diff --git "a/checkpoint-6000/trainer_state.json" "b/checkpoint-6000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-6000/trainer_state.json"
@@ -0,0 +1,36019 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.8836524300441826,
+  "eval_steps": 500,
+  "global_step": 6000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 2.0757,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 1.9875,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2e-05,
+      "loss": 1.9478,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 2.3627,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2e-05,
+      "loss": 1.9898,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4e-05,
+      "loss": 1.9836,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 1.9169,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 2.0051,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.6e-05,
+      "loss": 1.9131,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4e-05,
+      "loss": 1.9195,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 2.0058,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.8e-05,
+      "loss": 1.8733,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 1.9044,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 1.8994,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6e-05,
+      "loss": 1.5923,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 1.7187,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 1.742,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.2e-05,
+      "loss": 1.8281,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.6e-05,
+      "loss": 1.5861,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8e-05,
+      "loss": 1.6107,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.4e-05,
+      "loss": 1.7854,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 1.7075,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 1.8763,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.6e-05,
+      "loss": 1.8735,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001,
+      "loss": 1.7637,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010400000000000001,
+      "loss": 1.7543,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010800000000000001,
+      "loss": 1.7182,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011200000000000001,
+      "loss": 1.6764,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000116,
+      "loss": 1.8441,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012,
+      "loss": 1.7919,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000124,
+      "loss": 1.6958,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012800000000000002,
+      "loss": 1.6923,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000132,
+      "loss": 1.504,
+      "step": 33
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00013600000000000003,
+      "loss": 1.8161,
+      "step": 34
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00014,
+      "loss": 1.7934,
+      "step": 35
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000144,
+      "loss": 1.5316,
+      "step": 36
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000148,
+      "loss": 1.5065,
+      "step": 37
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000152,
+      "loss": 1.6563,
+      "step": 38
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00015600000000000002,
+      "loss": 1.7944,
+      "step": 39
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00016,
+      "loss": 1.73,
+      "step": 40
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000164,
+      "loss": 1.7552,
+      "step": 41
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000168,
+      "loss": 1.4764,
+      "step": 42
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000172,
+      "loss": 1.6605,
+      "step": 43
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00017600000000000002,
+      "loss": 1.6699,
+      "step": 44
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00018,
+      "loss": 1.6282,
+      "step": 45
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00018400000000000003,
+      "loss": 1.4225,
+      "step": 46
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000188,
+      "loss": 1.7011,
+      "step": 47
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000192,
+      "loss": 1.7409,
+      "step": 48
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000196,
+      "loss": 1.501,
+      "step": 49
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002,
+      "loss": 1.5354,
+      "step": 50
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00020400000000000003,
+      "loss": 1.6176,
+      "step": 51
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00020800000000000001,
+      "loss": 1.6752,
+      "step": 52
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00021200000000000003,
+      "loss": 1.6557,
+      "step": 53
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00021600000000000002,
+      "loss": 1.7425,
+      "step": 54
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00022000000000000003,
+      "loss": 1.5681,
+      "step": 55
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00022400000000000002,
+      "loss": 1.6588,
+      "step": 56
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00022799999999999999,
+      "loss": 1.9836,
+      "step": 57
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000232,
+      "loss": 1.6826,
+      "step": 58
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000236,
+      "loss": 1.4529,
+      "step": 59
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00024,
+      "loss": 1.4895,
+      "step": 60
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000244,
+      "loss": 1.6742,
+      "step": 61
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000248,
+      "loss": 1.568,
+      "step": 62
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000252,
+      "loss": 1.5808,
+      "step": 63
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00025600000000000004,
+      "loss": 1.7791,
+      "step": 64
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00026000000000000003,
+      "loss": 1.5176,
+      "step": 65
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000264,
+      "loss": 1.6139,
+      "step": 66
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000268,
+      "loss": 1.7887,
+      "step": 67
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00027200000000000005,
+      "loss": 1.856,
+      "step": 68
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000276,
+      "loss": 1.5028,
+      "step": 69
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00028,
+      "loss": 1.6957,
+      "step": 70
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000284,
+      "loss": 1.6721,
+      "step": 71
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000288,
+      "loss": 1.7098,
+      "step": 72
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000292,
+      "loss": 1.7618,
+      "step": 73
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000296,
+      "loss": 1.5841,
+      "step": 74
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00030000000000000003,
+      "loss": 1.7054,
+      "step": 75
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000304,
+      "loss": 1.6632,
+      "step": 76
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000308,
+      "loss": 1.4799,
+      "step": 77
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00031200000000000005,
+      "loss": 1.7129,
+      "step": 78
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00031600000000000004,
+      "loss": 1.5251,
+      "step": 79
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00032,
+      "loss": 1.7884,
+      "step": 80
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000324,
+      "loss": 1.5816,
+      "step": 81
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000328,
+      "loss": 1.5653,
+      "step": 82
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000332,
+      "loss": 1.7663,
+      "step": 83
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000336,
+      "loss": 1.6489,
+      "step": 84
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00034,
+      "loss": 1.8255,
+      "step": 85
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000344,
+      "loss": 1.5156,
+      "step": 86
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000348,
+      "loss": 1.3183,
+      "step": 87
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00035200000000000005,
+      "loss": 1.4092,
+      "step": 88
