505,
+      "loss": 0.7045,
+      "step": 35280
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010408265567561546,
+      "loss": 0.5646,
+      "step": 35290
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010402695778099587,
+      "loss": 0.5411,
+      "step": 35300
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010397125988637629,
+      "loss": 0.5279,
+      "step": 35310
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0001039155619917567,
+      "loss": 0.5623,
+      "step": 35320
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010385986409713711,
+      "loss": 0.5809,
+      "step": 35330
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010380416620251754,
+      "loss": 0.5448,
+      "step": 35340
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010374846830789795,
+      "loss": 0.592,
+      "step": 35350
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010369277041327836,
+      "loss": 0.6807,
+      "step": 35360
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010363707251865879,
+      "loss": 0.5957,
+      "step": 35370
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0001035813746240392,
+      "loss": 0.5448,
+      "step": 35380
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010352567672941962,
+      "loss": 0.5674,
+      "step": 35390
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010346997883480005,
+      "loss": 0.6582,
+      "step": 35400
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010341428094018046,
+      "loss": 0.5759,
+      "step": 35410
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010335858304556087,
+      "loss": 0.678,
+      "step": 35420
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010330288515094128,
+      "loss": 0.4513,
+      "step": 35430
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0001032471872563217,
+      "loss": 0.6136,
+      "step": 35440
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010319148936170212,
+      "loss": 0.5059,
+      "step": 35450
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010313579146708253,
+      "loss": 0.6887,
+      "step": 35460
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010308009357246295,
+      "loss": 0.5988,
+      "step": 35470
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00010302439567784336,
+      "loss": 0.5616,
+      "step": 35480
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010296869778322377,
+      "loss": 0.5207,
+      "step": 35490
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010291299988860421,
+      "loss": 0.4487,
+      "step": 35500
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010285730199398462,
+      "loss": 0.7288,
+      "step": 35510
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010280160409936503,
+      "loss": 0.6065,
+      "step": 35520
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010274590620474546,
+      "loss": 0.6433,
+      "step": 35530
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010269020831012587,
+      "loss": 0.4878,
+      "step": 35540
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010263451041550628,
+      "loss": 0.6387,
+      "step": 35550
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010257881252088669,
+      "loss": 0.6496,
+      "step": 35560
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010252311462626712,
+      "loss": 0.594,
+      "step": 35570
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010246741673164753,
+      "loss": 0.6354,
+      "step": 35580
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010241171883702794,
+      "loss": 0.5754,
+      "step": 35590
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010235602094240838,
+      "loss": 0.6031,
+      "step": 35600
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010230032304778879,
+      "loss": 0.6351,
+      "step": 35610
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001022446251531692,
+      "loss": 0.5952,
+      "step": 35620
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010218892725854962,
+      "loss": 0.5913,
+      "step": 35630
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010213322936393004,
+      "loss": 0.5138,
+      "step": 35640
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010207753146931045,
+      "loss": 0.5812,
+      "step": 35650
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010202183357469087,
+      "loss": 0.5434,
+      "step": 35660
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010196613568007128,
+      "loss": 0.6442,
+      "step": 35670
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001019104377854517,
+      "loss": 0.6107,
+      "step": 35680
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001018547398908321,
+      "loss": 0.6868,
+      "step": 35690
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010179904199621254,
+      "loss": 0.6235,
+      "step": 35700
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010174334410159295,
+      "loss": 0.5671,
+      "step": 35710
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010168764620697337,
+      "loss": 0.5926,
+      "step": 35720
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00010163194831235379,
+      "loss": 0.5502,
+      "step": 35730
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0001015762504177342,
+      "loss": 0.6109,
+      "step": 35740
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010152055252311461,
+      "loss": 0.6107,
+      "step": 35750
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010146485462849504,
+      "loss": 0.776,
+      "step": 35760
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010140915673387545,
+      "loss": 0.5754,
+      "step": 35770
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010135345883925586,
+      "loss": 0.6007,
+      "step": 35780
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0001012977609446363,
+      "loss": 0.5029,
+      "step": 35790
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010124206305001671,
+      "loss": 0.6831,
+      "step": 35800
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010118636515539712,
+      "loss": 0.6357,
+      "step": 35810
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010113066726077753,
+      "loss": 0.6079,
+      "step": 35820
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010107496936615796,
+      "loss": 0.5962,
+      "step": 35830
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010101927147153837,
+      "loss": 0.6177,
+      "step": 35840
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010096357357691878,
+      "loss": 0.4786,
+      "step": 35850
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0001009078756822992,
+      "loss": 0.5954,
+      "step": 35860
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010085217778767961,
+      "loss": 0.645,
+      "step": 35870
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010079647989306003,
+      "loss": 0.5621,
+      "step": 35880
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010074078199844045,
+      "loss": 0.5971,
+      "step": 35890
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010068508410382086,
+      "loss": 0.5841,
+      "step": 35900
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010062938620920129,
+      "loss": 0.4793,
+      "step": 35910
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010057368831458171,
+      "loss": 0.4483,
+      "step": 35920
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010051799041996212,
+      "loss": 0.6261,
+      "step": 35930
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010046229252534253,
+      "loss": 0.6121,
+      "step": 35940
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010040659463072294,
+      "loss": 0.6607,
+      "step": 35950
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010035089673610337,
+      "loss": 0.546,
+      "step": 35960
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010029519884148378,
+      "loss": 0.6361,
+      "step": 35970
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010023950094686419,
+      "loss": 0.6022,
+      "step": 35980
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010018380305224462,
+      "loss": 0.5962,
+      "step": 35990
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010012810515762503,
+      "loss": 0.6776,
+      "step": 36000
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00010007240726300544,
+      "loss": 0.4473,
+      "step": 36010
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00010001670936838588,
+      "loss": 0.5978,
+      "step": 36020
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.996101147376629e-05,
+      "loss": 0.5636,
+      "step": 36030
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.99053135791467e-05,
+      "loss": 0.5971,
+      "step": 36040
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.984961568452712e-05,
+      "loss": 0.6218,
+      "step": 36050
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.979391778990753e-05,
+      "loss": 0.4685,
+      "step": 36060
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.973821989528795e-05,
+      "loss": 0.5591,
+      "step": 36070
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.968252200066836e-05,
+      "loss": 0.4999,
+      "step": 36080
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.962682410604878e-05,
+      "loss": 0.5702,
+      "step": 36090
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.957112621142919e-05,
+      "loss": 0.503,
+      "step": 36100
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.95154283168096e-05,
+      "loss": 0.5753,
+      "step": 36110
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.945973042219004e-05,
+      "loss": 0.5989,
+      "step": 36120
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.940403252757045e-05,
+      "loss": 0.5798,
+      "step": 36130
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.934833463295086e-05,
+      "loss": 0.5491,
+      "step": 36140
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.929263673833129e-05,
+      "loss": 0.4819,
+      "step": 36150
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.92369388437117e-05,
+      "loss": 0.6227,
+      "step": 36160
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.918124094909211e-05,
+      "loss": 0.6807,
+      "step": 36170
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.912554305447254e-05,
+      "loss": 0.5521,
+      "step": 36180
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.906984515985295e-05,
+      "loss": 0.5717,
+      "step": 36190
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.901414726523336e-05,
+      "loss": 0.5666,
+      "step": 36200
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.89584493706138e-05,
+      "loss": 0.5859,
+      "step": 36210
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.890275147599421e-05,
+      "loss": 0.6087,
+      "step": 36220
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.884705358137462e-05,
+      "loss": 0.6007,
+      "step": 36230
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.879135568675503e-05,
+      "loss": 0.5538,
+      "step": 36240
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.873565779213545e-05,
+      "loss": 0.5587,
+      "step": 36250
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.867995989751587e-05,
+      "loss": 0.5805,
+      "step": 36260
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.862426200289628e-05,
+      "loss": 0.5035,
+      "step": 36270
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.85685641082767e-05,
+      "loss": 0.5851,
+      "step": 36280
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.851286621365711e-05,
+      "loss": 0.6262,
+      "step": 36290
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.845716831903752e-05,
+      "loss": 0.4777,
+      "step": 36300
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.840147042441796e-05,
+      "loss": 0.4962,
+      "step": 36310
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.834577252979837e-05,
+      "loss": 0.4893,
+      "step": 36320
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.829007463517878e-05,
+      "loss": 0.541,
+      "step": 36330
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.823437674055921e-05,
+      "loss": 0.6842,
+      "step": 36340
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.817867884593962e-05,
+      "loss": 0.6965,
+      "step": 36350
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.812298095132003e-05,
+      "loss": 0.4909,
+      "step": 36360
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.806728305670044e-05,
+      "loss": 0.5554,
+      "step": 36370
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.801158516208087e-05,
+      "loss": 0.589,
+      "step": 36380
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.795588726746128e-05,
+      "loss": 0.6397,
+      "step": 36390
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.790018937284169e-05,
+      "loss": 0.5537,
+      "step": 36400
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.784449147822211e-05,
+      "loss": 0.6147,
+      "step": 36410
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.778879358360254e-05,
+      "loss": 0.7166,
+      "step": 36420
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.773309568898295e-05,
+      "loss": 0.592,
+      "step": 36430
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.767739779436337e-05,
+      "loss": 0.5838,
+      "step": 36440
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.762169989974379e-05,
+      "loss": 0.571,
+      "step": 36450
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.75660020051242e-05,
+      "loss": 0.7476,
+      "step": 36460
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.751030411050462e-05,
+      "loss": 0.4528,
+      "step": 36470
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.745460621588503e-05,
+      "loss": 0.554,
+      "step": 36480
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.739890832126544e-05,
+      "loss": 0.5297,
+      "step": 36490
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.734321042664585e-05,
+      "loss": 0.6112,
+      "step": 36500
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.728751253202628e-05,
+      "loss": 0.4328,
+      "step": 36510
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.723181463740669e-05,
+      "loss": 0.5773,
+      "step": 36520
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.71761167427871e-05,
+      "loss": 0.6562,
+      "step": 36530
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.712041884816754e-05,
+      "loss": 0.5885,
+      "step": 36540
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.706472095354795e-05,
+      "loss": 0.5799,
+      "step": 36550
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.700902305892836e-05,
+      "loss": 0.5385,
+      "step": 36560
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.695332516430879e-05,
+      "loss": 0.7173,
+      "step": 36570
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.68976272696892e-05,
+      "loss": 0.4672,
+      "step": 36580
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.684192937506961e-05,
+      "loss": 0.666,
+      "step": 36590
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.678623148045003e-05,
+      "loss": 0.5006,
+      "step": 36600
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.673053358583044e-05,
+      "loss": 0.5442,
+      "step": 36610
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.667483569121086e-05,
+      "loss": 0.5064,
+      "step": 36620
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.662470758605323e-05,
+      "loss": 0.5305,
+      "step": 36630
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.656900969143367e-05,
+      "loss": 0.575,
+      "step": 36640
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.651331179681408e-05,
+      "loss": 0.4911,
+      "step": 36650
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.645761390219449e-05,
+      "loss": 0.5857,
+      "step": 36660
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.64019160075749e-05,
+      "loss": 0.5431,
+      "step": 36670
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.634621811295532e-05,
+      "loss": 0.5609,
+      "step": 36680
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.629052021833574e-05,
+      "loss": 0.6122,
+      "step": 36690
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.623482232371615e-05,
+      "loss": 0.6579,
+      "step": 36700
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.617912442909657e-05,
+      "loss": 0.6233,
+      "step": 36710
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.612342653447698e-05,
+      "loss": 0.6777,
+      "step": 36720
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.60677286398574e-05,
+      "loss": 0.5071,
+      "step": 36730
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.601203074523783e-05,
+      "loss": 0.6418,
+      "step": 36740
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.595633285061824e-05,
+      "loss": 0.5162,
+      "step": 36750
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.590063495599865e-05,
+      "loss": 0.6013,
+      "step": 36760
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.584493706137908e-05,
+      "loss": 0.487,
+      "step": 36770
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.578923916675949e-05,
+      "loss": 0.5323,
+      "step": 36780
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.57335412721399e-05,
+      "loss": 0.5625,
+      "step": 36790
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.567784337752031e-05,
+      "loss": 0.5449,
+      "step": 36800
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.562214548290074e-05,
+      "loss": 0.653,
+      "step": 36810
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.556644758828115e-05,
+      "loss": 0.5442,
+      "step": 36820
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.551074969366156e-05,
+      "loss": 0.745,
+      "step": 36830
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.5455051799042e-05,
+      "loss": 0.5711,
+      "step": 36840
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.539935390442241e-05,
+      "loss": 0.643,
+      "step": 36850
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.534365600980282e-05,
+      "loss": 0.6256,
+      "step": 36860
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.528795811518324e-05,
+      "loss": 0.6328,
+      "step": 36870
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.523226022056366e-05,
+      "loss": 0.6809,
+      "step": 36880
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.517656232594407e-05,
+      "loss": 0.5442,
+      "step": 36890
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.512086443132449e-05,
+      "loss": 0.5729,
+      "step": 36900
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.50651665367049e-05,
+      "loss": 0.5166,
+      "step": 36910
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.500946864208531e-05,
+      "loss": 0.4517,
+      "step": 36920
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.495377074746573e-05,
+      "loss": 0.6478,
+      "step": 36930
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.489807285284616e-05,
+      "loss": 0.5478,
+      "step": 36940
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.484237495822657e-05,
+      "loss": 0.5259,
+      "step": 36950
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.478667706360699e-05,
+      "loss": 0.5417,
+      "step": 36960
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.473097916898741e-05,
+      "loss": 0.5827,
+      "step": 36970
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.467528127436782e-05,
+      "loss": 0.609,
+      "step": 36980
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.461958337974823e-05,
+      "loss": 0.6195,
+      "step": 36990
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.456388548512866e-05,
+      "loss": 0.6875,
+      "step": 37000
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.450818759050907e-05,
+      "loss": 0.6355,
+      "step": 37010
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.445248969588948e-05,
+      "loss": 0.547,
+      "step": 37020
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.43967918012699e-05,
+      "loss": 0.4833,
+      "step": 37030
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.434109390665033e-05,
+      "loss": 0.7239,
+      "step": 37040
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.428539601203074e-05,
+      "loss": 0.5379,
+      "step": 37050
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.422969811741115e-05,
+      "loss": 0.6801,
+      "step": 37060
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.417400022279158e-05,
+      "loss": 0.4889,
+      "step": 37070
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.411830232817199e-05,
+      "loss": 0.6934,
+      "step": 37080
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.40626044335524e-05,
+      "loss": 0.6061,
+      "step": 37090
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.400690653893282e-05,
+      "loss": 0.4343,
+      "step": 37100
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.395120864431323e-05,
+      "loss": 0.6552,
+      "step": 37110
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.389551074969365e-05,
+      "loss": 0.6288,
+      "step": 37120
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.383981285507407e-05,
+      "loss": 0.4945,
+      "step": 37130
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.378411496045448e-05,
+      "loss": 0.492,
+      "step": 37140
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.372841706583489e-05,
+      "loss": 0.5001,
+      "step": 37150
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.367271917121533e-05,
+      "loss": 0.5464,
+      "step": 37160
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.361702127659574e-05,
+      "loss": 0.5427,
+      "step": 37170
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.356132338197615e-05,
+      "loss": 0.5644,
+      "step": 37180
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.350562548735658e-05,
+      "loss": 0.5245,
+      "step": 37190
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.344992759273699e-05,
+      "loss": 0.6931,
+      "step": 37200
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.33942296981174e-05,
+      "loss": 0.5938,
+      "step": 37210
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.333853180349781e-05,
+      "loss": 0.502,
+      "step": 37220
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.328283390887824e-05,
+      "loss": 0.5595,
+      "step": 37230
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.322713601425865e-05,
+      "loss": 0.6337,
+      "step": 37240
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.317143811963906e-05,
+      "loss": 0.5674,
+      "step": 37250
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.31157402250195e-05,
+      "loss": 0.6151,
+      "step": 37260
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.306004233039991e-05,
+      "loss": 0.6653,
+      "step": 37270
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.300434443578032e-05,
+      "loss": 0.566,
+      "step": 37280
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.294864654116074e-05,
+      "loss": 0.629,
+      "step": 37290
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.289294864654115e-05,
+      "loss": 0.6316,
+      "step": 37300
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.284282054138353e-05,
+      "loss": 0.6943,
+      "step": 37310
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.278712264676394e-05,
+      "loss": 0.5991,
+      "step": 37320
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.273142475214436e-05,
+      "loss": 0.5962,
+      "step": 37330
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.267572685752477e-05,
+      "loss": 0.6278,
+      "step": 37340
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.262002896290519e-05,
+      "loss": 0.584,
+      "step": 37350
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.256433106828562e-05,
+      "loss": 0.5016,
+      "step": 37360
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.250863317366603e-05,
+      "loss": 0.7003,
+      "step": 37370
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.245293527904645e-05,
+      "loss": 0.4483,
+      "step": 37380
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.239723738442686e-05,
+      "loss": 0.52,
+      "step": 37390
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.234153948980728e-05,
+      "loss": 0.588,
+      "step": 37400
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.228584159518769e-05,
+      "loss": 0.5803,
+      "step": 37410
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.22301437005681e-05,
+      "loss": 0.5044,
+      "step": 37420
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.217444580594853e-05,
+      "loss": 0.4983,
+      "step": 37430
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.211874791132894e-05,
+      "loss": 0.5265,
+      "step": 37440
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.206305001670935e-05,
+      "loss": 0.5209,
+      "step": 37450
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.200735212208979e-05,
+      "loss": 0.4863,
+      "step": 37460
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.19516542274702e-05,
+      "loss": 0.7458,
+      "step": 37470
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.189595633285061e-05,
+      "loss": 0.6365,
+      "step": 37480
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.184025843823104e-05,
+      "loss": 0.6431,
+      "step": 37490
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.178456054361145e-05,
+      "loss": 0.6689,
+      "step": 37500
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.172886264899186e-05,
+      "loss": 0.5844,
+      "step": 37510
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.167316475437227e-05,
+      "loss": 0.6368,
+      "step": 37520
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.16174668597527e-05,
+      "loss": 0.6164,
+      "step": 37530
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.15617689651331e-05,
+      "loss": 0.6848,
+      "step": 37540
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.150607107051352e-05,
+      "loss": 0.6895,
+      "step": 37550
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.145037317589395e-05,
+      "loss": 0.6693,
+      "step": 37560
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.139467528127437e-05,
+      "loss": 0.5838,
+      "step": 37570