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00035600000000000003,
+      "loss": 1.5494,
+      "step": 89
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00036,
+      "loss": 1.6793,
+      "step": 90
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000364,
+      "loss": 1.5925,
+      "step": 91
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00036800000000000005,
+      "loss": 1.5937,
+      "step": 92
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00037200000000000004,
+      "loss": 1.7434,
+      "step": 93
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000376,
+      "loss": 1.6751,
+      "step": 94
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00038,
+      "loss": 1.7545,
+      "step": 95
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000384,
+      "loss": 1.8709,
+      "step": 96
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000388,
+      "loss": 1.7035,
+      "step": 97
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000392,
+      "loss": 1.6354,
+      "step": 98
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00039600000000000003,
+      "loss": 1.697,
+      "step": 99
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0004,
+      "loss": 1.4428,
+      "step": 100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003999999779480245,
+      "loss": 1.7104,
+      "step": 101
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039999991179210276,
+      "loss": 1.5222,
+      "step": 102
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039999980153224943,
+      "loss": 1.6017,
+      "step": 103
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039999964716848883,
+      "loss": 1.6301,
+      "step": 104
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999994487008549,
+      "loss": 1.5535,
+      "step": 105
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039999920612939155,
+      "loss": 1.7935,
+      "step": 106
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999989194541522,
+      "loss": 1.7488,
+      "step": 107
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999985886752001,
+      "loss": 1.6292,
+      "step": 108
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039999821379260817,
+      "loss": 1.5223,
+      "step": 109
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000399997794806459,
+      "loss": 1.4636,
+      "step": 110
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999973317168451,
+      "loss": 1.818,
+      "step": 111
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039999682452386857,
+      "loss": 1.7942,
+      "step": 112
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999962732276412,
+      "loss": 1.7123,
+      "step": 113
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039999567782828465,
+      "loss": 1.6798,
+      "step": 114
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999950383259301,
+      "loss": 1.7004,
+      "step": 115
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039999435472071876,
+      "loss": 1.684,
+      "step": 116
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999936270128011,
+      "loss": 1.5815,
+      "step": 117
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999928552023378,
+      "loss": 1.6033,
+      "step": 118
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039999203928949906,
+      "loss": 1.5368,
+      "step": 119
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999911792744648,
+      "loss": 1.6893,
+      "step": 120
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039999027515742454,
+      "loss": 1.53,
+      "step": 121
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999893269385778,
+      "loss": 1.5987,
+      "step": 122
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039998833461813354,
+      "loss": 1.6644,
+      "step": 123
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999872981963107,
+      "loss": 1.685,
+      "step": 124
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999862176733378,
+      "loss": 1.45,
+      "step": 125
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999850930494532,
+      "loss": 1.4791,
+      "step": 126
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999839243249047,
+      "loss": 1.7671,
+      "step": 127
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999827114999502,
+      "loss": 1.7443,
+      "step": 128
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999814545748571,
+      "loss": 1.7581,
+      "step": 129
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039998015354990253,
+      "loss": 1.5924,
+      "step": 130
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999788084253735,
+      "loss": 1.4712,
+      "step": 131
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039997741920156656,
+      "loss": 1.6466,
+      "step": 132
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039997598587878806,
+      "loss": 1.6701,
+      "step": 133
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999745084573541,
+      "loss": 1.5634,
+      "step": 134
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999729869375905,
+      "loss": 1.695,
+      "step": 135
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999714213198327,
+      "loss": 1.5627,
+      "step": 136
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000399969811604426,
+      "loss": 1.5928,
+      "step": 137
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039996815779172544,
+      "loss": 1.7722,
+      "step": 138
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999664598820956,
+      "loss": 1.627,
+      "step": 139
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000399964717875911,
+      "loss": 1.6012,
+      "step": 140
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039996293177355574,
+      "loss": 1.5681,
+      "step": 141
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999611015754237,
+      "loss": 1.6342,
+      "step": 142
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999592272819184,
+      "loss": 1.6779,
+      "step": 143
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039995730889345334,
+      "loss": 1.6892,
+      "step": 144
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999553464104514,
+      "loss": 1.3043,
+      "step": 145
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999533398333454,
+      "loss": 1.5841,
+      "step": 146
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039995128916257784,
+      "loss": 1.6602,
+      "step": 147
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999491943986009,
+      "loss": 1.7782,
+      "step": 148
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999470555418766,
+      "loss": 1.7341,
+      "step": 149
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039994487259287654,
+      "loss": 1.5795,
+      "step": 150
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039994264555208203,
+      "loss": 1.6511,
+      "step": 151
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999403744199843,
+      "loss": 1.5869,
+      "step": 152
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039993805919708416,
+      "loss": 1.5598,
+      "step": 153
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999356998838921,
+      "loss": 1.4968,
+      "step": 154
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999332964809284,
+      "loss": 1.868,
+      "step": 155
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039993084898872313,
+      "loss": 1.5642,
+      "step": 156
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039992835740781593,
+      "loss": 1.7403,
+      "step": 157
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999258217387563,
+      "loss": 1.6301,
+      "step": 158
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039992324198210335,
+      "loss": 1.4405,
+      "step": 159
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039992061813842603,
+      "loss": 1.5882,
+      "step": 160
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039991795020830295,
+      "loss": 1.6731,
+      "step": 161
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039991523819232235,
+      "loss": 1.6462,
+      "step": 162
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00039991248209108243,
+      "loss": 1.5686,
+      "step": 163
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999096819051908,
+      "loss": 1.6835,
+      "step": 164
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999068376352651,
+      "loss": 1.4027,
+      "step": 165
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999039492819324,
+      "loss": 1.2967,