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.133897738665478e-05,
+      "loss": 0.6268,
+      "step": 37580
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.12832794920352e-05,
+      "loss": 0.5345,
+      "step": 37590
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.122758159741561e-05,
+      "loss": 0.6347,
+      "step": 37600
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.117188370279602e-05,
+      "loss": 0.6112,
+      "step": 37610
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.111618580817645e-05,
+      "loss": 0.5432,
+      "step": 37620
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.106048791355686e-05,
+      "loss": 0.5615,
+      "step": 37630
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.100479001893727e-05,
+      "loss": 0.6325,
+      "step": 37640
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.094909212431768e-05,
+      "loss": 0.6287,
+      "step": 37650
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.089339422969812e-05,
+      "loss": 0.6164,
+      "step": 37660
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.083769633507853e-05,
+      "loss": 0.5449,
+      "step": 37670
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.078199844045894e-05,
+      "loss": 0.5394,
+      "step": 37680
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.072630054583937e-05,
+      "loss": 0.5274,
+      "step": 37690
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.067060265121978e-05,
+      "loss": 0.5024,
+      "step": 37700
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.061490475660019e-05,
+      "loss": 0.5863,
+      "step": 37710
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.055920686198061e-05,
+      "loss": 0.5523,
+      "step": 37720
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.050350896736103e-05,
+      "loss": 0.6241,
+      "step": 37730
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.044781107274144e-05,
+      "loss": 0.5805,
+      "step": 37740
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.039211317812186e-05,
+      "loss": 0.6348,
+      "step": 37750
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.033641528350227e-05,
+      "loss": 0.5918,
+      "step": 37760
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.028071738888268e-05,
+      "loss": 0.6979,
+      "step": 37770
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.022501949426311e-05,
+      "loss": 0.4575,
+      "step": 37780
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.016932159964353e-05,
+      "loss": 0.4958,
+      "step": 37790
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.011362370502394e-05,
+      "loss": 0.5361,
+      "step": 37800
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.005792581040435e-05,
+      "loss": 0.599,
+      "step": 37810
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.000222791578478e-05,
+      "loss": 0.589,
+      "step": 37820
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.994653002116519e-05,
+      "loss": 0.7167,
+      "step": 37830
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.98908321265456e-05,
+      "loss": 0.6201,
+      "step": 37840
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.983513423192603e-05,
+      "loss": 0.4408,
+      "step": 37850
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.977943633730644e-05,
+      "loss": 0.624,
+      "step": 37860
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.972373844268685e-05,
+      "loss": 0.5171,
+      "step": 37870
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.966804054806729e-05,
+      "loss": 0.6237,
+      "step": 37880
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.96123426534477e-05,
+      "loss": 0.4729,
+      "step": 37890
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.955664475882811e-05,
+      "loss": 0.6422,
+      "step": 37900
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.950094686420852e-05,
+      "loss": 0.5589,
+      "step": 37910
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.944524896958895e-05,
+      "loss": 0.5995,
+      "step": 37920
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.938955107496936e-05,
+      "loss": 0.4762,
+      "step": 37930
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.933385318034977e-05,
+      "loss": 0.5674,
+      "step": 37940
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.927815528573019e-05,
+      "loss": 0.6598,
+      "step": 37950
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.92224573911106e-05,
+      "loss": 0.512,
+      "step": 37960
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.916675949649101e-05,
+      "loss": 0.5665,
+      "step": 37970
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.911106160187145e-05,
+      "loss": 0.472,
+      "step": 37980
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.905536370725186e-05,
+      "loss": 0.5788,
+      "step": 37990
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.899966581263227e-05,
+      "loss": 0.5859,
+      "step": 38000
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.89439679180127e-05,
+      "loss": 0.5586,
+      "step": 38010
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.888827002339311e-05,
+      "loss": 0.4824,
+      "step": 38020
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.883257212877352e-05,
+      "loss": 0.5916,
+      "step": 38030
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.877687423415393e-05,
+      "loss": 0.6081,
+      "step": 38040
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.872117633953436e-05,
+      "loss": 0.6108,
+      "step": 38050
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.866547844491477e-05,
+      "loss": 0.6746,
+      "step": 38060
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.860978055029518e-05,
+      "loss": 0.6062,
+      "step": 38070
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.855408265567562e-05,
+      "loss": 0.658,
+      "step": 38080
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.849838476105603e-05,
+      "loss": 0.5576,
+      "step": 38090
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.844268686643644e-05,
+      "loss": 0.6388,
+      "step": 38100
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.838698897181687e-05,
+      "loss": 0.5272,
+      "step": 38110
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.833129107719728e-05,
+      "loss": 0.5783,
+      "step": 38120
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.827559318257769e-05,
+      "loss": 0.6013,
+      "step": 38130
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.821989528795811e-05,
+      "loss": 0.6183,
+      "step": 38140
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.816419739333852e-05,
+      "loss": 0.5121,
+      "step": 38150
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.810849949871893e-05,
+      "loss": 0.7646,
+      "step": 38160
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.805280160409936e-05,
+      "loss": 0.7093,
+      "step": 38170
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.799710370947978e-05,
+      "loss": 0.7307,
+      "step": 38180
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.79414058148602e-05,
+      "loss": 0.578,
+      "step": 38190
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.78857079202406e-05,
+      "loss": 0.6395,
+      "step": 38200
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.783001002562103e-05,
+      "loss": 0.4526,
+      "step": 38210
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.777431213100144e-05,
+      "loss": 0.7063,
+      "step": 38220
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.771861423638185e-05,
+      "loss": 0.4838,
+      "step": 38230
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.766291634176228e-05,
+      "loss": 0.5524,
+      "step": 38240
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.760721844714269e-05,
+      "loss": 0.4842,
+      "step": 38250
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.75515205525231e-05,
+      "loss": 0.5478,
+      "step": 38260
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.749582265790352e-05,
+      "loss": 0.6879,
+      "step": 38270
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.744012476328394e-05,
+      "loss": 0.6129,
+      "step": 38280
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.738442686866435e-05,
+      "loss": 0.7345,
+      "step": 38290
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.732872897404479e-05,
+      "loss": 0.6483,
+      "step": 38300
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.72730310794252e-05,
+      "loss": 0.5696,
+      "step": 38310
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.721733318480561e-05,
+      "loss": 0.562,
+      "step": 38320
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.716163529018602e-05,
+      "loss": 0.5372,
+      "step": 38330
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.710593739556644e-05,
+      "loss": 0.5587,
+      "step": 38340
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.705023950094685e-05,
+      "loss": 0.6179,
+      "step": 38350
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.699454160632727e-05,
+      "loss": 0.5569,
+      "step": 38360
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.693884371170769e-05,
+      "loss": 0.5788,
+      "step": 38370
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.68831458170881e-05,
+      "loss": 0.573,
+      "step": 38380
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.682744792246851e-05,
+      "loss": 0.6139,
+      "step": 38390
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.677175002784895e-05,
+      "loss": 0.6121,
+      "step": 38400
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.671605213322936e-05,
+      "loss": 0.5259,
+      "step": 38410
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.666035423860977e-05,
+      "loss": 0.588,
+      "step": 38420
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.66046563439902e-05,
+      "loss": 0.4979,
+      "step": 38430
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.654895844937061e-05,
+      "loss": 0.6504,
+      "step": 38440
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.649326055475102e-05,
+      "loss": 0.6656,
+      "step": 38450
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.643756266013143e-05,
+      "loss": 0.5916,
+      "step": 38460
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.638186476551186e-05,
+      "loss": 0.5339,
+      "step": 38470
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.632616687089227e-05,
+      "loss": 0.6233,
+      "step": 38480
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.627046897627268e-05,
+      "loss": 0.5343,
+      "step": 38490
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.621477108165312e-05,
+      "loss": 0.5898,
+      "step": 38500
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.615907318703353e-05,
+      "loss": 0.4948,
+      "step": 38510
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.610337529241394e-05,
+      "loss": 0.6933,
+      "step": 38520
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.604767739779436e-05,
+      "loss": 0.5161,
+      "step": 38530
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.599197950317477e-05,
+      "loss": 0.6217,
+      "step": 38540
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.593628160855519e-05,
+      "loss": 0.5585,
+      "step": 38550
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.588058371393561e-05,
+      "loss": 0.7432,
+      "step": 38560
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.582488581931602e-05,
+      "loss": 0.7917,
+      "step": 38570
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.576918792469643e-05,
+      "loss": 0.6663,
+      "step": 38580
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.571349003007684e-05,
+      "loss": 0.6414,
+      "step": 38590
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.565779213545728e-05,
+      "loss": 0.6547,
+      "step": 38600
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.560209424083769e-05,
+      "loss": 0.4906,
+      "step": 38610
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.55463963462181e-05,
+      "loss": 0.6133,
+      "step": 38620
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.549069845159853e-05,
+      "loss": 0.6114,
+      "step": 38630
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.543500055697894e-05,
+      "loss": 0.6121,
+      "step": 38640
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.537930266235935e-05,
+      "loss": 0.5713,
+      "step": 38650
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.532360476773978e-05,
+      "loss": 0.6192,
+      "step": 38660
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.526790687312019e-05,
+      "loss": 0.5984,
+      "step": 38670
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.52122089785006e-05,
+      "loss": 0.5022,
+      "step": 38680
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.515651108388102e-05,
+      "loss": 0.594,
+      "step": 38690
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.510081318926145e-05,
+      "loss": 0.6233,
+      "step": 38700
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.504511529464186e-05,
+      "loss": 0.49,
+      "step": 38710
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.498941740002227e-05,
+      "loss": 0.5534,
+      "step": 38720
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.49337195054027e-05,
+      "loss": 0.5978,
+      "step": 38730
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.48780216107831e-05,
+      "loss": 0.5747,
+      "step": 38740
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.482232371616352e-05,
+      "loss": 0.6752,
+      "step": 38750
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.476662582154394e-05,
+      "loss": 0.6899,
+      "step": 38760
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.471092792692435e-05,
+      "loss": 0.4496,
+      "step": 38770
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.465523003230476e-05,
+      "loss": 0.4616,
+      "step": 38780
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.459953213768519e-05,
+      "loss": 0.6139,
+      "step": 38790
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.45438342430656e-05,
+      "loss": 0.4745,
+      "step": 38800
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.448813634844601e-05,
+      "loss": 0.6267,
+      "step": 38810
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.443243845382645e-05,
+      "loss": 0.4762,
+      "step": 38820
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.437674055920686e-05,
+      "loss": 0.6528,
+      "step": 38830
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.432104266458727e-05,
+      "loss": 0.6377,
+      "step": 38840
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.426534476996768e-05,
+      "loss": 0.5159,
+      "step": 38850
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.420964687534811e-05,
+      "loss": 0.4913,
+      "step": 38860
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.415394898072852e-05,
+      "loss": 0.7115,
+      "step": 38870
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.409825108610893e-05,
+      "loss": 0.4628,
+      "step": 38880
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.404255319148935e-05,
+      "loss": 0.6262,
+      "step": 38890
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.398685529686976e-05,
+      "loss": 0.6766,
+      "step": 38900
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.393115740225018e-05,
+      "loss": 0.4706,
+      "step": 38910
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.387545950763061e-05,
+      "loss": 0.646,
+      "step": 38920
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.381976161301103e-05,
+      "loss": 0.5226,
+      "step": 38930
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.376406371839144e-05,
+      "loss": 0.4093,
+      "step": 38940
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.370836582377186e-05,
+      "loss": 0.5973,
+      "step": 38950
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.365266792915227e-05,
+      "loss": 0.7618,
+      "step": 38960
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.359697003453268e-05,
+      "loss": 0.6285,
+      "step": 38970
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.35412721399131e-05,
+      "loss": 0.5303,
+      "step": 38980
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.348557424529352e-05,
+      "loss": 0.6095,
+      "step": 38990
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.342987635067393e-05,
+      "loss": 0.5461,
+      "step": 39000
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.337417845605434e-05,
+      "loss": 0.5979,
+      "step": 39010
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.331848056143478e-05,
+      "loss": 0.6598,
+      "step": 39020
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.326278266681519e-05,
+      "loss": 0.5353,
+      "step": 39030
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.32070847721956e-05,
+      "loss": 0.7684,
+      "step": 39040
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.315138687757603e-05,
+      "loss": 0.5507,
+      "step": 39050
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.309568898295644e-05,
+      "loss": 0.5111,
+      "step": 39060
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.303999108833685e-05,
+      "loss": 0.4786,
+      "step": 39070
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.298429319371727e-05,
+      "loss": 0.5672,
+      "step": 39080
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.292859529909768e-05,
+      "loss": 0.7179,
+      "step": 39090
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.28728974044781e-05,
+      "loss": 0.5183,
+      "step": 39100
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.281719950985853e-05,
+      "loss": 0.4978,
+      "step": 39110
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.276150161523895e-05,
+      "loss": 0.5553,
+      "step": 39120
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.270580372061936e-05,
+      "loss": 0.5143,
+      "step": 39130
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.265010582599977e-05,
+      "loss": 0.5938,
+      "step": 39140
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.259440793138019e-05,
+      "loss": 0.579,
+      "step": 39150
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.25387100367606e-05,
+      "loss": 0.6513,
+      "step": 39160
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.248301214214101e-05,
+      "loss": 0.527,
+      "step": 39170
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.242731424752144e-05,
+      "loss": 0.5359,
+      "step": 39180
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.237161635290185e-05,
+      "loss": 0.5852,
+      "step": 39190
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.231591845828226e-05,
+      "loss": 0.6273,
+      "step": 39200
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.22602205636627e-05,
+      "loss": 0.5162,
+      "step": 39210
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.220452266904311e-05,
+      "loss": 0.5746,
+      "step": 39220
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.214882477442352e-05,
+      "loss": 0.4622,
+      "step": 39230
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.209312687980395e-05,
+      "loss": 0.5128,
+      "step": 39240
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.203742898518436e-05,
+      "loss": 0.5382,
+      "step": 39250
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.198173109056477e-05,
+      "loss": 0.6015,
+      "step": 39260
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.192603319594518e-05,
+      "loss": 0.7172,
+      "step": 39270
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.18703353013256e-05,
+      "loss": 0.6598,
+      "step": 39280
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.181463740670602e-05,
+      "loss": 0.6191,
+      "step": 39290
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.175893951208643e-05,
+      "loss": 0.5241,
+      "step": 39300
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.170324161746685e-05,
+      "loss": 0.5757,
+      "step": 39310
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.164754372284726e-05,
+      "loss": 0.5595,
+      "step": 39320
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.159184582822769e-05,
+      "loss": 0.5609,
+      "step": 39330
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.153614793360811e-05,
+      "loss": 0.6168,
+      "step": 39340
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.148045003898852e-05,
+      "loss": 0.5133,
+      "step": 39350
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.142475214436893e-05,
+      "loss": 0.6391,
+      "step": 39360
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.136905424974936e-05,
+      "loss": 0.4574,
+      "step": 39370
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.131335635512977e-05,
+      "loss": 0.7241,
+      "step": 39380
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.125765846051018e-05,
+      "loss": 0.5844,
+      "step": 39390
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.120196056589059e-05,
+      "loss": 0.6103,
+      "step": 39400
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.114626267127102e-05,
+      "loss": 0.5897,
+      "step": 39410
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.109056477665143e-05,
+      "loss": 0.5035,
+      "step": 39420
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.103486688203184e-05,
+      "loss": 0.5493,
+      "step": 39430
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.097916898741228e-05,
+      "loss": 0.478,
+      "step": 39440
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.092347109279269e-05,
+      "loss": 0.5897,
+      "step": 39450
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.08677731981731e-05,
+      "loss": 0.5199,
+      "step": 39460
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.081207530355352e-05,
+      "loss": 0.6233,
+      "step": 39470
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.075637740893394e-05,
+      "loss": 0.6201,
+      "step": 39480
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.070067951431435e-05,
+      "loss": 0.7711,
+      "step": 39490
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.064498161969477e-05,
+      "loss": 0.5393,
+      "step": 39500
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.058928372507518e-05,
+      "loss": 0.6425,
+      "step": 39510
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.05335858304556e-05,
+      "loss": 0.4941,
+      "step": 39520
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.0477887935836e-05,
+      "loss": 0.6082,
+      "step": 39530
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.042219004121644e-05,
+      "loss": 0.5083,
+      "step": 39540
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.036649214659685e-05,
+      "loss": 0.5724,
+      "step": 39550
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.031079425197727e-05,
+      "loss": 0.5116,
+      "step": 39560
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.025509635735769e-05,
+      "loss": 0.6118,
+      "step": 39570
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.01993984627381e-05,
+      "loss": 0.5039,
+      "step": 39580
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.014370056811851e-05,
+      "loss": 0.5212,
+      "step": 39590
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.008800267349894e-05,
+      "loss": 0.5142,
+      "step": 39600
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.003230477887935e-05,
+      "loss": 0.575,
+      "step": 39610
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.997660688425976e-05,
+      "loss": 0.5748,
+      "step": 39620
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.99209089896402e-05,
+      "loss": 0.5913,
+      "step": 39630
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.986521109502061e-05,
+      "loss": 0.5368,
+      "step": 39640
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.980951320040102e-05,
+      "loss": 0.6079,
+      "step": 39650
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.975381530578143e-05,
+      "loss": 0.6189,
+      "step": 39660
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.969811741116186e-05,
+      "loss": 0.5868,
+      "step": 39670
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.964241951654227e-05,
+      "loss": 0.551,
+      "step": 39680
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.958672162192268e-05,
+      "loss": 0.5548,
+      "step": 39690
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.95310237273031e-05,
+      "loss": 0.6888,
+      "step": 39700
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.947532583268351e-05,
+      "loss": 0.6626,
+      "step": 39710
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.941962793806393e-05,
+      "loss": 0.5973,
+      "step": 39720
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.936393004344436e-05,
+      "loss": 0.5248,
+      "step": 39730
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.930823214882477e-05,
+      "loss": 0.5731,
+      "step": 39740
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.925253425420519e-05,
+      "loss": 0.7173,
+      "step": 39750
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.919683635958561e-05,
+      "loss": 0.4956,
+      "step": 39760
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.914113846496602e-05,
+      "loss": 0.6819,
+      "step": 39770
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.908544057034643e-05,
+      "loss": 0.5895,
+      "step": 39780
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 7.902974267572684e-05,
+      "loss": 0.6268,
+      "step": 39790
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.897404478110727e-05,
+      "loss": 0.6219,
+      "step": 39800
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.891834688648768e-05,
+      "loss": 0.6172,
+      "step": 39810
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.886264899186809e-05,
+      "loss": 0.6433,
+      "step": 39820
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.880695109724852e-05,
+      "loss": 0.597,
+      "step": 39830
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.875125320262893e-05,
+      "loss": 0.5338,
+      "step": 39840
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.869555530800935e-05,
+      "loss": 0.6401,
+      "step": 39850
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.863985741338978e-05,
+      "loss": 0.5635,
+      "step": 39860
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.858415951877019e-05,
+      "loss": 0.6666,
+      "step": 39870
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.85284616241506e-05,
+      "loss": 0.5134,
+      "step": 39880
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.847276372953102e-05,
+      "loss": 0.7161,
+      "step": 39890
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.841706583491143e-05,
+      "loss": 0.4848,
+      "step": 39900
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.836136794029185e-05,
+      "loss": 0.7374,
+      "step": 39910
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.830567004567226e-05,
+      "loss": 0.5928,
+      "step": 39920
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.824997215105268e-05,
+      "loss": 0.6739,
+      "step": 39930
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.819427425643309e-05,
+      "loss": 0.5335,
+      "step": 39940
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.81385763618135e-05,
+      "loss": 0.4856,
+      "step": 39950
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.808287846719394e-05,
+      "loss": 0.6146,
+      "step": 39960
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.802718057257435e-05,
+      "loss": 0.6421,
+      "step": 39970
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.797148267795476e-05,
+      "loss": 0.503,
+      "step": 39980
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.791578478333519e-05,
+      "loss": 0.5727,
+      "step": 39990
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.78600868887156e-05,
+      "loss": 0.4706,
+      "step": 40000
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.780438899409601e-05,
+      "loss": 0.4936,
+      "step": 40010
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.774869109947644e-05,
+      "loss": 0.6032,
+      "step": 40020
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.769299320485685e-05,
+      "loss": 0.5765,
+      "step": 40030
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.763729531023726e-05,
+      "loss": 0.5574,
+      "step": 40040
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.75815974156177e-05,
+      "loss": 0.6039,
+      "step": 40050
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.752589952099811e-05,
+      "loss": 0.5606,
+      "step": 40060
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.747020162637852e-05,
+      "loss": 0.5112,
+      "step": 40070
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.741450373175893e-05,
+      "loss": 0.5822,
+      "step": 40080
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.735880583713935e-05,
+      "loss": 0.7732,
+      "step": 40090
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.730310794251977e-05,
+      "loss": 0.5575,
+      "step": 40100
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.724741004790018e-05,
+      "loss": 0.6468,
+      "step": 40110
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.71917121532806e-05,
+      "loss": 0.5499,
+      "step": 40120
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.713601425866101e-05,
+      "loss": 0.6773,
+      "step": 40130
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.708031636404142e-05,
+      "loss": 0.6287,
+      "step": 40140
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.702461846942186e-05,
+      "loss": 0.4615,
+      "step": 40150
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.696892057480227e-05,
+      "loss": 0.5807,
+      "step": 40160
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.691322268018268e-05,
+      "loss": 0.5307,
+      "step": 40170
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.685752478556311e-05,
+      "loss": 0.6332,
+      "step": 40180
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.680182689094352e-05,
+      "loss": 0.4301,
+      "step": 40190
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.674612899632393e-05,
+      "loss": 0.439,
+      "step": 40200
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.669043110170434e-05,
+      "loss": 0.5922,
+      "step": 40210
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.663473320708477e-05,
+      "loss": 0.5614,
+      "step": 40220
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.658460510192714e-05,
+      "loss": 0.552,
+      "step": 40230
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.652890720730755e-05,
+      "loss": 0.422,
+      "step": 40240
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.647320931268796e-05,
+      "loss": 0.6389,
+      "step": 40250
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.64175114180684e-05,
+      "loss": 0.5084,
+      "step": 40260
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.636181352344881e-05,
+      "loss": 0.5967,
+      "step": 40270
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.630611562882922e-05,
+      "loss": 0.6254,
+      "step": 40280
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.625041773420965e-05,
+      "loss": 0.6903,
+      "step": 40290
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.619471983959006e-05,
+      "loss": 0.5762,
+      "step": 40300
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.613902194497047e-05,
+      "loss": 0.4594,
+      "step": 40310
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.60833240503509e-05,
+      "loss": 0.5053,
+      "step": 40320
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.60276261557313e-05,
+      "loss": 0.5539,
+      "step": 40330
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.597192826111172e-05,
+      "loss": 0.6412,
+      "step": 40340
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.591623036649215e-05,
+      "loss": 0.7587,
+      "step": 40350
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.586053247187257e-05,
+      "loss": 0.6224,
+      "step": 40360
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.580483457725298e-05,
+      "loss": 0.5366,
+      "step": 40370
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.574913668263339e-05,
+      "loss": 0.4758,
+      "step": 40380
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.569343878801381e-05,
+      "loss": 0.5738,
+      "step": 40390
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.563774089339422e-05,
+      "loss": 0.6283,
+      "step": 40400
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.558204299877463e-05,
+      "loss": 0.7389,
+      "step": 40410
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.552634510415506e-05,
+      "loss": 0.5932,
+      "step": 40420
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.547064720953547e-05,
+      "loss": 0.5267,
+      "step": 40430
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.541494931491588e-05,
+      "loss": 0.6217,
+      "step": 40440
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.53592514202963e-05,
+      "loss": 0.8216,
+      "step": 40450
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.530355352567672e-05,
+      "loss": 0.5606,
+      "step": 40460
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.524785563105714e-05,
+      "loss": 0.4766,
+      "step": 40470
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.519215773643757e-05,
+      "loss": 0.5511,
+      "step": 40480
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.513645984181798e-05,
+      "loss": 0.6062,
+      "step": 40490
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.508076194719839e-05,
+      "loss": 0.621,
+      "step": 40500
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.50250640525788e-05,
+      "loss": 0.5645,
+      "step": 40510
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.496936615795922e-05,
+      "loss": 0.6488,
+      "step": 40520
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.491366826333964e-05,
+      "loss": 0.5186,
+      "step": 40530
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.485797036872006e-05,
+      "loss": 0.5323,
+      "step": 40540
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.480227247410047e-05,
+      "loss": 0.6308,
+      "step": 40550
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.474657457948088e-05,
+      "loss": 0.5674,
+      "step": 40560
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.469087668486131e-05,
+      "loss": 0.6205,
+      "step": 40570
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.463517879024172e-05,
+      "loss": 0.5036,
+      "step": 40580
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.457948089562214e-05,
+      "loss": 0.5233,
+      "step": 40590
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.452378300100255e-05,
+      "loss": 0.6184,
+      "step": 40600
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.446808510638297e-05,
+      "loss": 0.5887,
+      "step": 40610
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.441238721176339e-05,
+      "loss": 0.4676,
+      "step": 40620
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.43566893171438e-05,
+      "loss": 0.4686,
+      "step": 40630
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.430099142252423e-05,
+      "loss": 0.5638,
+      "step": 40640
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.424529352790464e-05,
+      "loss": 0.5497,
+      "step": 40650
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.418959563328505e-05,
+      "loss": 0.6577,
+      "step": 40660
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.413389773866547e-05,
+      "loss": 0.5928,
+      "step": 40670
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.407819984404588e-05,
+      "loss": 0.6367,
+      "step": 40680
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.402250194942631e-05,
+      "loss": 0.5934,
+      "step": 40690
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.396680405480672e-05,
+      "loss": 0.5527,
+      "step": 40700
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.391110616018713e-05,
+      "loss": 0.6522,
+      "step": 40710
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.385540826556756e-05,
+      "loss": 0.5563,
+      "step": 40720
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.379971037094797e-05,
+      "loss": 0.6284,
+      "step": 40730
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.374401247632839e-05,
+      "loss": 0.4895,
+      "step": 40740
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.36883145817088e-05,
+      "loss": 0.5064,
+      "step": 40750
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.363261668708921e-05,
+      "loss": 0.5851,
+      "step": 40760
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.357691879246964e-05,
+      "loss": 0.6601,
+      "step": 40770
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.352122089785006e-05,
+      "loss": 0.6701,
+      "step": 40780
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.346552300323047e-05,
+      "loss": 0.464,
+      "step": 40790
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.340982510861089e-05,
+      "loss": 0.6026,
+      "step": 40800
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.33541272139913e-05,
+      "loss": 0.6416,
+      "step": 40810
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.329842931937172e-05,
+      "loss": 0.5212,
+      "step": 40820
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.324273142475215e-05,
+      "loss": 0.6097,
+      "step": 40830
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.318703353013256e-05,
+      "loss": 0.5378,
+      "step": 40840
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.313133563551297e-05,
+      "loss": 0.5055,
+      "step": 40850
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.30756377408934e-05,
+      "loss": 0.5933,
+      "step": 40860
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.30199398462738e-05,
+      "loss": 0.6016,
+      "step": 40870
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.296424195165423e-05,
+      "loss": 0.6743,
+      "step": 40880
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.290854405703464e-05,
+      "loss": 0.6074,
+      "step": 40890
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.285284616241505e-05,
+      "loss": 0.6888,
+      "step": 40900
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.279714826779548e-05,
+      "loss": 0.6754,
+      "step": 40910
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.274145037317589e-05,
+      "loss": 0.5157,
+      "step": 40920
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.268575247855631e-05,
+      "loss": 0.5088,
+      "step": 40930
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.263005458393672e-05,
+      "loss": 0.5971,
+      "step": 40940
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.257435668931713e-05,
+      "loss": 0.573,
+      "step": 40950
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.251865879469756e-05,
+      "loss": 0.6144,
+      "step": 40960
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.246296090007797e-05,
+      "loss": 0.6003,
+      "step": 40970
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.24072630054584e-05,
+      "loss": 0.5963,
+      "step": 40980
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.23515651108388e-05,
+      "loss": 0.6157,
+      "step": 40990
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.229586721621922e-05,
+      "loss": 0.6737,
+      "step": 41000
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.224016932159964e-05,
+      "loss": 0.5582,
+      "step": 41010
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.218447142698005e-05,
+      "loss": 0.5435,
+      "step": 41020
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.212877353236046e-05,
+      "loss": 0.5482,
+      "step": 41030
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.207307563774089e-05,
+      "loss": 0.5183,
+      "step": 41040
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.20173777431213e-05,
+      "loss": 0.6322,
+      "step": 41050
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.196167984850172e-05,
+      "loss": 0.5983,
+      "step": 41060
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.190598195388214e-05,
+      "loss": 0.4881,
+      "step": 41070
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.185028405926255e-05,
+      "loss": 0.6044,
+      "step": 41080
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.179458616464297e-05,
+      "loss": 0.4251,
+      "step": 41090
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.173888827002338e-05,
+      "loss": 0.6917,
+      "step": 41100
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.168319037540381e-05,
+      "loss": 0.6223,
+      "step": 41110
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.162749248078422e-05,
+      "loss": 0.536,
+      "step": 41120
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.157179458616463e-05,
+      "loss": 0.6086,
+      "step": 41130
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.151609669154505e-05,
+      "loss": 0.5963,
+      "step": 41140
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.146039879692547e-05,
+      "loss": 0.5298,
+      "step": 41150
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.140470090230589e-05,
+      "loss": 0.6451,
+      "step": 41160
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.13490030076863e-05,
+      "loss": 0.7617,
+      "step": 41170
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.129330511306671e-05,
+      "loss": 0.6068,
+      "step": 41180
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.123760721844714e-05,
+      "loss": 0.5597,
+      "step": 41190
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.118190932382755e-05,
+      "loss": 0.6196,
+      "step": 41200
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.112621142920797e-05,
+      "loss": 0.5273,
+      "step": 41210
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.107051353458838e-05,
+      "loss": 0.431,
+      "step": 41220
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.10148156399688e-05,
+      "loss": 0.587,
+      "step": 41230
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.095911774534922e-05,
+      "loss": 0.5756,
+      "step": 41240
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.090341985072964e-05,
+      "loss": 0.4591,
+      "step": 41250
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.084772195611006e-05,
+      "loss": 0.447,
+      "step": 41260
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.079202406149047e-05,
+      "loss": 0.5481,
+      "step": 41270
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.073632616687088e-05,
+      "loss": 0.5159,
+      "step": 41280
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.06806282722513e-05,
+      "loss": 0.5976,
+      "step": 41290
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.062493037763173e-05,
+      "loss": 0.4891,
+      "step": 41300
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.056923248301214e-05,
+      "loss": 0.694,
+      "step": 41310
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.051353458839255e-05,
+      "loss": 0.5772,
+      "step": 41320
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.045783669377297e-05,
+      "loss": 0.6043,
+      "step": 41330
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.040213879915339e-05,
+      "loss": 0.6163,
+      "step": 41340
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.034644090453381e-05,
+      "loss": 0.567,
+      "step": 41350
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.029074300991422e-05,
+      "loss": 0.5425,
+      "step": 41360
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.023504511529463e-05,
+      "loss": 0.6521,
+      "step": 41370
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.017934722067506e-05,
+      "loss": 0.7866,
+      "step": 41380
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.012364932605547e-05,
+      "loss": 0.5658,
+      "step": 41390
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.006795143143589e-05,
+      "loss": 0.5146,
+      "step": 41400
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.00122535368163e-05,
+      "loss": 0.5498,
+      "step": 41410
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.995655564219672e-05,
+      "loss": 0.6435,
+      "step": 41420
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.990085774757714e-05,
+      "loss": 0.4407,
+      "step": 41430
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.984515985295755e-05,
+      "loss": 0.5896,
+      "step": 41440
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.978946195833798e-05,
+      "loss": 0.6194,
+      "step": 41450
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.973376406371839e-05,
+      "loss": 0.5176,
+      "step": 41460
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.96780661690988e-05,
+      "loss": 0.6486,
+      "step": 41470
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.962236827447922e-05,
+      "loss": 0.7138,
+      "step": 41480
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.956667037985963e-05,
+      "loss": 0.6464,
+      "step": 41490
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.951097248524006e-05,
+      "loss": 0.5833,
+      "step": 41500
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.945527459062047e-05,
+      "loss": 0.56,
+      "step": 41510
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.939957669600088e-05,
+      "loss": 0.4471,
+      "step": 41520
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.93438788013813e-05,
+      "loss": 0.5446,
+      "step": 41530
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.928818090676172e-05,
+      "loss": 0.4813,
+      "step": 41540
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.923248301214213e-05,
+      "loss": 0.6535,
+      "step": 41550
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.917678511752255e-05,
+      "loss": 0.525,
+      "step": 41560
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.912108722290296e-05,
+      "loss": 0.6113,
+      "step": 41570
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.906538932828339e-05,
+      "loss": 0.5219,
+      "step": 41580
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.90096914336638e-05,
+      "loss": 0.6028,
+      "step": 41590
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.895399353904421e-05,
+      "loss": 0.4393,
+      "step": 41600
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.889829564442464e-05,
+      "loss": 0.575,
+      "step": 41610
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.884259774980505e-05,
+      "loss": 0.6329,
+      "step": 41620
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.878689985518547e-05,
+      "loss": 0.6352,
+      "step": 41630
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.873120196056588e-05,
+      "loss": 0.5936,
+      "step": 41640
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.867550406594629e-05,
+      "loss": 0.471,
+      "step": 41650
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.861980617132672e-05,
+      "loss": 0.6303,
+      "step": 41660
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.856410827670713e-05,
+      "loss": 0.5688,
+      "step": 41670
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 6.850841038208755e-05,
+      "loss": 0.4856,
+      "step": 41680
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.845271248746796e-05,
+      "loss": 0.5973,
+      "step": 41690
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.839701459284838e-05,
+      "loss": 0.5696,
+      "step": 41700
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.83413166982288e-05,
+      "loss": 0.5856,
+      "step": 41710
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.828561880360923e-05,
+      "loss": 0.7105,
+      "step": 41720
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.822992090898964e-05,
+      "loss": 0.6047,
+      "step": 41730
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.817422301437005e-05,
+      "loss": 0.5922,
+      "step": 41740
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.811852511975046e-05,
+      "loss": 0.5163,
+      "step": 41750
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.806282722513088e-05,
+      "loss": 0.5799,
+      "step": 41760
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.800712933051131e-05,
+      "loss": 0.6015,
+      "step": 41770
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.795143143589172e-05,
+      "loss": 0.6168,
+      "step": 41780
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.789573354127213e-05,
+      "loss": 0.546,
+      "step": 41790
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.784003564665256e-05,
+      "loss": 0.571,
+      "step": 41800
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.778433775203297e-05,
+      "loss": 0.4898,
+      "step": 41810
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.772863985741339e-05,
+      "loss": 0.5061,
+      "step": 41820
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.76729419627938e-05,
+      "loss": 0.5205,
+      "step": 41830
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.761724406817421e-05,
+      "loss": 0.5441,
+      "step": 41840
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.756154617355464e-05,
+      "loss": 0.5502,
+      "step": 41850
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.750584827893505e-05,
+      "loss": 0.5188,
+      "step": 41860
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.745015038431547e-05,
+      "loss": 0.5865,
+      "step": 41870
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.739445248969588e-05,
+      "loss": 0.48,
+      "step": 41880
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.73387545950763e-05,
+      "loss": 0.5491,
+      "step": 41890
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.728305670045672e-05,
+      "loss": 0.5788,
+      "step": 41900
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.722735880583713e-05,
+      "loss": 0.5824,
+      "step": 41910
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.717166091121756e-05,
+      "loss": 0.5147,
+      "step": 41920
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.711596301659797e-05,
+      "loss": 0.6484,
+      "step": 41930
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.706026512197838e-05,
+      "loss": 0.429,
+      "step": 41940
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 6.70045672273588e-05,
+      "loss": 0.6061,
+      "step": 41950
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.694886933273921e-05,
+      "loss": 0.6317,
+      "step": 41960
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.689317143811964e-05,
+      "loss": 0.5021,
+      "step": 41970
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.683747354350005e-05,
+      "loss": 0.6281,
+      "step": 41980
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.678177564888046e-05,
+      "loss": 0.5148,
+      "step": 41990
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.672607775426089e-05,
+      "loss": 0.584,
+      "step": 42000
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.66703798596413e-05,
+      "loss": 0.6127,
+      "step": 42010
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.661468196502172e-05,
+      "loss": 0.6296,
+      "step": 42020
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.655898407040213e-05,
+      "loss": 0.5686,
+      "step": 42030
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.650328617578254e-05,
+      "loss": 0.5753,
+      "step": 42040
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.644758828116297e-05,
+      "loss": 0.4686,
+      "step": 42050
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.639189038654338e-05,
+      "loss": 0.7022,
+      "step": 42060
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.633619249192379e-05,
+      "loss": 0.6933,
+      "step": 42070
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.628049459730422e-05,
+      "loss": 0.5129,
+      "step": 42080
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.622479670268463e-05,
+      "loss": 0.6067,
+      "step": 42090
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.616909880806505e-05,
+      "loss": 0.6279,
+      "step": 42100
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.611340091344546e-05,
+      "loss": 0.4647,
+      "step": 42110
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.605770301882587e-05,
+      "loss": 0.4879,
+      "step": 42120
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.60020051242063e-05,
+      "loss": 0.4727,
+      "step": 42130
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.594630722958671e-05,