+      "step": 166
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003999010168458297,
+      "loss": 1.3614,
+      "step": 167
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003998980403276038,
+      "loss": 1.5476,
+      "step": 168
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003998950197279109,
+      "loss": 1.6363,
+      "step": 169
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003998919550474172,
+      "loss": 1.5766,
+      "step": 170
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003998888462867985,
+      "loss": 1.807,
+      "step": 171
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003998856934467403,
+      "loss": 1.4256,
+      "step": 172
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003998824965279379,
+      "loss": 1.3884,
+      "step": 173
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039987925553109624,
+      "loss": 1.6485,
+      "step": 174
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039987597045693015,
+      "loss": 1.6324,
+      "step": 175
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003998726413061639,
+      "loss": 1.5961,
+      "step": 176
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039986926807953173,
+      "loss": 1.4983,
+      "step": 177
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039986585077777745,
+      "loss": 1.7092,
+      "step": 178
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039986238940165465,
+      "loss": 1.6379,
+      "step": 179
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039985888395192674,
+      "loss": 1.7078,
+      "step": 180
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039985533442936654,
+      "loss": 1.5577,
+      "step": 181
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000399851740834757,
+      "loss": 1.6209,
+      "step": 182
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003998481031688904,
+      "loss": 1.4901,
+      "step": 183
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000399844421432569,
+      "loss": 1.8356,
+      "step": 184
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003998406956266047,
+      "loss": 1.7403,
+      "step": 185
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003998369257518191,
+      "loss": 1.4272,
+      "step": 186
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003998331118090435,
+      "loss": 1.6484,
+      "step": 187
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003998292537991191,
+      "loss": 1.515,
+      "step": 188
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039982535172289644,
+      "loss": 1.7864,
+      "step": 189
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039982140558123615,
+      "loss": 1.5783,
+      "step": 190
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039981741537500843,
+      "loss": 1.6643,
+      "step": 191
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003998133811050931,
+      "loss": 1.6573,
+      "step": 192
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003998093027723799,
+      "loss": 1.4607,
+      "step": 193
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003998051803777682,
+      "loss": 1.4438,
+      "step": 194
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000399801013922167,
+      "loss": 1.7795,
+      "step": 195
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039979680340649504,
+      "loss": 1.7432,
+      "step": 196
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039979254883168095,
+      "loss": 1.5958,
+      "step": 197
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039978825019866283,
+      "loss": 1.541,
+      "step": 198
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039978390750838866,
+      "loss": 1.8556,
+      "step": 199
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003997795207618162,
+      "loss": 1.6584,
+      "step": 200
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003997750899599126,
+      "loss": 1.4267,
+      "step": 201
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003997706151036551,
+      "loss": 1.4841,
+      "step": 202
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003997660961940304,
+      "loss": 1.5104,
+      "step": 203
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039976153323203513,
+      "loss": 1.596,
+      "step": 204
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039975692621867535,
+      "loss": 1.5982,
+      "step": 205
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039975227515496713,
+      "loss": 1.7528,
+      "step": 206
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003997475800419361,
+      "loss": 1.3808,
+      "step": 207
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003997428408806175,
+      "loss": 1.8153,
+      "step": 208
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039973805767205665,
+      "loss": 1.5508,
+      "step": 209
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003997332304173081,
+      "loss": 1.6476,
+      "step": 210
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039972835911743644,
+      "loss": 1.4912,
+      "step": 211
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000399723443773516,
+      "loss": 1.4338,
+      "step": 212
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003997184843866305,
+      "loss": 1.5958,
+      "step": 213
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039971348095787373,
+      "loss": 1.606,
+      "step": 214
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039970843348834903,
+      "loss": 1.5469,
+      "step": 215
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003997033419791695,
+      "loss": 1.4804,
+      "step": 216
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003996982064314578,
+      "loss": 1.6223,
+      "step": 217
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003996930268463465,
+      "loss": 1.5996,
+      "step": 218
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039968780322497777,
+      "loss": 1.608,
+      "step": 219
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039968253556850355,
+      "loss": 1.5006,
+      "step": 220
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003996772238780855,
+      "loss": 1.5793,
+      "step": 221
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039967186815489486,
+      "loss": 1.4451,
+      "step": 222
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003996664684001127,
+      "loss": 1.7581,
+      "step": 223
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003996610246149298,
+      "loss": 1.497,
+      "step": 224
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003996555368005466,
+      "loss": 1.4664,
+      "step": 225
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039965000495817334,
+      "loss": 1.6605,
+      "step": 226
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003996444290890298,
+      "loss": 1.8601,
+      "step": 227
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003996388091943456,
+      "loss": 1.5943,
+      "step": 228
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003996331452753601,
+      "loss": 1.405,
+      "step": 229
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039962743733332217,
+      "loss": 1.8435,
+      "step": 230
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039962168536949067,
+      "loss": 1.7668,
+      "step": 231
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000399615889385134,
+      "loss": 1.5809,
+      "step": 232
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003996100493815302,
+      "loss": 1.3963,
+      "step": 233
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039960416535996717,
+      "loss": 1.7476,
+      "step": 234
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003995982373217424,
+      "loss": 1.5147,
+      "step": 235
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039959226526816317,
+      "loss": 1.8521,
+      "step": 236
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00039958624920054654,
+      "loss": 1.5722,
+      "step": 237
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039958018912021896,
+      "loss": 1.6618,
+      "step": 238
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039957408502851703,
+      "loss": 1.5729,
+      "step": 239
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003995679369267866,
+      "loss": 1.5775,
+      "step": 240
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039956174481638364,
+      "loss": 1.5679,
+      "step": 241
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003995555086986735,