+      "loss": 0.5251,
+      "step": 42140
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.589060933496713e-05,
+      "loss": 0.5472,
+      "step": 42150
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.583491144034755e-05,
+      "loss": 0.6579,
+      "step": 42160
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.577921354572796e-05,
+      "loss": 0.6365,
+      "step": 42170
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.572351565110838e-05,
+      "loss": 0.5025,
+      "step": 42180
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.56678177564888e-05,
+      "loss": 0.6741,
+      "step": 42190
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.561211986186922e-05,
+      "loss": 0.5789,
+      "step": 42200
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.555642196724963e-05,
+      "loss": 0.563,
+      "step": 42210
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 6.550072407263004e-05,
+      "loss": 0.4188,
+      "step": 42220
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.544502617801046e-05,
+      "loss": 0.4501,
+      "step": 42230
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.538932828339089e-05,
+      "loss": 0.6077,
+      "step": 42240
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.53336303887713e-05,
+      "loss": 0.448,
+      "step": 42250
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.527793249415171e-05,
+      "loss": 0.5281,
+      "step": 42260
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.522223459953214e-05,
+      "loss": 0.5279,
+      "step": 42270
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.516653670491255e-05,
+      "loss": 0.587,
+      "step": 42280
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.511083881029297e-05,
+      "loss": 0.6235,
+      "step": 42290
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.505514091567338e-05,
+      "loss": 0.5993,
+      "step": 42300
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.49994430210538e-05,
+      "loss": 0.491,
+      "step": 42310
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.494374512643422e-05,
+      "loss": 0.627,
+      "step": 42320
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.488804723181463e-05,
+      "loss": 0.545,
+      "step": 42330
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.483234933719505e-05,
+      "loss": 0.71,
+      "step": 42340
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.477665144257547e-05,
+      "loss": 0.5984,
+      "step": 42350
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.472095354795588e-05,
+      "loss": 0.7135,
+      "step": 42360
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.46652556533363e-05,
+      "loss": 0.4812,
+      "step": 42370
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.460955775871671e-05,
+      "loss": 0.6753,
+      "step": 42380
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.455385986409714e-05,
+      "loss": 0.5908,
+      "step": 42390
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.449816196947755e-05,
+      "loss": 0.5926,
+      "step": 42400
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.444246407485796e-05,
+      "loss": 0.4748,
+      "step": 42410
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.438676618023838e-05,
+      "loss": 0.6817,
+      "step": 42420
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.43310682856188e-05,
+      "loss": 0.4508,
+      "step": 42430
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.427537039099922e-05,
+      "loss": 0.5497,
+      "step": 42440
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.421967249637963e-05,
+      "loss": 0.5681,
+      "step": 42450
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.416397460176004e-05,
+      "loss": 0.6295,
+      "step": 42460
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.410827670714047e-05,
+      "loss": 0.4416,
+      "step": 42470
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.405257881252088e-05,
+      "loss": 0.5661,
+      "step": 42480
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 6.39968809179013e-05,
+      "loss": 0.5799,
+      "step": 42490
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.394118302328171e-05,
+      "loss": 0.4385,
+      "step": 42500
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.388548512866213e-05,
+      "loss": 0.6964,
+      "step": 42510
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.382978723404255e-05,
+      "loss": 0.5238,
+      "step": 42520
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.377408933942297e-05,
+      "loss": 0.5934,
+      "step": 42530
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.371839144480339e-05,
+      "loss": 0.5819,
+      "step": 42540
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.36626935501838e-05,
+      "loss": 0.5154,
+      "step": 42550
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.360699565556421e-05,
+      "loss": 0.513,
+      "step": 42560
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.355129776094463e-05,
+      "loss": 0.5352,
+      "step": 42570
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.349559986632504e-05,
+      "loss": 0.6589,
+      "step": 42580
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.343990197170547e-05,
+      "loss": 0.5297,
+      "step": 42590
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.338420407708588e-05,
+      "loss": 0.6475,
+      "step": 42600
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.332850618246629e-05,
+      "loss": 0.624,
+      "step": 42610
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.327280828784672e-05,
+      "loss": 0.55,
+      "step": 42620
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.321711039322713e-05,
+      "loss": 0.5857,
+      "step": 42630
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.316141249860754e-05,
+      "loss": 0.6391,
+      "step": 42640
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.310571460398796e-05,
+      "loss": 0.7421,
+      "step": 42650
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.305001670936839e-05,
+      "loss": 0.7518,
+      "step": 42660
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.29943188147488e-05,
+      "loss": 0.5515,
+      "step": 42670
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.293862092012921e-05,
+      "loss": 0.6017,
+      "step": 42680
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.288292302550962e-05,
+      "loss": 0.4212,
+      "step": 42690
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.282722513089005e-05,
+      "loss": 0.449,
+      "step": 42700
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.277152723627047e-05,
+      "loss": 0.5952,
+      "step": 42710
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.271582934165088e-05,
+      "loss": 0.6312,
+      "step": 42720
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.266013144703129e-05,
+      "loss": 0.5841,
+      "step": 42730
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.260443355241172e-05,
+      "loss": 0.6086,
+      "step": 42740
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.254873565779213e-05,
+      "loss": 0.6145,
+      "step": 42750
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 6.249303776317255e-05,
+      "loss": 0.5464,
+      "step": 42760
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.243733986855296e-05,
+      "loss": 0.586,
+      "step": 42770
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.238164197393337e-05,
+      "loss": 0.4917,
+      "step": 42780
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.23259440793138e-05,
+      "loss": 0.614,
+      "step": 42790
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.227024618469421e-05,
+      "loss": 0.702,
+      "step": 42800
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.221454829007464e-05,
+      "loss": 0.6493,
+      "step": 42810
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.215885039545505e-05,
+      "loss": 0.6244,
+      "step": 42820
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.210315250083546e-05,
+      "loss": 0.5299,
+      "step": 42830
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.204745460621588e-05,
+      "loss": 0.5824,
+      "step": 42840
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.19917567115963e-05,
+      "loss": 0.5691,
+      "step": 42850
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.193605881697672e-05,
+      "loss": 0.5034,
+      "step": 42860
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.188036092235713e-05,
+      "loss": 0.6444,
+      "step": 42870
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.182466302773754e-05,
+      "loss": 0.6555,
+      "step": 42880
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.176896513311797e-05,
+      "loss": 0.5187,
+      "step": 42890
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.171326723849838e-05,
+      "loss": 0.5819,
+      "step": 42900
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.16575693438788e-05,
+      "loss": 0.5772,
+      "step": 42910
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.160187144925921e-05,
+      "loss": 0.5957,
+      "step": 42920
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.154617355463962e-05,
+      "loss": 0.5934,
+      "step": 42930
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.149047566002005e-05,
+      "loss": 0.5445,
+      "step": 42940
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.143477776540046e-05,
+      "loss": 0.6011,
+      "step": 42950
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.137907987078088e-05,
+      "loss": 0.5211,
+      "step": 42960
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.13233819761613e-05,
+      "loss": 0.5328,
+      "step": 42970
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.12676840815417e-05,
+      "loss": 0.5839,
+      "step": 42980
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.121198618692213e-05,
+      "loss": 0.5628,
+      "step": 42990
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.115628829230256e-05,
+      "loss": 0.5969,
+      "step": 43000
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.110059039768297e-05,
+      "loss": 0.6406,
+      "step": 43010
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.104489250306338e-05,
+      "loss": 0.4462,
+      "step": 43020
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 6.098919460844379e-05,
+      "loss": 0.5414,
+      "step": 43030
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.0933496713824214e-05,
+      "loss": 0.742,
+      "step": 43040
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.087779881920463e-05,
+      "loss": 0.5354,
+      "step": 43050
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.082210092458504e-05,
+      "loss": 0.5085,
+      "step": 43060
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.076640302996546e-05,
+      "loss": 0.4648,
+      "step": 43070
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.071070513534587e-05,
+      "loss": 0.5113,
+      "step": 43080
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.0655007240726296e-05,
+      "loss": 0.723,
+      "step": 43090
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.0599309346106714e-05,
+      "loss": 0.6523,
+      "step": 43100
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.0543611451487125e-05,
+      "loss": 0.565,
+      "step": 43110
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.048791355686754e-05,
+      "loss": 0.5951,
+      "step": 43120
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.043221566224797e-05,
+      "loss": 0.5108,
+      "step": 43130
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.037651776762838e-05,
+      "loss": 0.5616,
+      "step": 43140
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.03208198730088e-05,
+      "loss": 0.4755,
+      "step": 43150
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.026512197838921e-05,
+      "loss": 0.5337,
+      "step": 43160
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.0209424083769626e-05,
+      "loss": 0.5642,
+      "step": 43170
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.015372618915005e-05,
+      "loss": 0.5967,
+      "step": 43180
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.009802829453046e-05,
+      "loss": 0.5457,
+      "step": 43190
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 6.004233039991088e-05,
+      "loss": 0.614,
+      "step": 43200
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 5.99866325052913e-05,
+      "loss": 0.6379,
+      "step": 43210
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 5.993093461067171e-05,
+      "loss": 0.4907,
+      "step": 43220
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 5.987523671605213e-05,
+      "loss": 0.586,
+      "step": 43230
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 5.9819538821432545e-05,
+      "loss": 0.7276,
+      "step": 43240
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 5.976384092681296e-05,
+      "loss": 0.6253,
+      "step": 43250
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 5.970814303219338e-05,
+      "loss": 0.5205,
+      "step": 43260
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 5.965244513757379e-05,
+      "loss": 0.6194,
+      "step": 43270
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 5.959674724295421e-05,
+      "loss": 0.537,
+      "step": 43280
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 5.954104934833462e-05,
+      "loss": 0.4461,
+      "step": 43290
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 5.9485351453715045e-05,
+      "loss": 0.6683,
+      "step": 43300
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.942965355909546e-05,
+      "loss": 0.5558,
+      "step": 43310
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.9373955664475874e-05,
+      "loss": 0.7098,
+      "step": 43320
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.931825776985629e-05,
+      "loss": 0.5677,
+      "step": 43330
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.926255987523672e-05,
+      "loss": 0.5365,
+      "step": 43340
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.920686198061713e-05,
+      "loss": 0.4692,
+      "step": 43350
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.9151164085997546e-05,
+      "loss": 0.6152,
+      "step": 43360
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.909546619137796e-05,
+      "loss": 0.5987,
+      "step": 43370
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.9039768296758375e-05,
+      "loss": 0.4675,
+      "step": 43380
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.89840704021388e-05,
+      "loss": 0.5813,
+      "step": 43390
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.892837250751921e-05,
+      "loss": 0.5771,
+      "step": 43400
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.887267461289963e-05,
+      "loss": 0.7069,
+      "step": 43410
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.881697671828004e-05,
+      "loss": 0.6672,
+      "step": 43420
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.876127882366046e-05,
+      "loss": 0.5319,
+      "step": 43430
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.870558092904088e-05,
+      "loss": 0.622,
+      "step": 43440
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.8649883034421294e-05,
+      "loss": 0.6422,
+      "step": 43450
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.859418513980171e-05,
+      "loss": 0.6164,
+      "step": 43460
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.853848724518213e-05,
+      "loss": 0.5758,
+      "step": 43470
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.848278935056254e-05,
+      "loss": 0.5833,
+      "step": 43480
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.8427091455942965e-05,
+      "loss": 0.5216,
+      "step": 43490
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.8371393561323377e-05,
+      "loss": 0.7481,
+      "step": 43500
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.8315695666703794e-05,
+      "loss": 0.6219,
+      "step": 43510
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.825999777208421e-05,
+      "loss": 0.6885,
+      "step": 43520
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.8204299877464624e-05,
+      "loss": 0.6049,
+      "step": 43530
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.814860198284504e-05,
+      "loss": 0.5224,
+      "step": 43540
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.809290408822546e-05,
+      "loss": 0.5185,
+      "step": 43550
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.803720619360588e-05,
+      "loss": 0.685,
+      "step": 43560
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.7981508298986295e-05,
+      "loss": 0.6797,
+      "step": 43570
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.7925810404366706e-05,
+      "loss": 0.5858,
+      "step": 43580
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.7870112509747124e-05,
+      "loss": 0.5643,
+      "step": 43590
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.781441461512755e-05,
+      "loss": 0.5286,
+      "step": 43600
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.775871672050796e-05,
+      "loss": 0.478,
+      "step": 43610
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.770301882588838e-05,
+      "loss": 0.5956,
+      "step": 43620
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.764732093126879e-05,
+      "loss": 0.5075,
+      "step": 43630
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.759162303664921e-05,
+      "loss": 0.5067,
+      "step": 43640
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.753592514202963e-05,
+      "loss": 0.6501,
+      "step": 43650
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.748022724741004e-05,
+      "loss": 0.6296,
+      "step": 43660
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.742452935279046e-05,
+      "loss": 0.5735,
+      "step": 43670
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.736883145817088e-05,
+      "loss": 0.7211,
+      "step": 43680
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.731313356355129e-05,
+      "loss": 0.6313,
+      "step": 43690
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.7257435668931714e-05,
+      "loss": 0.5505,
+      "step": 43700
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.7201737774312126e-05,
+      "loss": 0.5815,
+      "step": 43710
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.7146039879692544e-05,
+      "loss": 0.5547,
+      "step": 43720
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.709034198507296e-05,
+      "loss": 0.5678,
+      "step": 43730
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.703464409045337e-05,
+      "loss": 0.5824,
+      "step": 43740
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.69789461958338e-05,
+      "loss": 0.6383,
+      "step": 43750
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.692324830121421e-05,
+      "loss": 0.5237,
+      "step": 43760
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.6867550406594626e-05,
+      "loss": 0.491,
+      "step": 43770
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.6811852511975044e-05,
+      "loss": 0.5946,
+      "step": 43780
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.6756154617355455e-05,
+      "loss": 0.5582,
+      "step": 43790
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.670045672273587e-05,
+      "loss": 0.6475,
+      "step": 43800
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.66447588281163e-05,
+      "loss": 0.5595,
+      "step": 43810
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.658906093349671e-05,
+      "loss": 0.5113,
+      "step": 43820
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.653336303887713e-05,
+      "loss": 0.545,
+      "step": 43830
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.647766514425754e-05,
+      "loss": 0.653,
+      "step": 43840
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.6421967249637956e-05,
+      "loss": 0.6246,
+      "step": 43850
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.636626935501838e-05,
+      "loss": 0.4969,
+      "step": 43860
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.631057146039879e-05,
+      "loss": 0.6338,
+      "step": 43870
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.625487356577921e-05,
+      "loss": 0.5956,
+      "step": 43880
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.619917567115962e-05,
+      "loss": 0.522,
+      "step": 43890
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.614347777654004e-05,
+      "loss": 0.7748,
+      "step": 43900
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.6087779881920464e-05,
+      "loss": 0.4356,
+      "step": 43910
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.6032081987300875e-05,
+      "loss": 0.3851,
+      "step": 43920
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.597638409268129e-05,
+      "loss": 0.5757,
+      "step": 43930
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.592068619806171e-05,
+      "loss": 0.441,
+      "step": 43940
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.586498830344212e-05,
+      "loss": 0.548,
+      "step": 43950
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.5809290408822546e-05,
+      "loss": 0.4554,
+      "step": 43960
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.575359251420296e-05,
+      "loss": 0.4906,
+      "step": 43970
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.5697894619583375e-05,
+      "loss": 0.5354,
+      "step": 43980
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.564219672496379e-05,
+      "loss": 0.5002,
+      "step": 43990
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.5586498830344204e-05,
+      "loss": 0.5484,
+      "step": 44000
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.553080093572463e-05,
+      "loss": 0.5703,
+      "step": 44010
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.547510304110504e-05,
+      "loss": 0.6664,
+      "step": 44020
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.541940514648546e-05,
+      "loss": 0.4722,
+      "step": 44030
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.5363707251865876e-05,
+      "loss": 0.4626,
+      "step": 44040
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.530800935724629e-05,
+      "loss": 0.5479,
+      "step": 44050
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.5252311462626705e-05,
+      "loss": 0.525,
+      "step": 44060
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.519661356800713e-05,
+      "loss": 0.6428,
+      "step": 44070
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.514091567338754e-05,
+      "loss": 0.5166,
+      "step": 44080
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.508521777876796e-05,
+      "loss": 0.6326,
+      "step": 44090
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.502951988414837e-05,
+      "loss": 0.5729,
+      "step": 44100
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.497382198952879e-05,
+      "loss": 0.6143,
+      "step": 44110
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.491812409490921e-05,
+      "loss": 0.6401,
+      "step": 44120
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.4862426200289624e-05,
+      "loss": 0.4702,
+      "step": 44130
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.480672830567004e-05,
+      "loss": 0.6347,
+      "step": 44140
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.475103041105046e-05,
+      "loss": 0.5589,
+      "step": 44150
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.469533251643087e-05,
+      "loss": 0.561,
+      "step": 44160
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.4639634621811295e-05,
+      "loss": 0.5932,
+      "step": 44170
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.4583936727191706e-05,
+      "loss": 0.5326,
+      "step": 44180
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.4528238832572124e-05,
+      "loss": 0.6079,
+      "step": 44190
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.447254093795254e-05,
+      "loss": 0.6451,
+      "step": 44200
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.4416843043332953e-05,
+      "loss": 0.7125,
+      "step": 44210
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.436114514871338e-05,
+      "loss": 0.5346,
+      "step": 44220
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.430544725409379e-05,
+      "loss": 0.5436,
+      "step": 44230
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.424974935947421e-05,
+      "loss": 0.7746,
+      "step": 44240
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.4194051464854625e-05,
+      "loss": 0.6162,
+      "step": 44250
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.4138353570235036e-05,
+      "loss": 0.5579,
+      "step": 44260
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.408265567561546e-05,
+      "loss": 0.5597,
+      "step": 44270
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.402695778099588e-05,
+      "loss": 0.594,
+      "step": 44280
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.397125988637629e-05,
+      "loss": 0.5743,
+      "step": 44290
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.391556199175671e-05,
+      "loss": 0.6061,
+      "step": 44300
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.385986409713712e-05,
+      "loss": 0.5044,
+      "step": 44310
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.380416620251754e-05,
+      "loss": 0.4822,
+      "step": 44320
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.374846830789796e-05,
+      "loss": 0.5765,
+      "step": 44330
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.369277041327837e-05,
+      "loss": 0.505,
+      "step": 44340
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.363707251865879e-05,
+      "loss": 0.6134,
+      "step": 44350
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.35813746240392e-05,
+      "loss": 0.5075,
+      "step": 44360
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.352567672941962e-05,
+      "loss": 0.5673,
+      "step": 44370
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.3469978834800044e-05,
+      "loss": 0.6172,
+      "step": 44380
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.3414280940180456e-05,
+      "loss": 0.5002,
+      "step": 44390
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.3358583045560873e-05,
+      "loss": 0.6122,
+      "step": 44400
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.330288515094129e-05,
+      "loss": 0.5741,
+      "step": 44410
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.32471872563217e-05,
+      "loss": 0.5744,
+      "step": 44420
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.319148936170213e-05,
+      "loss": 0.5806,
+      "step": 44430
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.313579146708254e-05,
+      "loss": 0.464,
+      "step": 44440
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.3080093572462956e-05,
+      "loss": 0.5507,
+      "step": 44450
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.302996546730533e-05,
+      "loss": 0.5791,
+      "step": 44460
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.297426757268575e-05,
+      "loss": 0.592,
+      "step": 44470
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.291856967806616e-05,
+      "loss": 0.636,
+      "step": 44480
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.286287178344658e-05,
+      "loss": 0.6583,
+      "step": 44490
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.2807173888827e-05,
+      "loss": 0.5768,
+      "step": 44500
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.275147599420741e-05,
+      "loss": 0.5204,
+      "step": 44510
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.269577809958783e-05,
+      "loss": 0.5837,
+      "step": 44520
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.264008020496824e-05,
+      "loss": 0.5983,
+      "step": 44530
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.2584382310348665e-05,
+      "loss": 0.5578,
+      "step": 44540
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.2528684415729083e-05,
+      "loss": 0.5907,
+      "step": 44550
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.2472986521109495e-05,
+      "loss": 0.7457,
+      "step": 44560
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.241728862648991e-05,
+      "loss": 0.5937,
+      "step": 44570
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.236159073187034e-05,
+      "loss": 0.6664,
+      "step": 44580
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.230589283725075e-05,
+      "loss": 0.5235,
+      "step": 44590
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.2250194942631166e-05,
+      "loss": 0.7534,
+      "step": 44600
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.219449704801158e-05,
+      "loss": 0.6803,
+      "step": 44610
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.2138799153391995e-05,
+      "loss": 0.5623,
+      "step": 44620
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.208310125877242e-05,
+      "loss": 0.5566,
+      "step": 44630
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.202740336415283e-05,
+      "loss": 0.6319,
+      "step": 44640
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.197170546953325e-05,
+      "loss": 0.6485,
+      "step": 44650
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.191600757491367e-05,
+      "loss": 0.5822,
+      "step": 44660
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.186030968029408e-05,
+      "loss": 0.5371,
+      "step": 44670
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.1804611785674496e-05,
+      "loss": 0.4855,
+      "step": 44680
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.1748913891054914e-05,
+      "loss": 0.538,
+      "step": 44690
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.169321599643533e-05,
+      "loss": 0.7173,
+      "step": 44700
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.163751810181575e-05,
+      "loss": 0.6047,
+      "step": 44710
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.158182020719616e-05,
+      "loss": 0.5279,
+      "step": 44720
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.152612231257658e-05,
+      "loss": 0.5356,
+      "step": 44730
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.147042441795699e-05,
+      "loss": 0.7094,
+      "step": 44740
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.1414726523337415e-05,
+      "loss": 0.4974,
+      "step": 44750
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.135902862871783e-05,
+      "loss": 0.5157,
+      "step": 44760
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.1303330734098244e-05,
+      "loss": 0.5486,
+      "step": 44770
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.124763283947866e-05,
+      "loss": 0.6363,
+      "step": 44780
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.1191934944859086e-05,
+      "loss": 0.4916,
+      "step": 44790
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.11362370502395e-05,
+      "loss": 0.5139,
+      "step": 44800
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.1080539155619915e-05,
+      "loss": 0.4681,
+      "step": 44810
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.1024841261000326e-05,
+      "loss": 0.569,
+      "step": 44820
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.0969143366380744e-05,
+      "loss": 0.5176,
+      "step": 44830
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.091344547176117e-05,
+      "loss": 0.6875,
+      "step": 44840
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.085774757714158e-05,
+      "loss": 0.5992,
+      "step": 44850
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.0802049682522e-05,
+      "loss": 0.6228,
+      "step": 44860
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.074635178790241e-05,
+      "loss": 0.5097,
+      "step": 44870
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.069065389328283e-05,
+      "loss": 0.5447,
+      "step": 44880
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.063495599866325e-05,
+      "loss": 0.5607,
+      "step": 44890
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.057925810404366e-05,
+      "loss": 0.5853,
+      "step": 44900
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.052356020942408e-05,
+      "loss": 0.462,
+      "step": 44910
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.04678623148045e-05,
+      "loss": 0.5318,
+      "step": 44920
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 5.041216442018491e-05,
+      "loss": 0.538,
+      "step": 44930
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 5.035646652556533e-05,
+      "loss": 0.53,
+      "step": 44940
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 5.0300768630945746e-05,
+      "loss": 0.5686,
+      "step": 44950
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 5.0245070736326164e-05,
+      "loss": 0.5887,
+      "step": 44960
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 5.018937284170658e-05,
+      "loss": 0.5257,
+      "step": 44970
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 5.013367494708699e-05,
+      "loss": 0.6711,
+      "step": 44980
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 5.007797705246741e-05,
+      "loss": 0.538,
+      "step": 44990
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 5.002227915784782e-05,
+      "loss": 0.5684,
+      "step": 45000
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.9966581263228246e-05,
+      "loss": 0.5431,
+      "step": 45010
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.9910883368608664e-05,
+      "loss": 0.5484,
+      "step": 45020
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.9855185473989075e-05,
+      "loss": 0.4859,
+      "step": 45030
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.979948757936949e-05,
+      "loss": 0.5966,
+      "step": 45040
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.974378968474992e-05,
+      "loss": 0.5927,
+      "step": 45050
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.968809179013033e-05,
+      "loss": 0.4849,
+      "step": 45060
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.963239389551075e-05,
+      "loss": 0.5171,
+      "step": 45070
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.957669600089116e-05,
+      "loss": 0.664,
+      "step": 45080
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.9520998106271576e-05,
+      "loss": 0.6009,
+      "step": 45090
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.9465300211652e-05,
+      "loss": 0.5988,
+      "step": 45100
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.940960231703241e-05,
+      "loss": 0.5559,
+      "step": 45110
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.935390442241283e-05,
+      "loss": 0.508,
+      "step": 45120
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.929820652779325e-05,
+      "loss": 0.6703,
+      "step": 45130
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.924250863317366e-05,
+      "loss": 0.5753,
+      "step": 45140
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.9186810738554084e-05,
+      "loss": 0.5609,
+      "step": 45150
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.9131112843934495e-05,
+      "loss": 0.6179,
+      "step": 45160
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.907541494931491e-05,
+      "loss": 0.4829,
+      "step": 45170
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.901971705469533e-05,
+      "loss": 0.5374,
+      "step": 45180
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.896401916007574e-05,
+      "loss": 0.4785,
+      "step": 45190
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.890832126545616e-05,
+      "loss": 0.4718,
+      "step": 45200
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.885262337083658e-05,
+      "loss": 0.5101,
+      "step": 45210
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.8796925476216995e-05,
+      "loss": 0.6352,
+      "step": 45220
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.874122758159741e-05,
+      "loss": 0.4965,
+      "step": 45230
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.8685529686977824e-05,
+      "loss": 0.5496,
+      "step": 45240
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.862983179235824e-05,
+      "loss": 0.6111,
+      "step": 45250
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.857413389773867e-05,
+      "loss": 0.6453,
+      "step": 45260
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.851843600311908e-05,
+      "loss": 0.5842,
+      "step": 45270
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.8462738108499496e-05,
+      "loss": 0.5155,
+      "step": 45280
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.840704021387991e-05,
+      "loss": 0.6652,
+      "step": 45290
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.8351342319260325e-05,
+      "loss": 0.5439,
+      "step": 45300
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.829564442464075e-05,
+      "loss": 0.5285,
+      "step": 45310
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.823994653002116e-05,
+      "loss": 0.605,
+      "step": 45320
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.818424863540158e-05,
+      "loss": 0.5689,
+      "step": 45330
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.812855074078199e-05,
+      "loss": 0.6165,
+      "step": 45340
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.807285284616241e-05,
+      "loss": 0.5203,
+      "step": 45350
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.801715495154283e-05,
+      "loss": 0.5506,
+      "step": 45360
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.7961457056923244e-05,
+      "loss": 0.4898,
+      "step": 45370
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.790575916230366e-05,
+      "loss": 0.5658,
+      "step": 45380
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.785006126768408e-05,
+      "loss": 0.4978,
+      "step": 45390
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.779436337306449e-05,
+      "loss": 0.5876,
+      "step": 45400
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.7738665478444915e-05,
+      "loss": 0.5455,
+      "step": 45410
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.7682967583825327e-05,
+      "loss": 0.5432,
+      "step": 45420
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.7627269689205744e-05,
+      "loss": 0.5204,
+      "step": 45430
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.757157179458616e-05,
+      "loss": 0.578,
+      "step": 45440
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.7515873899966574e-05,
+      "loss": 0.4366,
+      "step": 45450
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.7460176005347e-05,
+      "loss": 0.5416,
+      "step": 45460
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.740447811072741e-05,
+      "loss": 0.5189,
+      "step": 45470
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.734878021610783e-05,
+      "loss": 0.641,
+      "step": 45480
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.7293082321488245e-05,
+      "loss": 0.5771,
+      "step": 45490
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.7237384426868656e-05,
+      "loss": 0.6699,
+      "step": 45500
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.7181686532249074e-05,
+      "loss": 0.4789,
+      "step": 45510
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.71259886376295e-05,
+      "loss": 0.6007,
+      "step": 45520
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.707029074300991e-05,
+      "loss": 0.608,
+      "step": 45530
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.701459284839033e-05,
+      "loss": 0.4538,
+      "step": 45540
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.695889495377074e-05,
+      "loss": 0.4578,
+      "step": 45550
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.690319705915116e-05,
+      "loss": 0.5789,
+      "step": 45560
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.684749916453158e-05,
+      "loss": 0.5112,
+      "step": 45570
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.679180126991199e-05,
+      "loss": 0.5292,
+      "step": 45580
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.673610337529241e-05,
+      "loss": 0.552,
+      "step": 45590
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.668040548067283e-05,
+      "loss": 0.588,
+      "step": 45600
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.662470758605324e-05,
+      "loss": 0.5784,
+      "step": 45610
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.6569009691433664e-05,
+      "loss": 0.7478,
+      "step": 45620
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.6513311796814076e-05,
+      "loss": 0.5005,
+      "step": 45630
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.6457613902194494e-05,
+      "loss": 0.6307,
+      "step": 45640
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.640191600757491e-05,
+      "loss": 0.486,
+      "step": 45650
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.634621811295532e-05,
+      "loss": 0.5088,
+      "step": 45660
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.629052021833575e-05,
+      "loss": 0.5886,
+      "step": 45670
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.623482232371616e-05,
+      "loss": 0.4676,
+      "step": 45680
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.6179124429096576e-05,
+      "loss": 0.4557,
+      "step": 45690
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.6123426534476994e-05,
+      "loss": 0.5614,
+      "step": 45700
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.6067728639857405e-05,
+      "loss": 0.5603,
+      "step": 45710
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.601203074523783e-05,
+      "loss": 0.6486,
+      "step": 45720
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.595633285061825e-05,
+      "loss": 0.6395,
+      "step": 45730
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.590063495599866e-05,
+      "loss": 0.6031,
+      "step": 45740
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.584493706137908e-05,
+      "loss": 0.6164,
+      "step": 45750
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.578923916675949e-05,
+      "loss": 0.6158,
+      "step": 45760
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.5733541272139906e-05,
+      "loss": 0.4844,
+      "step": 45770
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.567784337752033e-05,
+      "loss": 0.5709,
+      "step": 45780
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.562214548290074e-05,
+      "loss": 0.6556,
+      "step": 45790
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.556644758828116e-05,
+      "loss": 0.495,
+      "step": 45800
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.551074969366157e-05,
+      "loss": 0.5378,
+      "step": 45810
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.545505179904199e-05,
+      "loss": 0.5235,
+      "step": 45820
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.5399353904422414e-05,
+      "loss": 0.61,
+      "step": 45830
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.5343656009802825e-05,
+      "loss": 0.5406,
+      "step": 45840
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.528795811518324e-05,
+      "loss": 0.5562,
+      "step": 45850
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.523226022056366e-05,
+      "loss": 0.4969,
+      "step": 45860
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.517656232594407e-05,
+      "loss": 0.523,
+      "step": 45870
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.5120864431324496e-05,
+      "loss": 0.6725,
+      "step": 45880
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.506516653670491e-05,
+      "loss": 0.5714,
+      "step": 45890
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.5009468642085325e-05,
+      "loss": 0.5511,
+      "step": 45900
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.495377074746574e-05,
+      "loss": 0.5412,
+      "step": 45910
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.4898072852846154e-05,
+      "loss": 0.4855,
+      "step": 45920
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.484237495822658e-05,
+      "loss": 0.5127,
+      "step": 45930
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.478667706360699e-05,
+      "loss": 0.6152,
+      "step": 45940
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.473097916898741e-05,
+      "loss": 0.6062,
+      "step": 45950
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.4675281274367826e-05,
+      "loss": 0.5077,
+      "step": 45960
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.461958337974824e-05,
+      "loss": 0.5717,
+      "step": 45970
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.456388548512866e-05,
+      "loss": 0.5763,
+      "step": 45980
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.450818759050908e-05,
+      "loss": 0.6363,
+      "step": 45990
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.445248969588949e-05,
+      "loss": 0.545,
+      "step": 46000
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.439679180126991e-05,
+      "loss": 0.5906,
+      "step": 46010
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.434109390665032e-05,
+      "loss": 0.6867,
+      "step": 46020
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.428539601203074e-05,
+      "loss": 0.6252,
+      "step": 46030
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.422969811741116e-05,
+      "loss": 0.4837,
+      "step": 46040
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.4174000222791574e-05,
+      "loss": 0.4988,
+      "step": 46050
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.411830232817199e-05,
+      "loss": 0.5887,
+      "step": 46060
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.406260443355241e-05,
+      "loss": 0.5581,
+      "step": 46070
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.400690653893282e-05,
+      "loss": 0.6246,
+      "step": 46080
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.3951208644313245e-05,
+      "loss": 0.5884,
+      "step": 46090
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.3895510749693656e-05,
+      "loss": 0.5568,
+      "step": 46100
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.3839812855074074e-05,
+      "loss": 0.6271,
+      "step": 46110
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.378411496045449e-05,
+      "loss": 0.6573,
+      "step": 46120
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.3728417065834903e-05,
+      "loss": 0.7016,
+      "step": 46130
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.367271917121533e-05,
+      "loss": 0.7095,
+      "step": 46140
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.361702127659574e-05,
+      "loss": 0.4233,
+      "step": 46150
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.356132338197616e-05,
+      "loss": 0.504,
+      "step": 46160
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.3505625487356575e-05,
+      "loss": 0.5638,
+      "step": 46170
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.3449927592736986e-05,
+      "loss": 0.5206,
+      "step": 46180
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.339422969811741e-05,
+      "loss": 0.4868,
+      "step": 46190
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.333853180349783e-05,
+      "loss": 0.4563,
+      "step": 46200
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.328283390887824e-05,
+      "loss": 0.6311,
+      "step": 46210
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.322713601425866e-05,
+      "loss": 0.5634,
+      "step": 46220
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.317143811963907e-05,
+      "loss": 0.6034,
+      "step": 46230
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.3115740225019494e-05,
+      "loss": 0.6102,
+      "step": 46240
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.306004233039991e-05,
+      "loss": 0.6777,
+      "step": 46250
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.300434443578032e-05,
+      "loss": 0.6366,
+      "step": 46260
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.294864654116074e-05,
+      "loss": 0.7463,
+      "step": 46270
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.289294864654115e-05,
+      "loss": 0.6662,
+      "step": 46280
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.283725075192157e-05,
+      "loss": 0.7086,
+      "step": 46290
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.2781552857301994e-05,
+      "loss": 0.5628,
+      "step": 46300
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.2725854962682406e-05,
+      "loss": 0.5446,
+      "step": 46310
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.2670157068062823e-05,
+      "loss": 0.6231,
+      "step": 46320
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.261445917344324e-05,
+      "loss": 0.6283,
+      "step": 46330
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.255876127882365e-05,
+      "loss": 0.4289,
+      "step": 46340
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.250306338420408e-05,
+      "loss": 0.5489,
+      "step": 46350
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.244736548958449e-05,
+      "loss": 0.5115,
+      "step": 46360
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.2391667594964906e-05,
+      "loss": 0.5882,
+      "step": 46370
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.2335969700345324e-05,
+      "loss": 0.6087,
+      "step": 46380
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.2280271805725735e-05,
+      "loss": 0.6942,
+      "step": 46390
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.222457391110616e-05,
+      "loss": 0.6756,
+      "step": 46400
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.216887601648657e-05,
+      "loss": 0.6181,
+      "step": 46410
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.211317812186699e-05,
+      "loss": 0.6081,
+      "step": 46420
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.205748022724741e-05,
+      "loss": 0.5893,
+      "step": 46430
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.200178233262782e-05,
+      "loss": 0.5367,
+      "step": 46440
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.194608443800824e-05,
+      "loss": 0.5065,
+      "step": 46450
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.189038654338866e-05,
+      "loss": 0.5657,
+      "step": 46460
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.183468864876907e-05,
+      "loss": 0.5578,
+      "step": 46470
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.177899075414949e-05,
+      "loss": 0.5015,
+      "step": 46480
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.17232928595299e-05,
+      "loss": 0.6406,
+      "step": 46490
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.1667594964910326e-05,
+      "loss": 0.5087,
+      "step": 46500
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.1611897070290743e-05,
+      "loss": 0.5553,
+      "step": 46510
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.1556199175671155e-05,
+      "loss": 0.669,
+      "step": 46520
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.150050128105157e-05,
+      "loss": 0.4915,
+      "step": 46530
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.144480338643199e-05,
+      "loss": 0.5879,
+      "step": 46540
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.139467528127437e-05,
+      "loss": 0.664,
+      "step": 46550
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.133897738665478e-05,
+      "loss": 0.5826,
+      "step": 46560
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.12832794920352e-05,
+      "loss": 0.6061,
+      "step": 46570
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.122758159741562e-05,
+      "loss": 0.4465,
+      "step": 46580
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.117188370279603e-05,
+      "loss": 0.5513,
+      "step": 46590
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.111618580817645e-05,
+      "loss": 0.4745,
+      "step": 46600
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.1060487913556864e-05,
+      "loss": 0.6092,
+      "step": 46610
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.100479001893728e-05,
+      "loss": 0.7792,
+      "step": 46620
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.09490921243177e-05,
+      "loss": 0.6153,
+      "step": 46630
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.089339422969811e-05,
+      "loss": 0.5581,
+      "step": 46640
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.083769633507853e-05,
+      "loss": 0.526,
+      "step": 46650
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.078199844045895e-05,
+      "loss": 0.5867,
+      "step": 46660
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.0726300545839365e-05,
+      "loss": 0.5199,
+      "step": 46670
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.067060265121978e-05,
+      "loss": 0.496,
+      "step": 46680
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.0614904756600194e-05,
+      "loss": 0.5653,
+      "step": 46690