+      "loss": 1.706,
+      "step": 242
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003995492285750315,
+      "loss": 1.5541,
+      "step": 243
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039954290444684236,
+      "loss": 1.4252,
+      "step": 244
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003995365363155008,
+      "loss": 1.3999,
+      "step": 245
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039953012418241104,
+      "loss": 1.581,
+      "step": 246
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003995236680489871,
+      "loss": 1.6188,
+      "step": 247
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039951716791665275,
+      "loss": 1.4069,
+      "step": 248
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003995106237868414,
+      "loss": 1.4419,
+      "step": 249
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003995040356609961,
+      "loss": 1.599,
+      "step": 250
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003994974035405696,
+      "loss": 1.5566,
+      "step": 251
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039949072742702455,
+      "loss": 1.4832,
+      "step": 252
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003994840073218331,
+      "loss": 1.5697,
+      "step": 253
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003994772432264771,
+      "loss": 1.4782,
+      "step": 254
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039947043514244836,
+      "loss": 1.5058,
+      "step": 255
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000399463583071248,
+      "loss": 1.6365,
+      "step": 256
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003994566870143872,
+      "loss": 1.7114,
+      "step": 257
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039944974697338643,
+      "loss": 1.5717,
+      "step": 258
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003994427629497764,
+      "loss": 1.6165,
+      "step": 259
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039943573494509704,
+      "loss": 1.5864,
+      "step": 260
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039942866296089816,
+      "loss": 1.4914,
+      "step": 261
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039942154699873945,
+      "loss": 1.7077,
+      "step": 262
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039941438706018995,
+      "loss": 1.5882,
+      "step": 263
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039940718314682855,
+      "loss": 1.3808,
+      "step": 264
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039939993526024397,
+      "loss": 1.6445,
+      "step": 265
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039939264340203444,
+      "loss": 1.4559,
+      "step": 266
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000399385307573808,
+      "loss": 1.6483,
+      "step": 267
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039937792777718236,
+      "loss": 1.5486,
+      "step": 268
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039937050401378483,
+      "loss": 1.5724,
+      "step": 269
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039936303628525254,
+      "loss": 1.4924,
+      "step": 270
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039935552459323224,
+      "loss": 1.601,
+      "step": 271
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039934796893938054,
+      "loss": 1.7831,
+      "step": 272
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003993403693253634,
+      "loss": 1.5741,
+      "step": 273
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003993327257528569,
+      "loss": 1.2097,
+      "step": 274
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039932503822354647,
+      "loss": 1.6514,
+      "step": 275
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039931730673912737,
+      "loss": 1.6751,
+      "step": 276
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039930953130130453,
+      "loss": 1.5137,
+      "step": 277
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039930171191179266,
+      "loss": 1.5319,
+      "step": 278
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039929384857231604,
+      "loss": 1.4837,
+      "step": 279
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003992859412846087,
+      "loss": 1.4373,
+      "step": 280
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039927799005041434,
+      "loss": 1.7161,
+      "step": 281
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003992699948714864,
+      "loss": 1.6907,
+      "step": 282
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039926195574958796,
+      "loss": 1.6317,
+      "step": 283
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039925387268649183,
+      "loss": 1.6084,
+      "step": 284
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003992457456839804,
+      "loss": 1.557,
+      "step": 285
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039923757474384587,
+      "loss": 1.5411,
+      "step": 286
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003992293598678902,
+      "loss": 1.4989,
+      "step": 287
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003992211010579248,
+      "loss": 1.4845,
+      "step": 288
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039921279831577094,
+      "loss": 1.7035,
+      "step": 289
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039920445164325953,
+      "loss": 1.5035,
+      "step": 290
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039919606104223116,
+      "loss": 1.5472,
+      "step": 291
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003991876265145362,
+      "loss": 1.7518,
+      "step": 292
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039917914806203463,
+      "loss": 1.515,
+      "step": 293
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039917062568659603,
+      "loss": 1.7659,
+      "step": 294
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003991620593900998,
+      "loss": 1.5993,
+      "step": 295
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003991534491744349,
+      "loss": 1.4352,
+      "step": 296
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003991447950415002,
+      "loss": 1.6372,
+      "step": 297
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039913609699320407,
+      "loss": 1.571,
+      "step": 298
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003991273550314645,
+      "loss": 1.5746,
+      "step": 299
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003991185691582093,
+      "loss": 1.291,
+      "step": 300
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000399109739375376,
+      "loss": 1.5993,
+      "step": 301
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039910086568491166,
+      "loss": 1.7251,
+      "step": 302
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039909194808877314,
+      "loss": 1.7101,
+      "step": 303
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000399082986588927,
+      "loss": 1.4058,
+      "step": 304
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00039907398118734937,
+      "loss": 1.6663,
+      "step": 305
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003990649318860261,
+      "loss": 1.5224,
+      "step": 306
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003990558386869528,
+      "loss": 1.4564,
+      "step": 307
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003990467015921346,
+      "loss": 1.5236,
+      "step": 308
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039903752060358653,
+      "loss": 1.7184,
+      "step": 309
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003990282957233331,
+      "loss": 1.5195,
+      "step": 310
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039901902695340855,
+      "loss": 1.491,
+      "step": 311
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003990097142958569,
+      "loss": 1.4775,
+      "step": 312
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039900035775273184,
+      "loss": 1.5668,
+      "step": 313
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003989909573260965,
+      "loss": 1.7499,
+      "step": 314
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003989815130180239,
+      "loss": 1.5615,
+      "step": 315
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003989720248305969,
+      "loss": 1.6831,
+      "step": 316
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039896249276590755,
+      "loss": 1.4251,
+      "step": 317
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039895291682605805,
+      "loss": 1.7124,
+      "step": 318
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039894329701316,
+      "loss": 1.5372,
+      "step": 319
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039893363332933475,
+      "loss": 1.5441,
+      "step": 320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039892392577671335,
+      "loss": 1.6676,
+      "step": 321
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003989141743574365,
+      "loss": 1.6574,
+      "step": 322
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003989043790736547,
+      "loss": 1.8681,
+      "step": 323
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003988945399275278,
+      "loss": 1.4901,
+      "step": 324
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039888465692122565,
+      "loss": 1.7363,
+      "step": 325
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003988747300569277,
+      "loss": 1.522,
+      "step": 326
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003988647593368229,
+      "loss": 1.6674,
+      "step": 327
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039885474476311,
+      "loss": 1.7895,
+      "step": 328
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003988446863379974,
+      "loss": 1.8207,
+      "step": 329
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003988345840637033,
+      "loss": 1.6472,
+      "step": 330
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003988244379424554,
+      "loss": 1.5229,
+      "step": 331
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003988142479764911,
+      "loss": 1.6058,
+      "step": 332
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039880401416805746,
+      "loss": 1.5832,
+      "step": 333
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039879373651941126,
+      "loss": 1.6998,
+      "step": 334
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000398783415032819,
+      "loss": 1.5201,
+      "step": 335
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003987730497105566,
+      "loss": 1.6113,
+      "step": 336
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003987626405549099,
+      "loss": 1.4334,
+      "step": 337
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003987521875681744,
+      "loss": 1.4381,
+      "step": 338
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039874169075265514,
+      "loss": 1.696,
+      "step": 339
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003987311501106669,
+      "loss": 1.4626,
+      "step": 340
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039872056564453406,
+      "loss": 1.6239,
+      "step": 341
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003987099373565906,
+      "loss": 1.4817,
+      "step": 342
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003986992652491805,
+      "loss": 1.8449,
+      "step": 343
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000398688549324657,
+      "loss": 1.8038,
+      "step": 344
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039867778958538325,
+      "loss": 1.7869,
+      "step": 345
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000398666986033732,
+      "loss": 1.7971,
+      "step": 346
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039865613867208553,
+      "loss": 1.6161,
+      "step": 347
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000398645247502836,
+      "loss": 1.5515,
+      "step": 348
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003986343125283851,
+      "loss": 1.6161,
+      "step": 349
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003986233337511442,
+      "loss": 1.8468,
+      "step": 350
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003986123111735343,
+      "loss": 1.4182,
+      "step": 351
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003986012447979863,
+      "loss": 1.4161,
+      "step": 352
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003985901346269403,
+      "loss": 1.5226,
+      "step": 353
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039857898066284636,
+      "loss": 1.6461,
+      "step": 354
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039856778290816425,
+      "loss": 1.5097,
+      "step": 355
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039855654136536323,
+      "loss": 1.6536,
+      "step": 356
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003985452560369223,
+      "loss": 1.632,
+      "step": 357
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003985339269253301,
+      "loss": 1.385,
+      "step": 358
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039852255403308493,
+      "loss": 1.531,
+      "step": 359
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003985111373626947,
+      "loss": 1.3888,
+      "step": 360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039849967691667703,
+      "loss": 1.6587,
+      "step": 361
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003984881726975592,
+      "loss": 1.6238,
+      "step": 362
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039847662470787813,
+      "loss": 1.6448,
+      "step": 363
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003984650329501803,
+      "loss": 1.6387,
+      "step": 364
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000398453397427022,
+      "loss": 1.6623,
+      "step": 365
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003984417181409691,
+      "loss": 1.5612,
+      "step": 366
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039842999509459695,
+      "loss": 1.5383,
+      "step": 367
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039841822829049094,
+      "loss": 1.6209,
+      "step": 368
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003984064177312457,
+      "loss": 1.5839,
+      "step": 369
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003983945634194659,
+      "loss": 1.6077,
+      "step": 370
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039838266535776543,
+      "loss": 1.4211,
+      "step": 371
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003983707235487682,
+      "loss": 1.4477,
+      "step": 372
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00039835873799510754,
+      "loss": 1.6463,
+      "step": 373
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003983467086994265,
+      "loss": 1.5374,
+      "step": 374
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039833463566437777,
+      "loss": 1.4111,
+      "step": 375
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039832251889262375,
+      "loss": 1.6402,
+      "step": 376
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039831035838683646,
+      "loss": 1.4274,
+      "step": 377
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039829815414969735,
+      "loss": 1.4816,
+      "step": 378
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039828590618389784,
+      "loss": 1.4251,
+      "step": 379
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003982736144921389,
+      "loss": 1.7404,
+      "step": 380
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003982612790771309,
+      "loss": 1.5814,
+      "step": 381
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003982488999415942,
+      "loss": 1.6694,
+      "step": 382
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003982364770882586,
+      "loss": 1.6656,
+      "step": 383
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003982240105198636,
+      "loss": 1.43,
+      "step": 384
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039821150023915835,
+      "loss": 1.3906,
+      "step": 385
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003981989462489015,
+      "loss": 1.51,
+      "step": 386
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039818634855186155,
+      "loss": 1.466,
+      "step": 387
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039817370715081646,
+      "loss": 1.4311,
+      "step": 388
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000398161022048554,
+      "loss": 1.6227,
+      "step": 389
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003981482932478715,
+      "loss": 1.6498,
+      "step": 390
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039813552075157585,
+      "loss": 1.6288,
+      "step": 391
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003981227045624836,
+      "loss": 1.5591,
+      "step": 392
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039810984468342103,
+      "loss": 1.8217,