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.055920686198061e-05,
+      "loss": 0.6168,
+      "step": 46700
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.0503508967361036e-05,
+      "loss": 0.5163,
+      "step": 46710
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.044781107274145e-05,
+      "loss": 0.5686,
+      "step": 46720
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.0392113178121865e-05,
+      "loss": 0.6113,
+      "step": 46730
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.0336415283502276e-05,
+      "loss": 0.5519,
+      "step": 46740
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.0280717388882694e-05,
+      "loss": 0.4987,
+      "step": 46750
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.022501949426312e-05,
+      "loss": 0.5284,
+      "step": 46760
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.016932159964353e-05,
+      "loss": 0.6033,
+      "step": 46770
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.011362370502395e-05,
+      "loss": 0.6149,
+      "step": 46780
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.005792581040436e-05,
+      "loss": 0.5985,
+      "step": 46790
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.000222791578478e-05,
+      "loss": 0.6426,
+      "step": 46800
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 3.99465300211652e-05,
+      "loss": 0.6223,
+      "step": 46810
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.989083212654561e-05,
+      "loss": 0.5255,
+      "step": 46820
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.983513423192603e-05,
+      "loss": 0.564,
+      "step": 46830
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.977943633730645e-05,
+      "loss": 0.5405,
+      "step": 46840
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.972373844268686e-05,
+      "loss": 0.5923,
+      "step": 46850
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.9668040548067285e-05,
+      "loss": 0.5136,
+      "step": 46860
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.9612342653447696e-05,
+      "loss": 0.5915,
+      "step": 46870
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.9556644758828114e-05,
+      "loss": 0.4871,
+      "step": 46880
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.950094686420853e-05,
+      "loss": 0.5138,
+      "step": 46890
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.944524896958894e-05,
+      "loss": 0.4976,
+      "step": 46900
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.938955107496936e-05,
+      "loss": 0.4912,
+      "step": 46910
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.933385318034978e-05,
+      "loss": 0.7713,
+      "step": 46920
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.9278155285730196e-05,
+      "loss": 0.6141,
+      "step": 46930
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.9222457391110614e-05,
+      "loss": 0.5806,
+      "step": 46940
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.9166759496491025e-05,
+      "loss": 0.5268,
+      "step": 46950
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.911106160187144e-05,
+      "loss": 0.6086,
+      "step": 46960
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.905536370725187e-05,
+      "loss": 0.4905,
+      "step": 46970
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.899966581263228e-05,
+      "loss": 0.6449,
+      "step": 46980
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.89439679180127e-05,
+      "loss": 0.5924,
+      "step": 46990
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.888827002339311e-05,
+      "loss": 0.6073,
+      "step": 47000
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.8832572128773526e-05,
+      "loss": 0.5415,
+      "step": 47010
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.877687423415395e-05,
+      "loss": 0.5702,
+      "step": 47020
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.872117633953436e-05,
+      "loss": 0.4638,
+      "step": 47030
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.866547844491478e-05,
+      "loss": 0.6281,
+      "step": 47040
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.86097805502952e-05,
+      "loss": 0.5637,
+      "step": 47050
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.855408265567561e-05,
+      "loss": 0.5895,
+      "step": 47060
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.8498384761056034e-05,
+      "loss": 0.3958,
+      "step": 47070
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.8442686866436445e-05,
+      "loss": 0.4596,
+      "step": 47080
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.838698897181686e-05,
+      "loss": 0.5062,
+      "step": 47090
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.833129107719728e-05,
+      "loss": 0.6428,
+      "step": 47100
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.827559318257769e-05,
+      "loss": 0.5513,
+      "step": 47110
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.8219895287958116e-05,
+      "loss": 0.5119,
+      "step": 47120
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.816419739333853e-05,
+      "loss": 0.5603,
+      "step": 47130
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.8108499498718945e-05,
+      "loss": 0.6028,
+      "step": 47140
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.805280160409936e-05,
+      "loss": 0.6344,
+      "step": 47150
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.7997103709479774e-05,
+      "loss": 0.5946,
+      "step": 47160
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.794140581486019e-05,
+      "loss": 0.5356,
+      "step": 47170
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.788570792024062e-05,
+      "loss": 0.5491,
+      "step": 47180
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.783001002562103e-05,
+      "loss": 0.5036,
+      "step": 47190
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.7774312131001446e-05,
+      "loss": 0.5586,
+      "step": 47200
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.771861423638186e-05,
+      "loss": 0.5447,
+      "step": 47210
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.7662916341762275e-05,
+      "loss": 0.4533,
+      "step": 47220
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.76072184471427e-05,
+      "loss": 0.5152,
+      "step": 47230
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.755152055252311e-05,
+      "loss": 0.5746,
+      "step": 47240
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.749582265790353e-05,
+      "loss": 0.5726,
+      "step": 47250
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.744012476328395e-05,
+      "loss": 0.6571,
+      "step": 47260
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.738442686866436e-05,
+      "loss": 0.5132,
+      "step": 47270
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.7328728974044776e-05,
+      "loss": 0.5532,
+      "step": 47280
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.7273031079425194e-05,
+      "loss": 0.5015,
+      "step": 47290
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.721733318480561e-05,
+      "loss": 0.4945,
+      "step": 47300
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.716163529018603e-05,
+      "loss": 0.6086,
+      "step": 47310
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.710593739556644e-05,
+      "loss": 0.4906,
+      "step": 47320
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.7050239500946865e-05,
+      "loss": 0.6731,
+      "step": 47330
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.6994541606327277e-05,
+      "loss": 0.5482,
+      "step": 47340
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.6938843711707694e-05,
+      "loss": 0.4853,
+      "step": 47350
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.688314581708811e-05,
+      "loss": 0.487,
+      "step": 47360
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.682744792246853e-05,
+      "loss": 0.5661,
+      "step": 47370
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.677175002784895e-05,
+      "loss": 0.4321,
+      "step": 47380
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.671605213322936e-05,
+      "loss": 0.6578,
+      "step": 47390
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.666035423860978e-05,
+      "loss": 0.654,
+      "step": 47400
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.6604656343990195e-05,
+      "loss": 0.5783,
+      "step": 47410
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.654895844937061e-05,
+      "loss": 0.4982,
+      "step": 47420
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.649326055475103e-05,
+      "loss": 0.5617,
+      "step": 47430
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.643756266013144e-05,
+      "loss": 0.5977,
+      "step": 47440
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.638186476551186e-05,
+      "loss": 0.5348,
+      "step": 47450
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.632616687089228e-05,
+      "loss": 0.5685,
+      "step": 47460
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.6270468976272696e-05,
+      "loss": 0.5385,
+      "step": 47470
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.621477108165311e-05,
+      "loss": 0.6513,
+      "step": 47480
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.6159073187033525e-05,
+      "loss": 0.5679,
+      "step": 47490
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.610337529241394e-05,
+      "loss": 0.502,
+      "step": 47500
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.604767739779436e-05,
+      "loss": 0.5285,
+      "step": 47510
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.599197950317478e-05,
+      "loss": 0.405,
+      "step": 47520
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.593628160855519e-05,
+      "loss": 0.6419,
+      "step": 47530
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.588058371393561e-05,
+      "loss": 0.5418,
+      "step": 47540
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.5824885819316026e-05,
+      "loss": 0.515,
+      "step": 47550
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.5769187924696444e-05,
+      "loss": 0.4902,
+      "step": 47560
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.571349003007686e-05,
+      "loss": 0.5365,
+      "step": 47570
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.565779213545727e-05,
+      "loss": 0.5141,
+      "step": 47580
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.56020942408377e-05,
+      "loss": 0.6049,
+      "step": 47590
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.554639634621811e-05,
+      "loss": 0.5635,
+      "step": 47600
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.5490698451598526e-05,
+      "loss": 0.6949,
+      "step": 47610
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.5435000556978944e-05,
+      "loss": 0.5147,
+      "step": 47620
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.537930266235936e-05,
+      "loss": 0.6236,
+      "step": 47630
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.532360476773978e-05,
+      "loss": 0.7037,
+      "step": 47640
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.526790687312019e-05,
+      "loss": 0.5828,
+      "step": 47650
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.521220897850061e-05,
+      "loss": 0.51,
+      "step": 47660
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.515651108388103e-05,
+      "loss": 0.5796,
+      "step": 47670
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.5100813189261445e-05,
+      "loss": 0.5492,
+      "step": 47680
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.504511529464186e-05,
+      "loss": 0.6647,
+      "step": 47690
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.4989417400022274e-05,
+      "loss": 0.5784,
+      "step": 47700
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.493371950540269e-05,
+      "loss": 0.7018,
+      "step": 47710
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.487802161078311e-05,
+      "loss": 0.6299,
+      "step": 47720
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.482232371616353e-05,
+      "loss": 0.5654,
+      "step": 47730
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.476662582154394e-05,
+      "loss": 0.5152,
+      "step": 47740
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.471092792692436e-05,
+      "loss": 0.5748,
+      "step": 47750
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.4655230032304775e-05,
+      "loss": 0.5715,
+      "step": 47760
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.459953213768519e-05,
+      "loss": 0.4943,
+      "step": 47770
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.454383424306561e-05,
+      "loss": 0.5994,
+      "step": 47780
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.448813634844602e-05,
+      "loss": 0.614,
+      "step": 47790
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.4432438453826446e-05,
+      "loss": 0.6671,
+      "step": 47800
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.437674055920686e-05,
+      "loss": 0.5859,
+      "step": 47810
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.4321042664587275e-05,
+      "loss": 0.5051,
+      "step": 47820
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.426534476996769e-05,
+      "loss": 0.6863,
+      "step": 47830
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.4209646875348104e-05,
+      "loss": 0.5615,
+      "step": 47840
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.415394898072853e-05,
+      "loss": 0.558,
+      "step": 47850
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.409825108610894e-05,
+      "loss": 0.5991,
+      "step": 47860
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.404255319148936e-05,
+      "loss": 0.6507,
+      "step": 47870
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.3986855296869776e-05,
+      "loss": 0.5231,
+      "step": 47880
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.3931157402250194e-05,
+      "loss": 0.6547,
+      "step": 47890
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.387545950763061e-05,
+      "loss": 0.6191,
+      "step": 47900
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.381976161301102e-05,
+      "loss": 0.5995,
+      "step": 47910
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.376406371839144e-05,
+      "loss": 0.6403,
+      "step": 47920
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.370836582377186e-05,
+      "loss": 0.5227,
+      "step": 47930
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.365266792915228e-05,
+      "loss": 0.6298,
+      "step": 47940
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.3596970034532695e-05,
+      "loss": 0.5616,
+      "step": 47950
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.3541272139913106e-05,
+      "loss": 0.563,
+      "step": 47960
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.3485574245293524e-05,
+      "loss": 0.6102,
+      "step": 47970
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.342987635067394e-05,
+      "loss": 0.5515,
+      "step": 47980
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.337417845605436e-05,
+      "loss": 0.5193,
+      "step": 47990
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.331848056143477e-05,
+      "loss": 0.6968,
+      "step": 48000
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.326278266681519e-05,
+      "loss": 0.5517,
+      "step": 48010
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.3207084772195606e-05,
+      "loss": 0.6192,
+      "step": 48020
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.3151386877576024e-05,
+      "loss": 0.6097,
+      "step": 48030
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.309568898295644e-05,
+      "loss": 0.5692,
+      "step": 48040
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.3039991088336853e-05,
+      "loss": 0.5861,
+      "step": 48050
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.298429319371728e-05,
+      "loss": 0.5805,
+      "step": 48060
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.292859529909769e-05,
+      "loss": 0.5638,
+      "step": 48070
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.287289740447811e-05,
+      "loss": 0.5188,
+      "step": 48080
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.2817199509858525e-05,
+      "loss": 0.5722,
+      "step": 48090
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.276150161523894e-05,
+      "loss": 0.5402,
+      "step": 48100
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.270580372061936e-05,
+      "loss": 0.5593,
+      "step": 48110
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.265010582599977e-05,
+      "loss": 0.7037,
+      "step": 48120
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.259440793138019e-05,
+      "loss": 0.5058,
+      "step": 48130
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.253871003676061e-05,
+      "loss": 0.5378,
+      "step": 48140
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.2483012142141026e-05,
+      "loss": 0.6221,
+      "step": 48150
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.2427314247521444e-05,
+      "loss": 0.5499,
+      "step": 48160
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.2371616352901855e-05,
+      "loss": 0.4452,
+      "step": 48170
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.231591845828227e-05,
+      "loss": 0.4904,
+      "step": 48180
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.226022056366269e-05,
+      "loss": 0.6345,
+      "step": 48190
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.220452266904311e-05,
+      "loss": 0.7162,
+      "step": 48200
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.2148824774423526e-05,
+      "loss": 0.4464,
+      "step": 48210
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.209312687980394e-05,
+      "loss": 0.5543,
+      "step": 48220
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.2037428985184356e-05,
+      "loss": 0.5664,
+      "step": 48230
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.1981731090564773e-05,
+      "loss": 0.5523,
+      "step": 48240
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.192603319594519e-05,
+      "loss": 0.5612,
+      "step": 48250
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.18703353013256e-05,
+      "loss": 0.4993,
+      "step": 48260
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.181463740670603e-05,
+      "loss": 0.6767,
+      "step": 48270
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.175893951208644e-05,
+      "loss": 0.5287,
+      "step": 48280
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.1703241617466856e-05,
+      "loss": 0.5508,
+      "step": 48290
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.1647543722847274e-05,
+      "loss": 0.6466,
+      "step": 48300
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.1591845828227685e-05,
+      "loss": 0.5184,
+      "step": 48310
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.153614793360811e-05,
+      "loss": 0.4867,
+      "step": 48320
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.148045003898852e-05,
+      "loss": 0.5099,
+      "step": 48330
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.142475214436894e-05,
+      "loss": 0.5857,
+      "step": 48340
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.136905424974936e-05,
+      "loss": 0.5391,
+      "step": 48350
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.1313356355129775e-05,
+      "loss": 0.5064,
+      "step": 48360
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.125765846051019e-05,
+      "loss": 0.6499,
+      "step": 48370
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.1201960565890604e-05,
+      "loss": 0.5281,
+      "step": 48380
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.114626267127102e-05,
+      "loss": 0.6135,
+      "step": 48390
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.109056477665144e-05,
+      "loss": 0.5571,
+      "step": 48400
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.103486688203186e-05,
+      "loss": 0.6009,
+      "step": 48410
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.0979168987412276e-05,
+      "loss": 0.523,
+      "step": 48420
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.092347109279269e-05,
+      "loss": 0.4927,
+      "step": 48430
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.086777319817311e-05,
+      "loss": 0.739,
+      "step": 48440
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.081207530355352e-05,
+      "loss": 0.4902,
+      "step": 48450
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.075637740893394e-05,
+      "loss": 0.5542,
+      "step": 48460
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.070067951431436e-05,
+      "loss": 0.484,
+      "step": 48470
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.064498161969477e-05,
+      "loss": 0.5727,
+      "step": 48480
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.058928372507519e-05,
+      "loss": 0.5418,
+      "step": 48490
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.0533585830455605e-05,
+      "loss": 0.5902,
+      "step": 48500
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.0477887935836023e-05,
+      "loss": 0.5889,
+      "step": 48510
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.0422190041216438e-05,
+      "loss": 0.6031,
+      "step": 48520
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.036649214659686e-05,
+      "loss": 0.5224,
+      "step": 48530
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.0310794251977274e-05,
+      "loss": 0.5233,
+      "step": 48540
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.0255096357357688e-05,
+      "loss": 0.542,
+      "step": 48550
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.0199398462738106e-05,
+      "loss": 0.531,
+      "step": 48560
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.0149270357580482e-05,
+      "loss": 0.6225,
+      "step": 48570
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.0093572462960897e-05,
+      "loss": 0.6596,
+      "step": 48580
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.0037874568341315e-05,
+      "loss": 0.7282,
+      "step": 48590
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.998217667372173e-05,
+      "loss": 0.5325,
+      "step": 48600
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.992647877910215e-05,
+      "loss": 0.6451,
+      "step": 48610
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.9870780884482565e-05,
+      "loss": 0.64,
+      "step": 48620
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.981508298986298e-05,
+      "loss": 0.5696,
+      "step": 48630
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.9759385095243397e-05,
+      "loss": 0.5518,
+      "step": 48640
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.9703687200623815e-05,
+      "loss": 0.5646,
+      "step": 48650
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.964798930600423e-05,
+      "loss": 0.4954,
+      "step": 48660
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.9592291411384648e-05,
+      "loss": 0.7372,
+      "step": 48670
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.9536593516765062e-05,
+      "loss": 0.5751,
+      "step": 48680
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.9480895622145477e-05,
+      "loss": 0.5941,
+      "step": 48690
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.9425197727525898e-05,
+      "loss": 0.5785,
+      "step": 48700
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.9369499832906313e-05,
+      "loss": 0.4956,
+      "step": 48710
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.931380193828673e-05,
+      "loss": 0.7268,
+      "step": 48720
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.9258104043667145e-05,
+      "loss": 0.5952,
+      "step": 48730
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.9202406149047566e-05,
+      "loss": 0.6268,
+      "step": 48740
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.914670825442798e-05,
+      "loss": 0.4495,
+      "step": 48750
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.9091010359808395e-05,
+      "loss": 0.7017,
+      "step": 48760
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.9035312465188813e-05,
+      "loss": 0.5829,
+      "step": 48770
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.897961457056923e-05,
+      "loss": 0.5505,
+      "step": 48780
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.892391667594965e-05,
+      "loss": 0.5899,
+      "step": 48790
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.8868218781330064e-05,
+      "loss": 0.5719,
+      "step": 48800
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.8812520886710478e-05,
+      "loss": 0.5271,
+      "step": 48810
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.87568229920909e-05,
+      "loss": 0.507,
+      "step": 48820
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.8701125097471314e-05,
+      "loss": 0.5156,
+      "step": 48830
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.864542720285173e-05,
+      "loss": 0.5059,
+      "step": 48840
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.8589729308232146e-05,
+      "loss": 0.5778,
+      "step": 48850
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.853403141361256e-05,
+      "loss": 0.6351,
+      "step": 48860
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.8478333518992982e-05,
+      "loss": 0.6452,
+      "step": 48870
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.8422635624373397e-05,
+      "loss": 0.5668,
+      "step": 48880
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.836693772975381e-05,
+      "loss": 0.7163,
+      "step": 48890
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.831123983513423e-05,
+      "loss": 0.5634,
+      "step": 48900
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.8255541940514647e-05,
+      "loss": 0.6061,
+      "step": 48910
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.8199844045895065e-05,
+      "loss": 0.5385,
+      "step": 48920
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.814414615127548e-05,
+      "loss": 0.4921,
+      "step": 48930
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.8094018046117852e-05,
+      "loss": 0.502,
+      "step": 48940
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.8038320151498274e-05,
+      "loss": 0.6084,
+      "step": 48950
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.7982622256878688e-05,
+      "loss": 0.5424,
+      "step": 48960
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.7926924362259103e-05,