+      "step": 393
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039809694111722405,
+      "loss": 1.5938,
+      "step": 394
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000398083993866738,
+      "loss": 1.6056,
+      "step": 395
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003980710029348182,
+      "loss": 1.4995,
+      "step": 396
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003980579683243292,
+      "loss": 1.8153,
+      "step": 397
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003980448900381456,
+      "loss": 1.3875,
+      "step": 398
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003980317680791512,
+      "loss": 1.4348,
+      "step": 399
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039801860245023985,
+      "loss": 1.5889,
+      "step": 400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039800539315431465,
+      "loss": 1.4764,
+      "step": 401
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003979921401942887,
+      "loss": 1.7089,
+      "step": 402
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039797884357308444,
+      "loss": 1.6527,
+      "step": 403
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000397965503293634,
+      "loss": 1.5651,
+      "step": 404
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039795211935887925,
+      "loss": 1.6245,
+      "step": 405
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003979386917717715,
+      "loss": 1.707,
+      "step": 406
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000397925220535272,
+      "loss": 1.5326,
+      "step": 407
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003979117056523512,
+      "loss": 1.2171,
+      "step": 408
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039789814712598955,
+      "loss": 1.2504,
+      "step": 409
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003978845449591769,
+      "loss": 1.4966,
+      "step": 410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039787089915491284,
+      "loss": 1.4069,
+      "step": 411
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039785720971620653,
+      "loss": 1.6532,
+      "step": 412
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039784347664607677,
+      "loss": 1.5782,
+      "step": 413
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003978296999475519,
+      "loss": 1.5442,
+      "step": 414
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039781587962367006,
+      "loss": 1.6325,
+      "step": 415
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003978020156774788,
+      "loss": 1.63,
+      "step": 416
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003977881081120355,
+      "loss": 1.6714,
+      "step": 417
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003977741569304069,
+      "loss": 1.656,
+      "step": 418
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003977601621356697,
+      "loss": 1.5981,
+      "step": 419
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039774612373090994,
+      "loss": 1.6093,
+      "step": 420
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003977320417192234,
+      "loss": 1.7316,
+      "step": 421
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039771791610371525,
+      "loss": 1.768,
+      "step": 422
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039770374688750075,
+      "loss": 1.3698,
+      "step": 423
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003976895340737044,
+      "loss": 1.5422,
+      "step": 424
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039767527766546023,
+      "loss": 1.6188,
+      "step": 425
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039766097766591235,
+      "loss": 1.8201,
+      "step": 426
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039764663407821396,
+      "loss": 1.4786,
+      "step": 427
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003976322469055282,
+      "loss": 1.5325,
+      "step": 428
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039761781615102774,
+      "loss": 1.6914,
+      "step": 429
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003976033418178948,
+      "loss": 1.7092,
+      "step": 430
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003975888239093213,
+      "loss": 1.5761,
+      "step": 431
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003975742624285087,
+      "loss": 1.4052,
+      "step": 432
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039755965737866807,
+      "loss": 1.5484,
+      "step": 433
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039754500876302017,
+      "loss": 1.3791,
+      "step": 434
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039753031658479527,
+      "loss": 1.7864,
+      "step": 435
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003975155808472333,
+      "loss": 1.4996,
+      "step": 436
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003975008015535838,
+      "loss": 1.5726,
+      "step": 437
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003974859787071058,
+      "loss": 1.5527,
+      "step": 438
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003974711123110681,
+      "loss": 1.6359,
+      "step": 439
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003974562023687491,
+      "loss": 1.4634,
+      "step": 440
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00039744124888343665,
+      "loss": 1.6526,
+      "step": 441
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003974262518584283,
+      "loss": 1.5784,
+      "step": 442
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003974112112970312,
+      "loss": 1.6034,
+      "step": 443
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003973961272025621,
+      "loss": 1.6236,
+      "step": 444
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039738099957834736,
+      "loss": 1.5095,
+      "step": 445
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003973658284277229,
+      "loss": 1.5907,
+      "step": 446
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039735061375403423,
+      "loss": 1.4992,
+      "step": 447
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003973353555606365,
+      "loss": 1.8073,
+      "step": 448
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039732005385089444,
+      "loss": 1.7525,
+      "step": 449
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039730470862818246,
+      "loss": 1.6149,
+      "step": 450
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039728931989588436,
+      "loss": 1.2932,
+      "step": 451
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003972738876573937,
+      "loss": 1.6065,
+      "step": 452
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003972584119161137,
+      "loss": 1.404,
+      "step": 453
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039724289267545684,
+      "loss": 1.7282,
+      "step": 454
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039722732993884566,
+      "loss": 1.4959,
+      "step": 455
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003972117237097119,
+      "loss": 1.5699,
+      "step": 456
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039719607399149706,
+      "loss": 1.746,
+      "step": 457
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003971803807876523,
+      "loss": 1.4402,
+      "step": 458
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039716464410163816,
+      "loss": 1.4108,
+      "step": 459
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000397148863936925,
+      "loss": 1.5417,
+      "step": 460
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003971330402969926,
+      "loss": 1.599,
+      "step": 461
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003971171731853304,
+      "loss": 1.5958,
+      "step": 462
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039710126260543736,
+      "loss": 1.6444,
+      "step": 463
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003970853085608222,
+      "loss": 1.4439,
+      "step": 464
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039706931105500294,
+      "loss": 1.754,
+      "step": 465
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003970532700915075,
+      "loss": 1.6078,
+      "step": 466
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039703718567387313,
+      "loss": 1.4583,
+      "step": 467
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003970210578056468,
+      "loss": 1.7857,
+      "step": 468
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000397004886490385,