+      "loss": 0.5544,
+      "step": 48970
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.787122646763952e-05,
+      "loss": 0.5997,
+      "step": 48980
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.781552857301994e-05,
+      "loss": 0.642,
+      "step": 48990
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.7759830678400356e-05,
+      "loss": 0.5417,
+      "step": 49000
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.770413278378077e-05,
+      "loss": 0.5882,
+      "step": 49010
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.7648434889161185e-05,
+      "loss": 0.5367,
+      "step": 49020
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.7592736994541603e-05,
+      "loss": 0.6134,
+      "step": 49030
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.753703909992202e-05,
+      "loss": 0.6281,
+      "step": 49040
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.7481341205302436e-05,
+      "loss": 0.5598,
+      "step": 49050
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.7425643310682854e-05,
+      "loss": 0.4418,
+      "step": 49060
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.7369945416063268e-05,
+      "loss": 0.5377,
+      "step": 49070
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.731424752144369e-05,
+      "loss": 0.5303,
+      "step": 49080
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.7258549626824104e-05,
+      "loss": 0.6383,
+      "step": 49090
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.720285173220452e-05,
+      "loss": 0.5223,
+      "step": 49100
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.7147153837584936e-05,
+      "loss": 0.5903,
+      "step": 49110
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.7091455942965354e-05,
+      "loss": 0.5445,
+      "step": 49120
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.7035758048345772e-05,
+      "loss": 0.5708,
+      "step": 49130
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.6980060153726187e-05,
+      "loss": 0.5344,
+      "step": 49140
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.69243622591066e-05,
+      "loss": 0.5863,
+      "step": 49150
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.6868664364487023e-05,
+      "loss": 0.574,
+      "step": 49160
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.6812966469867437e-05,
+      "loss": 0.4661,
+      "step": 49170
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.675726857524785e-05,
+      "loss": 0.5607,
+      "step": 49180
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.670157068062827e-05,
+      "loss": 0.6135,
+      "step": 49190
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.6645872786008684e-05,
+      "loss": 0.5096,
+      "step": 49200
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.6590174891389105e-05,
+      "loss": 0.6515,
+      "step": 49210
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.653447699676952e-05,
+      "loss": 0.6139,
+      "step": 49220
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.6478779102149934e-05,
+      "loss": 0.606,
+      "step": 49230
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.6423081207530352e-05,
+      "loss": 0.5841,
+      "step": 49240
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.636738331291077e-05,
+      "loss": 0.5122,
+      "step": 49250
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.6311685418291188e-05,
+      "loss": 0.67,
+      "step": 49260
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.6255987523671603e-05,
+      "loss": 0.6388,
+      "step": 49270
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.6200289629052017e-05,
+      "loss": 0.5996,
+      "step": 49280
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.614459173443244e-05,
+      "loss": 0.5778,
+      "step": 49290
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.6088893839812853e-05,
+      "loss": 0.6016,
+      "step": 49300
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.6033195945193268e-05,
+      "loss": 0.5354,
+      "step": 49310
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.5977498050573685e-05,
+      "loss": 0.5299,
+      "step": 49320
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.5921800155954103e-05,
+      "loss": 0.6332,
+      "step": 49330
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.586610226133452e-05,
+      "loss": 0.559,
+      "step": 49340
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.5810404366714936e-05,
+      "loss": 0.5202,
+      "step": 49350
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.575470647209535e-05,
+      "loss": 0.6037,
+      "step": 49360
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.5699008577475768e-05,
+      "loss": 0.5897,
+      "step": 49370
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.5643310682856186e-05,
+      "loss": 0.4251,
+      "step": 49380
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.5587612788236604e-05,
+      "loss": 0.5238,
+      "step": 49390
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.553191489361702e-05,
+      "loss": 0.65,
+      "step": 49400
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.5476216998997433e-05,
+      "loss": 0.4997,
+      "step": 49410
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.5420519104377854e-05,
+      "loss": 0.5812,
+      "step": 49420
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.536482120975827e-05,
+      "loss": 0.529,
+      "step": 49430
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.5309123315138687e-05,
+      "loss": 0.447,
+      "step": 49440
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.52534254205191e-05,
+      "loss": 0.5349,
+      "step": 49450
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.519772752589952e-05,
+      "loss": 0.6312,
+      "step": 49460
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.5142029631279937e-05,
+      "loss": 0.4897,
+      "step": 49470
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.5086331736660352e-05,
+      "loss": 0.6419,
+      "step": 49480
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.5030633842040766e-05,
+      "loss": 0.5988,
+      "step": 49490
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4974935947421184e-05,
+      "loss": 0.3872,
+      "step": 49500
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4919238052801602e-05,
+      "loss": 0.5554,
+      "step": 49510
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.486354015818202e-05,
+      "loss": 0.569,
+      "step": 49520
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4807842263562435e-05,
+      "loss": 0.7036,
+      "step": 49530
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.475214436894285e-05,
+      "loss": 0.4347,
+      "step": 49540
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.469644647432327e-05,
+      "loss": 0.5357,
+      "step": 49550
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4640748579703685e-05,
+      "loss": 0.5829,
+      "step": 49560
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4585050685084103e-05,
+      "loss": 0.5407,
+      "step": 49570
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4529352790464517e-05,
+      "loss": 0.5562,
+      "step": 49580
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4473654895844935e-05,
+      "loss": 0.5497,
+      "step": 49590
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4417957001225353e-05,
+      "loss": 0.5843,
+      "step": 49600
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4362259106605768e-05,
+      "loss": 0.6302,
+      "step": 49610
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4306561211986182e-05,
+      "loss": 0.5279,
+      "step": 49620
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4250863317366603e-05,
+      "loss": 0.6376,
+      "step": 49630
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4195165422747018e-05,
+      "loss": 0.49,
+      "step": 49640
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4139467528127436e-05,
+      "loss": 0.5348,
+      "step": 49650
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.408376963350785e-05,
+      "loss": 0.6197,
+      "step": 49660
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4028071738888265e-05,
+      "loss": 0.5806,
+      "step": 49670
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.3972373844268686e-05,
+      "loss": 0.5545,
+      "step": 49680
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.39166759496491e-05,
+      "loss": 0.6144,
+      "step": 49690
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.386097805502952e-05,
+      "loss": 0.4794,
+      "step": 49700
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.3805280160409933e-05,
+      "loss": 0.5592,
+      "step": 49710
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.374958226579035e-05,
+      "loss": 0.5108,
+      "step": 49720
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.369388437117077e-05,
+      "loss": 0.4816,
+      "step": 49730
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.3638186476551184e-05,
+      "loss": 0.534,
+      "step": 49740
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.3582488581931598e-05,
+      "loss": 0.7042,
+      "step": 49750
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.352679068731202e-05,
+      "loss": 0.6161,
+      "step": 49760
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.3471092792692434e-05,
+      "loss": 0.4955,
+      "step": 49770
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.3415394898072852e-05,
+      "loss": 0.5406,
+      "step": 49780
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.3359697003453266e-05,
+      "loss": 0.47,
+      "step": 49790
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.3303999108833688e-05,
+      "loss": 0.5203,
+      "step": 49800
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.3248301214214102e-05,
+      "loss": 0.4684,
+      "step": 49810
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.3192603319594517e-05,
+      "loss": 0.64,
+      "step": 49820
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.3136905424974935e-05,
+      "loss": 0.5931,
+      "step": 49830
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.308120753035535e-05,
+      "loss": 0.607,
+      "step": 49840
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.3025509635735767e-05,
+      "loss": 0.4432,
+      "step": 49850
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.2969811741116185e-05,
+      "loss": 0.5324,
+      "step": 49860
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.29141138464966e-05,
+      "loss": 0.5889,
+      "step": 49870
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.2858415951877014e-05,
+      "loss": 0.4878,
+      "step": 49880
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.2802718057257435e-05,
+      "loss": 0.5944,
+      "step": 49890
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.274702016263785e-05,
+      "loss": 0.6185,
+      "step": 49900
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.2691322268018268e-05,
+      "loss": 0.656,
+      "step": 49910
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.2635624373398682e-05,
+      "loss": 0.5864,
+      "step": 49920
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.2579926478779104e-05,
+      "loss": 0.4542,
+      "step": 49930
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.2524228584159518e-05,
+      "loss": 0.6639,
+      "step": 49940
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.2468530689539933e-05,
+      "loss": 0.7369,
+      "step": 49950
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.241283279492035e-05,
+      "loss": 0.5868,
+      "step": 49960
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.2357134900300765e-05,
+      "loss": 0.5113,
+      "step": 49970
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.2301437005681183e-05,
+      "loss": 0.5578,
+      "step": 49980
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.22457391110616e-05,
+      "loss": 0.526,
+      "step": 49990
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.2190041216442015e-05,
+      "loss": 0.6395,
+      "step": 50000
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.213434332182243e-05,
+      "loss": 0.7001,
+      "step": 50010
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.207864542720285e-05,
+      "loss": 0.4927,
+      "step": 50020
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.2022947532583266e-05,
+      "loss": 0.58,
+      "step": 50030
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.1967249637963684e-05,
+      "loss": 0.5076,
+      "step": 50040
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.1911551743344098e-05,
+      "loss": 0.5777,
+      "step": 50050
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.185585384872452e-05,
+      "loss": 0.6302,
+      "step": 50060
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.1800155954104934e-05,
+      "loss": 0.5982,
+      "step": 50070
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.174445805948535e-05,
+      "loss": 0.4589,
+      "step": 50080
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.1688760164865766e-05,
+      "loss": 0.5075,
+      "step": 50090
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.1633062270246184e-05,
+      "loss": 0.5355,
+      "step": 50100
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.15773643756266e-05,
+      "loss": 0.5174,
+      "step": 50110
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.1521666481007017e-05,
+      "loss": 0.6385,
+      "step": 50120
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.146596858638743e-05,
+      "loss": 0.648,
+      "step": 50130
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.1410270691767846e-05,
+      "loss": 0.4629,
+      "step": 50140
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.1354572797148267e-05,
+      "loss": 0.5356,
+      "step": 50150
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.129887490252868e-05,
+      "loss": 0.589,
+      "step": 50160
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.12431770079091e-05,
+      "loss": 0.485,
+      "step": 50170
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.1187479113289514e-05,
+      "loss": 0.6254,
+      "step": 50180
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.1131781218669935e-05,
+      "loss": 0.4952,
+      "step": 50190
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.107608332405035e-05,
+      "loss": 0.5772,
+      "step": 50200
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.1020385429430764e-05,
+      "loss": 0.5264,
+      "step": 50210
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.0964687534811182e-05,
+      "loss": 0.448,
+      "step": 50220
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.09089896401916e-05,
+      "loss": 0.523,
+      "step": 50230
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.0853291745572015e-05,
+      "loss": 0.5539,
+      "step": 50240
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.0797593850952433e-05,
+      "loss": 0.5861,
+      "step": 50250
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.0741895956332847e-05,
+      "loss": 0.5885,
+      "step": 50260
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.068619806171327e-05,
+      "loss": 0.6078,
+      "step": 50270
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.0630500167093683e-05,
+      "loss": 0.6467,
+      "step": 50280
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.0574802272474098e-05,
+      "loss": 0.5114,
+      "step": 50290
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.0519104377854516e-05,
+      "loss": 0.5043,
+      "step": 50300
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.046340648323493e-05,
+      "loss": 0.7764,
+      "step": 50310
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.040770858861535e-05,
+      "loss": 0.5989,
+      "step": 50320
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.0352010693995766e-05,
+      "loss": 0.5397,
+      "step": 50330
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.029631279937618e-05,
+      "loss": 0.6624,
+      "step": 50340
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.0240614904756598e-05,
+      "loss": 0.5074,
+      "step": 50350
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.0184917010137016e-05,
+      "loss": 0.6125,
+      "step": 50360
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.012921911551743e-05,
+      "loss": 0.6758,
+      "step": 50370
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.007352122089785e-05,
+      "loss": 0.7117,
+      "step": 50380
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.0017823326278263e-05,
+      "loss": 0.7379,
+      "step": 50390
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.9962125431658684e-05,
+      "loss": 0.4451,
+      "step": 50400
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.99064275370391e-05,
+      "loss": 0.4612,
+      "step": 50410
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.9850729642419514e-05,
+      "loss": 0.5497,
+      "step": 50420
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.979503174779993e-05,
+      "loss": 0.6251,
+      "step": 50430
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.9739333853180346e-05,
+      "loss": 0.5437,
+      "step": 50440
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.9683635958560767e-05,
+      "loss": 0.5941,
+      "step": 50450
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.9627938063941182e-05,
+      "loss": 0.5474,
+      "step": 50460
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.9572240169321596e-05,
+      "loss": 0.5998,
+      "step": 50470
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.9516542274702014e-05,
+      "loss": 0.5055,
+      "step": 50480
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.9460844380082432e-05,
+      "loss": 0.5535,
+      "step": 50490
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.9405146485462847e-05,
+      "loss": 0.5476,
+      "step": 50500
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.9349448590843265e-05,
+      "loss": 0.5119,
+      "step": 50510
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.929375069622368e-05,
+      "loss": 0.5713,
+      "step": 50520
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.92380528016041e-05,
+      "loss": 0.4773,
+      "step": 50530
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.9182354906984515e-05,
+      "loss": 0.5768,
+      "step": 50540
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.912665701236493e-05,
+      "loss": 0.5221,
+      "step": 50550
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.9070959117745347e-05,
+      "loss": 0.7206,
+      "step": 50560
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.9015261223125765e-05,
+      "loss": 0.4983,
+      "step": 50570
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.8959563328506183e-05,
+      "loss": 0.5234,
+      "step": 50580
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8903865433886598e-05,
+      "loss": 0.521,
+      "step": 50590
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8848167539267012e-05,
+      "loss": 0.5795,
+      "step": 50600
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.879246964464743e-05,
+      "loss": 0.6918,
+      "step": 50610
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8736771750027848e-05,
+      "loss": 0.6866,
+      "step": 50620
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8681073855408263e-05,
+      "loss": 0.6993,
+      "step": 50630
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.862537596078868e-05,
+      "loss": 0.6497,
+      "step": 50640
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.85696780661691e-05,
+      "loss": 0.4933,
+      "step": 50650
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8513980171549513e-05,
+      "loss": 0.6868,
+      "step": 50660
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.845828227692993e-05,
+      "loss": 0.5818,
+      "step": 50670
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8402584382310345e-05,
+      "loss": 0.5683,
+      "step": 50680
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8346886487690763e-05,
+      "loss": 0.4418,
+      "step": 50690
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.829118859307118e-05,
+      "loss": 0.5005,
+      "step": 50700
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.82354906984516e-05,
+      "loss": 0.5944,
+      "step": 50710
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8179792803832014e-05,
+      "loss": 0.7682,
+      "step": 50720
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.812409490921243e-05,
+      "loss": 0.4511,
+      "step": 50730
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8068397014592846e-05,
+      "loss": 0.5313,
+      "step": 50740
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.8012699119973264e-05,
+      "loss": 0.5761,
+      "step": 50750
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7957001225353682e-05,
+      "loss": 0.6557,
+      "step": 50760
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7901303330734096e-05,
+      "loss": 0.5167,
+      "step": 50770
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7845605436114514e-05,
+      "loss": 0.5412,
+      "step": 50780
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.778990754149493e-05,
+      "loss": 0.5521,
+      "step": 50790
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7734209646875347e-05,
+      "loss": 0.537,
+      "step": 50800
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.767851175225576e-05,
+      "loss": 0.5333,
+      "step": 50810
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.762281385763618e-05,
+      "loss": 0.5173,
+      "step": 50820
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7567115963016597e-05,
+      "loss": 0.4469,
+      "step": 50830
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.7511418068397015e-05,
+      "loss": 0.5463,
+      "step": 50840
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.745572017377743e-05,
+      "loss": 0.5388,
+      "step": 50850
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.7400022279157847e-05,
+      "loss": 0.6881,
+      "step": 50860
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.7344324384538262e-05,
+      "loss": 0.5108,
+      "step": 50870
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.728862648991868e-05,
+      "loss": 0.4259,
+      "step": 50880
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.7232928595299098e-05,
+      "loss": 0.5143,
+      "step": 50890
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.7177230700679512e-05,
+      "loss": 0.4401,
+      "step": 50900
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.712153280605993e-05,
+      "loss": 0.5805,
+      "step": 50910
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.7065834911440345e-05,
+      "loss": 0.5834,
+      "step": 50920
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.7010137016820763e-05,
+      "loss": 0.6101,
+      "step": 50930
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6954439122201177e-05,
+      "loss": 0.6045,
+      "step": 50940
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6898741227581595e-05,
+      "loss": 0.5697,
+      "step": 50950
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6843043332962013e-05,
+      "loss": 0.5748,
+      "step": 50960
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.678734543834243e-05,
+      "loss": 0.5727,
+      "step": 50970
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6731647543722845e-05,
+      "loss": 0.4418,
+      "step": 50980
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6675949649103263e-05,
+      "loss": 0.6387,
+      "step": 50990
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6620251754483678e-05,
+      "loss": 0.5258,
+      "step": 51000
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6564553859864096e-05,
+      "loss": 0.4365,
+      "step": 51010
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6508855965244514e-05,
+      "loss": 0.5777,
+      "step": 51020
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6453158070624928e-05,
+      "loss": 0.5431,
+      "step": 51030
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6397460176005346e-05,
+      "loss": 0.6415,
+      "step": 51040
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6341762281385764e-05,
+      "loss": 0.5364,
+      "step": 51050
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.628606438676618e-05,
+      "loss": 0.5547,
+      "step": 51060
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6230366492146593e-05,
+      "loss": 0.531,
+      "step": 51070
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.617466859752701e-05,
+      "loss": 0.5839,
+      "step": 51080
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.611897070290743e-05,
+      "loss": 0.6383,
+      "step": 51090
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6063272808287847e-05,
+      "loss": 0.4692,
+      "step": 51100
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.600757491366826e-05,
+      "loss": 0.4786,
+      "step": 51110
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.595187701904868e-05,
+      "loss": 0.741,
+      "step": 51120
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.5896179124429094e-05,
+      "loss": 0.5523,
+      "step": 51130
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.5840481229809512e-05,
+      "loss": 0.6348,
+      "step": 51140
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.578478333518993e-05,
+      "loss": 0.5444,
+      "step": 51150
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.5729085440570344e-05,
+      "loss": 0.551,
+      "step": 51160
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.5673387545950762e-05,
+      "loss": 0.5785,
+      "step": 51170
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.561768965133118e-05,
+      "loss": 0.6517,
+      "step": 51180
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.5561991756711594e-05,
+      "loss": 0.4782,
+      "step": 51190
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.5506293862092012e-05,
+      "loss": 0.6313,
+      "step": 51200
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.5450595967472427e-05,
+      "loss": 0.6293,
+      "step": 51210
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.5394898072852845e-05,
+      "loss": 0.5385,
+      "step": 51220
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.5339200178233263e-05,
+      "loss": 0.5399,
+      "step": 51230
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.5283502283613677e-05,
+      "loss": 0.6111,