+      "loss": 1.6099,
+      "step": 469
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039698867173165384,
+      "loss": 1.5627,
+      "step": 470
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039697241353302905,
+      "loss": 1.3948,
+      "step": 471
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039695611189809573,
+      "loss": 1.5131,
+      "step": 472
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039693976683044893,
+      "loss": 1.5071,
+      "step": 473
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039692337833369286,
+      "loss": 1.4996,
+      "step": 474
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039690694641144155,
+      "loss": 1.5247,
+      "step": 475
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003968904710673187,
+      "loss": 1.6779,
+      "step": 476
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003968739523049573,
+      "loss": 1.4853,
+      "step": 477
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039685739012800013,
+      "loss": 1.5086,
+      "step": 478
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039684078454009947,
+      "loss": 1.4432,
+      "step": 479
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039682413554491714,
+      "loss": 1.7621,
+      "step": 480
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003968074431461246,
+      "loss": 1.679,
+      "step": 481
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039679070734740293,
+      "loss": 1.6102,
+      "step": 482
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003967739281524426,
+      "loss": 1.7221,
+      "step": 483
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039675710556494375,
+      "loss": 1.6246,
+      "step": 484
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003967402395886161,
+      "loss": 1.4506,
+      "step": 485
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003967233302271791,
+      "loss": 1.4963,
+      "step": 486
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039670637748436135,
+      "loss": 1.5775,
+      "step": 487
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003966893813639014,
+      "loss": 1.6431,
+      "step": 488
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003966723418695472,
+      "loss": 1.5764,
+      "step": 489
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039665525900505627,
+      "loss": 1.4797,
+      "step": 490
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039663813277419576,
+      "loss": 1.4832,
+      "step": 491
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039662096318074234,
+      "loss": 1.5731,
+      "step": 492
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003966037502284823,
+      "loss": 1.5636,
+      "step": 493
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039658649392121126,
+      "loss": 1.4426,
+      "step": 494
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039656919426273474,
+      "loss": 1.4866,
+      "step": 495
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003965518512568676,
+      "loss": 1.5041,
+      "step": 496
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003965344649074343,
+      "loss": 1.6363,
+      "step": 497
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039651703521826893,
+      "loss": 1.7862,
+      "step": 498
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000396499562193215,
+      "loss": 1.601,
+      "step": 499
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003964820458361258,
+      "loss": 1.7703,
+      "step": 500
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039646448615086383,
+      "loss": 1.3512,
+      "step": 501
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039644688314130147,
+      "loss": 1.6212,
+      "step": 502
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039642923681132056,
+      "loss": 1.5604,
+      "step": 503
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003964115471648123,
+      "loss": 1.6653,
+      "step": 504
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039639381420567785,
+      "loss": 1.4794,
+      "step": 505
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003963760379378275,
+      "loss": 1.6225,
+      "step": 506
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003963582183651813,
+      "loss": 1.5677,
+      "step": 507
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003963403554916689,
+      "loss": 1.6847,
+      "step": 508
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00039632244932122926,
+      "loss": 1.3744,
+      "step": 509
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003963044998578112,
+      "loss": 1.7241,
+      "step": 510
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039628650710537286,
+      "loss": 1.5658,
+      "step": 511
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039626847106788204,
+      "loss": 1.4954,
+      "step": 512
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000396250391749316,
+      "loss": 1.6087,
+      "step": 513
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039623226915366154,
+      "loss": 1.7414,
+      "step": 514
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003962141032849151,
+      "loss": 1.5169,
+      "step": 515
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039619589414708267,
+      "loss": 1.3421,
+      "step": 516
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039617764174417966,
+      "loss": 1.5662,
+      "step": 517
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003961593460802311,
+      "loss": 1.5684,
+      "step": 518
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003961410071592716,
+      "loss": 1.5723,
+      "step": 519
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003961226249853451,
+      "loss": 1.5047,
+      "step": 520
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039610419956250534,
+      "loss": 1.569,
+      "step": 521
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039608573089481555,
+      "loss": 1.5559,
+      "step": 522
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003960672189863483,
+      "loss": 1.8505,
+      "step": 523
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039604866384118593,
+      "loss": 1.2918,
+      "step": 524
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039603006546342024,
+      "loss": 1.5761,
+      "step": 525
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039601142385715243,
+      "loss": 1.419,
+      "step": 526
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003959927390264934,
+      "loss": 1.4382,
+      "step": 527
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003959740109755635,
+      "loss": 1.5256,
+      "step": 528
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039595523970849267,
+      "loss": 1.7207,
+      "step": 529
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039593642522942036,
+      "loss": 1.4946,
+      "step": 530
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039591756754249545,
+      "loss": 1.4288,
+      "step": 531
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003958986666518765,
+      "loss": 1.6372,
+      "step": 532
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003958797225617316,
+      "loss": 1.586,
+      "step": 533
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003958607352762381,
+      "loss": 1.2555,
+      "step": 534
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039584170479958325,
+      "loss": 1.7151,
+      "step": 535
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003958226311359636,
+      "loss": 1.7707,
+      "step": 536
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003958035142895852,
+      "loss": 1.5752,
+      "step": 537
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003957843542646638,
+      "loss": 1.5623,
+      "step": 538
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039576515106542445,
+      "loss": 1.6971,
+      "step": 539
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003957459046961019,
+      "loss": 1.4505,
+      "step": 540
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003957266151609404,
+      "loss": 1.4605,
+      "step": 541
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003957072824641936,
+      "loss": 1.7238,
+      "step": 542
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039568790661012474,
+      "loss": 1.5428,
+      "step": 543
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039566848760300664,
+      "loss": 1.5456,
+      "step": 544