+      "step": 51240
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.5227804388994095e-05,
+      "loss": 0.7043,
+      "step": 51250
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.5172106494374511e-05,
+      "loss": 0.4684,
+      "step": 51260
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.5116408599754928e-05,
+      "loss": 0.5684,
+      "step": 51270
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.5060710705135344e-05,
+      "loss": 0.3748,
+      "step": 51280
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.5005012810515762e-05,
+      "loss": 0.6251,
+      "step": 51290
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4949314915896178e-05,
+      "loss": 0.5439,
+      "step": 51300
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4893617021276596e-05,
+      "loss": 0.5666,
+      "step": 51310
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.483791912665701e-05,
+      "loss": 0.6856,
+      "step": 51320
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4782221232037428e-05,
+      "loss": 0.5606,
+      "step": 51330
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4726523337417845e-05,
+      "loss": 0.5947,
+      "step": 51340
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4670825442798262e-05,
+      "loss": 0.5522,
+      "step": 51350
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4615127548178677e-05,
+      "loss": 0.5981,
+      "step": 51360
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4559429653559093e-05,
+      "loss": 0.5534,
+      "step": 51370
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4503731758939511e-05,
+      "loss": 0.4942,
+      "step": 51380
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.4448033864319927e-05,
+      "loss": 0.4048,
+      "step": 51390
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.4392335969700344e-05,
+      "loss": 0.4822,
+      "step": 51400
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.433663807508076e-05,
+      "loss": 0.5823,
+      "step": 51410
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.4280940180461178e-05,
+      "loss": 0.6042,
+      "step": 51420
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.4225242285841594e-05,
+      "loss": 0.5983,
+      "step": 51430
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.4169544391222012e-05,
+      "loss": 0.4227,
+      "step": 51440
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.4113846496602426e-05,
+      "loss": 0.4886,
+      "step": 51450
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.4058148601982844e-05,
+      "loss": 0.5644,
+      "step": 51460
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.400245070736326e-05,
+      "loss": 0.5254,
+      "step": 51470
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.3946752812743678e-05,
+      "loss": 0.4171,
+      "step": 51480
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.3891054918124093e-05,
+      "loss": 0.586,
+      "step": 51490
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.383535702350451e-05,
+      "loss": 0.4959,
+      "step": 51500
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.3779659128884927e-05,
+      "loss": 0.6238,
+      "step": 51510
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.3723961234265345e-05,
+      "loss": 0.53,
+      "step": 51520
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.366826333964576e-05,
+      "loss": 0.428,
+      "step": 51530
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.3612565445026176e-05,
+      "loss": 0.576,
+      "step": 51540
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.3556867550406594e-05,
+      "loss": 0.5181,
+      "step": 51550
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.350116965578701e-05,
+      "loss": 0.5524,
+      "step": 51560
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.3445471761167428e-05,
+      "loss": 0.4988,
+      "step": 51570
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.3389773866547842e-05,
+      "loss": 0.5608,
+      "step": 51580
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.333407597192826e-05,
+      "loss": 0.5204,
+      "step": 51590
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.3278378077308676e-05,
+      "loss": 0.5184,
+      "step": 51600
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.3222680182689094e-05,
+      "loss": 0.5935,
+      "step": 51610
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.3166982288069509e-05,
+      "loss": 0.4812,
+      "step": 51620
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.3111284393449927e-05,
+      "loss": 0.6615,
+      "step": 51630
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.3055586498830343e-05,
+      "loss": 0.5362,
+      "step": 51640
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.2999888604210761e-05,
+      "loss": 0.6029,
+      "step": 51650
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.2944190709591175e-05,
+      "loss": 0.698,
+      "step": 51660
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.2888492814971593e-05,
+      "loss": 0.5691,
+      "step": 51670
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.283279492035201e-05,
+      "loss": 0.5093,
+      "step": 51680
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.2777097025732426e-05,
+      "loss": 0.6178,
+      "step": 51690
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.2721399131112844e-05,
+      "loss": 0.4218,
+      "step": 51700
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.2665701236493258e-05,
+      "loss": 0.6245,
+      "step": 51710
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.2610003341873676e-05,
+      "loss": 0.5696,
+      "step": 51720
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.2554305447254092e-05,
+      "loss": 0.552,
+      "step": 51730
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.249860755263451e-05,
+      "loss": 0.5483,
+      "step": 51740
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.2442909658014925e-05,
+      "loss": 0.5139,
+      "step": 51750
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.2387211763395343e-05,
+      "loss": 0.5659,
+      "step": 51760
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.2331513868775759e-05,
+      "loss": 0.557,
+      "step": 51770
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.2275815974156177e-05,
+      "loss": 0.622,
+      "step": 51780
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.2220118079536593e-05,
+      "loss": 0.5288,
+      "step": 51790
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.216442018491701e-05,
+      "loss": 0.5918,
+      "step": 51800
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.2108722290297425e-05,
+      "loss": 0.4625,
+      "step": 51810
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.2053024395677843e-05,
+      "loss": 0.7206,
+      "step": 51820
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.199732650105826e-05,
+      "loss": 0.4559,
+      "step": 51830
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.1941628606438674e-05,
+      "loss": 0.5578,
+      "step": 51840
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.1885930711819092e-05,
+      "loss": 0.6598,
+      "step": 51850
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.1830232817199508e-05,
+      "loss": 0.5262,
+      "step": 51860
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.1774534922579926e-05,
+      "loss": 0.5228,
+      "step": 51870
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.171883702796034e-05,
+      "loss": 0.5485,
+      "step": 51880
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.1663139133340759e-05,
+      "loss": 0.5205,
+      "step": 51890
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.1607441238721175e-05,
+      "loss": 0.5783,
+      "step": 51900
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.1551743344101593e-05,
+      "loss": 0.5976,
+      "step": 51910
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.1496045449482009e-05,
+      "loss": 0.5747,
+      "step": 51920
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.1440347554862425e-05,
+      "loss": 0.6186,
+      "step": 51930
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.1384649660242841e-05,
+      "loss": 0.4475,
+      "step": 51940
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.132895176562326e-05,
+      "loss": 0.8409,
+      "step": 51950
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.1273253871003675e-05,
+      "loss": 0.5911,
+      "step": 51960
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.1217555976384093e-05,
+      "loss": 0.6518,
+      "step": 51970
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.1161858081764508e-05,
+      "loss": 0.5714,
+      "step": 51980
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.1106160187144924e-05,
+      "loss": 0.5216,
+      "step": 51990
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.1050462292525342e-05,
+      "loss": 0.5451,
+      "step": 52000
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0994764397905757e-05,
+      "loss": 0.5307,
+      "step": 52010
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0939066503286174e-05,
+      "loss": 0.5078,
+      "step": 52020
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.088336860866659e-05,
+      "loss": 0.7074,
+      "step": 52030
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0827670714047009e-05,
+      "loss": 0.4657,
+      "step": 52040
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0771972819427425e-05,
+      "loss": 0.6012,
+      "step": 52050
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0716274924807841e-05,
+      "loss": 0.5066,
+      "step": 52060
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0660577030188257e-05,
+      "loss": 0.4725,
+      "step": 52070
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0604879135568675e-05,
+      "loss": 0.5967,
+      "step": 52080
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0549181240949091e-05,
+      "loss": 0.4897,
+      "step": 52090
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.049348334632951e-05,
+      "loss": 0.5391,
+      "step": 52100
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0437785451709924e-05,
+      "loss": 0.4885,
+      "step": 52110
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0382087557090342e-05,
+      "loss": 0.577,
+      "step": 52120
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0326389662470758e-05,
+      "loss": 0.5491,
+      "step": 52130
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0270691767851176e-05,
+      "loss": 0.5647,
+      "step": 52140
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.021499387323159e-05,
+      "loss": 0.4779,
+      "step": 52150
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0159295978612007e-05,
+      "loss": 0.5625,
+      "step": 52160
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.0103598083992425e-05,
+      "loss": 0.555,
+      "step": 52170
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.004790018937284e-05,
+      "loss": 0.5129,
+      "step": 52180
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 9.992202294753257e-06,
+      "loss": 0.5839,
+      "step": 52190
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 9.936504400133673e-06,
+      "loss": 0.5261,
+      "step": 52200
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.880806505514091e-06,
+      "loss": 0.6553,
+      "step": 52210
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.825108610894507e-06,
+      "loss": 0.4706,
+      "step": 52220
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.769410716274925e-06,
+      "loss": 0.6074,
+      "step": 52230
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.71371282165534e-06,
+      "loss": 0.4789,
+      "step": 52240
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.658014927035758e-06,
+      "loss": 0.5131,
+      "step": 52250
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.602317032416174e-06,
+      "loss": 0.5238,
+      "step": 52260
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.546619137796592e-06,
+      "loss": 0.606,
+      "step": 52270
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.490921243177006e-06,
+      "loss": 0.6541,
+      "step": 52280
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.435223348557424e-06,
+      "loss": 0.6714,
+      "step": 52290
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.37952545393784e-06,
+      "loss": 0.6529,
+      "step": 52300
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.323827559318257e-06,
+      "loss": 0.5762,
+      "step": 52310
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.268129664698673e-06,
+      "loss": 0.5856,
+      "step": 52320
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.21243177007909e-06,
+      "loss": 0.5428,
+      "step": 52330
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.156733875459507e-06,
+      "loss": 0.5138,
+      "step": 52340
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.101035980839923e-06,
+      "loss": 0.4142,
+      "step": 52350
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.045338086220341e-06,
+      "loss": 0.5272,
+      "step": 52360
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.989640191600757e-06,
+      "loss": 0.5971,
+      "step": 52370
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.933942296981174e-06,
+      "loss": 0.6514,
+      "step": 52380
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.87824440236159e-06,
+      "loss": 0.5154,
+      "step": 52390
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.822546507742006e-06,
+      "loss": 0.4522,
+      "step": 52400
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.766848613122422e-06,
+      "loss": 0.7203,
+      "step": 52410
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.71115071850284e-06,
+      "loss": 0.4902,
+      "step": 52420
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.655452823883256e-06,
+      "loss": 0.4589,
+      "step": 52430
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.599754929263673e-06,
+      "loss": 0.5249,
+      "step": 52440
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.54405703464409e-06,
+      "loss": 0.571,
+      "step": 52450
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.488359140024507e-06,
+      "loss": 0.5981,
+      "step": 52460
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.432661245404923e-06,
+      "loss": 0.5864,
+      "step": 52470
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.376963350785339e-06,
+      "loss": 0.625,
+      "step": 52480
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.321265456165757e-06,
+      "loss": 0.7408,
+      "step": 52490
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.265567561546173e-06,
+      "loss": 0.5908,
+      "step": 52500
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.20986966692659e-06,
+      "loss": 0.558,
+      "step": 52510
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.154171772307006e-06,
+      "loss": 0.6495,
+      "step": 52520
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.098473877687424e-06,
+      "loss": 0.6201,
+      "step": 52530
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.042775983067838e-06,
+      "loss": 0.4979,
+      "step": 52540
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.987078088448256e-06,
+      "loss": 0.605,
+      "step": 52550
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.931380193828672e-06,
+      "loss": 0.4532,
+      "step": 52560
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.875682299209088e-06,
+      "loss": 0.5578,
+      "step": 52570
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.819984404589506e-06,
+      "loss": 0.5913,
+      "step": 52580
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.764286509969923e-06,
+      "loss": 0.6065,
+      "step": 52590
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.708588615350339e-06,
+      "loss": 0.575,
+      "step": 52600
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.652890720730755e-06,
+      "loss": 0.5402,
+      "step": 52610
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.597192826111172e-06,
+      "loss": 0.5447,
+      "step": 52620
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.541494931491589e-06,
+      "loss": 0.7365,
+      "step": 52630
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.485797036872005e-06,
+      "loss": 0.5948,
+      "step": 52640
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.4300991422524225e-06,
+      "loss": 0.613,
+      "step": 52650
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.3744012476328395e-06,
+      "loss": 0.5444,
+      "step": 52660
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.318703353013256e-06,
+      "loss": 0.6483,
+      "step": 52670
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.263005458393673e-06,
+      "loss": 0.5607,
+      "step": 52680
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.207307563774088e-06,
+      "loss": 0.438,
+      "step": 52690
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.151609669154505e-06,
+      "loss": 0.555,
+      "step": 52700
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.0959117745349215e-06,
+      "loss": 0.7265,
+      "step": 52710
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.0402138799153385e-06,
+      "loss": 0.559,
+      "step": 52720
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.984515985295755e-06,
+      "loss": 0.6297,
+      "step": 52730
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.928818090676172e-06,
+      "loss": 0.6368,
+      "step": 52740
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.873120196056588e-06,
+      "loss": 0.6207,
+      "step": 52750
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.817422301437005e-06,
+      "loss": 0.6341,
+      "step": 52760
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.761724406817421e-06,
+      "loss": 0.491,
+      "step": 52770
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.706026512197838e-06,
+      "loss": 0.4479,
+      "step": 52780
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.6503286175782554e-06,
+      "loss": 0.5128,
+      "step": 52790
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.594630722958672e-06,
+      "loss": 0.4408,
+      "step": 52800
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.538932828339089e-06,
+      "loss": 0.5033,
+      "step": 52810
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.483234933719505e-06,
+      "loss": 0.6035,
+      "step": 52820
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.427537039099922e-06,
+      "loss": 0.501,
+      "step": 52830
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.371839144480338e-06,
+      "loss": 0.5874,
+      "step": 52840
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.3161412498607545e-06,
+      "loss": 0.4779,
+      "step": 52850
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.260443355241171e-06,
+      "loss": 0.6621,
+      "step": 52860
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.204745460621588e-06,
+      "loss": 0.5001,
+      "step": 52870
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.149047566002004e-06,
+      "loss": 0.5606,
+      "step": 52880
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.093349671382421e-06,
+      "loss": 0.566,
+      "step": 52890
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.037651776762838e-06,
+      "loss": 0.553,
+      "step": 52900
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.981953882143254e-06,
+      "loss": 0.5356,
+      "step": 52910
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.926255987523671e-06,
+      "loss": 0.5597,
+      "step": 52920
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.870558092904088e-06,
+      "loss": 0.5982,
+      "step": 52930
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.814860198284505e-06,
+      "loss": 0.5086,
+      "step": 52940
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.759162303664921e-06,
+      "loss": 0.4721,
+      "step": 52950
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.703464409045338e-06,
+      "loss": 0.7,
+      "step": 52960
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.647766514425754e-06,
+      "loss": 0.4575,
+      "step": 52970
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.592068619806171e-06,
+      "loss": 0.5864,
+      "step": 52980
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.536370725186588e-06,
+      "loss": 0.5436,
+      "step": 52990
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.486242620028963e-06,
+      "loss": 0.5825,
+      "step": 53000
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 5.430544725409379e-06,
+      "loss": 0.5749,
+      "step": 53010
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.374846830789796e-06,
+      "loss": 0.5666,
+      "step": 53020
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.319148936170213e-06,
+      "loss": 0.5751,
+      "step": 53030
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.263451041550629e-06,
+      "loss": 0.7009,
+      "step": 53040
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.207753146931045e-06,
+      "loss": 0.6883,
+      "step": 53050
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.152055252311462e-06,
+      "loss": 0.5404,
+      "step": 53060
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.096357357691879e-06,
+      "loss": 0.4765,
+      "step": 53070
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 5.040659463072295e-06,
+      "loss": 0.4525,
+      "step": 53080
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.984961568452712e-06,
+      "loss": 0.5388,
+      "step": 53090
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.929263673833128e-06,
+      "loss": 0.43,
+      "step": 53100
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.873565779213545e-06,
+      "loss": 0.5274,
+      "step": 53110
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.8178678845939616e-06,
+      "loss": 0.5246,
+      "step": 53120
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.762169989974379e-06,
+      "loss": 0.5637,
+      "step": 53130
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.706472095354795e-06,
+      "loss": 0.5562,
+      "step": 53140
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.650774200735212e-06,
+      "loss": 0.582,
+      "step": 53150
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.595076306115629e-06,
+      "loss": 0.5761,
+      "step": 53160
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.539378411496045e-06,
+      "loss": 0.7622,
+      "step": 53170
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.483680516876461e-06,
+      "loss": 0.5229,
+      "step": 53180
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.4279826222568785e-06,
+      "loss": 0.5456,
+      "step": 53190
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.372284727637295e-06,
+      "loss": 0.5782,
+      "step": 53200
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.316586833017712e-06,
+      "loss": 0.6034,
+      "step": 53210
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.260888938398128e-06,
+      "loss": 0.369,
+      "step": 53220
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.205191043778545e-06,
+      "loss": 0.6394,
+      "step": 53230
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.149493149158961e-06,
+      "loss": 0.6187,
+      "step": 53240
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.0937952545393775e-06,
+      "loss": 0.5428,
+      "step": 53250
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.0380973599197945e-06,
+      "loss": 0.6027,
+      "step": 53260
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 3.982399465300212e-06,
+      "loss": 0.533,
+      "step": 53270
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 3.926701570680628e-06,
+      "loss": 0.4162,
+      "step": 53280
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.871003676061045e-06,
+      "loss": 0.5177,
+      "step": 53290
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.815305781441461e-06,
+      "loss": 0.5224,
+      "step": 53300
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.759607886821878e-06,
+      "loss": 0.6041,
+      "step": 53310
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.703909992202295e-06,
+      "loss": 0.5063,
+      "step": 53320
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.648212097582711e-06,
+      "loss": 0.6732,
+      "step": 53330
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.5925142029631277e-06,
+      "loss": 0.5758,
+      "step": 53340
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.5368163083435443e-06,
+      "loss": 0.6321,
+      "step": 53350
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.481118413723961e-06,
+      "loss": 0.4584,
+      "step": 53360
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.4254205191043776e-06,
+      "loss": 0.5924,
+      "step": 53370
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.3697226244847942e-06,
+      "loss": 0.5927,
+      "step": 53380
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.314024729865211e-06,
+      "loss": 0.594,
+      "step": 53390
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.2583268352456275e-06,
+      "loss": 0.5333,
+      "step": 53400
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.2026289406260437e-06,
+      "loss": 0.5619,
+      "step": 53410
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.1469310460064604e-06,
+      "loss": 0.742,
+      "step": 53420
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.0912331513868775e-06,
+      "loss": 0.5809,
+      "step": 53430
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.035535256767294e-06,
+      "loss": 0.4999,
+      "step": 53440
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.9798373621477107e-06,
+      "loss": 0.5197,
+      "step": 53450
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.9241394675281274e-06,
+      "loss": 0.6381,
+      "step": 53460
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.868441572908544e-06,
+      "loss": 0.706,
+      "step": 53470
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.8127436782889602e-06,
+      "loss": 0.483,
+      "step": 53480
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.757045783669377e-06,
+      "loss": 0.6196,
+      "step": 53490
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.7013478890497935e-06,
+      "loss": 0.542,
+      "step": 53500
+    }
+  ],
+  "max_steps": 53962,
+  "num_train_epochs": 2,
+  "total_flos": 1.463218548922368e+17,
+  "trial_name": null,
+  "trial_params": null
+}