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003956490254471215,
+      "loss": 1.4786,
+      "step": 545
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039562952014676116,
+      "loss": 1.7559,
+      "step": 546
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003956099717062269,
+      "loss": 1.772,
+      "step": 547
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039559038012982955,
+      "loss": 1.4457,
+      "step": 548
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003955707454218894,
+      "loss": 1.5248,
+      "step": 549
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003955510675867364,
+      "loss": 1.5311,
+      "step": 550
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003955313466287098,
+      "loss": 1.4755,
+      "step": 551
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039551158255215845,
+      "loss": 1.4659,
+      "step": 552
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039549177536144084,
+      "loss": 1.6177,
+      "step": 553
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003954719250609247,
+      "loss": 1.7363,
+      "step": 554
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003954520316549875,
+      "loss": 1.7455,
+      "step": 555
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003954320951480161,
+      "loss": 1.313,
+      "step": 556
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039541211554440693,
+      "loss": 1.589,
+      "step": 557
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039539209284856586,
+      "loss": 1.7506,
+      "step": 558
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039537202706490827,
+      "loss": 1.5365,
+      "step": 559
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039535191819785906,
+      "loss": 1.5052,
+      "step": 560
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039533176625185267,
+      "loss": 1.4812,
+      "step": 561
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000395311571231333,
+      "loss": 1.503,
+      "step": 562
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003952913331407534,
+      "loss": 1.6071,
+      "step": 563
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003952710519845768,
+      "loss": 1.6373,
+      "step": 564
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003952507277672757,
+      "loss": 1.7429,
+      "step": 565
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003952303604933318,
+      "loss": 1.6952,
+      "step": 566
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003952099501672366,
+      "loss": 1.5997,
+      "step": 567
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039518949679349095,
+      "loss": 1.4028,
+      "step": 568
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003951690003766053,
+      "loss": 1.3166,
+      "step": 569
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039514846092109936,
+      "loss": 1.654,
+      "step": 570
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039512787843150256,
+      "loss": 1.5074,
+      "step": 571
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039510725291235383,
+      "loss": 1.624,
+      "step": 572
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039508658436820146,
+      "loss": 1.4702,
+      "step": 573
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039506587280360315,
+      "loss": 1.6578,
+      "step": 574
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039504511822312634,
+      "loss": 1.5257,
+      "step": 575
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003950243206313478,
+      "loss": 1.6093,
+      "step": 576
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00039500348003285385,
+      "loss": 1.6823,
+      "step": 577
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039498259643224007,
+      "loss": 1.5242,
+      "step": 578
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039496166983411194,
+      "loss": 1.6566,
+      "step": 579
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039494070024308404,
+      "loss": 1.7802,
+      "step": 580
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039491968766378066,
+      "loss": 1.5347,
+      "step": 581
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003948986321008355,
+      "loss": 1.5615,
+      "step": 582
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003948775335588916,
+      "loss": 1.5319,
+      "step": 583
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003948563920426017,
+      "loss": 1.6137,
+      "step": 584
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039483520755662793,
+      "loss": 1.4292,
+      "step": 585
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039481398010564187,
+      "loss": 1.3204,
+      "step": 586
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039479270969432456,
+      "loss": 1.7767,
+      "step": 587
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039477139632736663,
+      "loss": 1.4207,
+      "step": 588
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000394750040009468,
+      "loss": 1.5523,
+      "step": 589
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039472864074533825,
+      "loss": 1.7062,
+      "step": 590
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039470719853969624,
+      "loss": 1.5417,
+      "step": 591
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003946857133972705,
+      "loss": 1.682,
+      "step": 592
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039466418532279886,
+      "loss": 1.7556,
+      "step": 593
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003946426143210287,
+      "loss": 1.6422,
+      "step": 594
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039462100039671684,
+      "loss": 1.5993,
+      "step": 595
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003945993435546296,
+      "loss": 1.4091,
+      "step": 596
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003945776437995428,
+      "loss": 1.4334,
+      "step": 597
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003945559011362415,
+      "loss": 1.5028,
+      "step": 598
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039453411556952063,
+      "loss": 1.5019,
+      "step": 599
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003945122871041841,
+      "loss": 1.4702,
+      "step": 600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039449041574504565,
+      "loss": 1.3659,
+      "step": 601
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003944685014969284,
+      "loss": 1.4598,
+      "step": 602
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003944465443646647,
+      "loss": 1.4831,
+      "step": 603
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039442454435309655,
+      "loss": 1.5262,
+      "step": 604
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003944025014670755,
+      "loss": 1.4696,
+      "step": 605
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003943804157114625,
+      "loss": 1.2458,
+      "step": 606
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039435828709112765,
+      "loss": 1.5858,
+      "step": 607
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000394336115610951,
+      "loss": 1.3698,
+      "step": 608
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003943139012758217,
+      "loss": 1.6065,
+      "step": 609
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039429164409063836,
+      "loss": 1.5268,
+      "step": 610
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039426934406030926,
+      "loss": 1.5227,
+      "step": 611
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039424700118975193,
+      "loss": 1.6872,
+      "step": 612
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003942246154838935,
+      "loss": 1.5023,
+      "step": 613
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003942021869476703,
+      "loss": 1.726,
+      "step": 614
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039417971558602835,
+      "loss": 1.6059,
+      "step": 615
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003941572014039231,
+      "loss": 1.5997,
+      "step": 616
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003941346444063193,
+      "loss": 1.8913,
+      "step": 617
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003941120445981912,
+      "loss": 1.4707,
+      "step": 618
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003940894019845225,
+      "loss": 1.5097,
+      "step": 619
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039406671657030645,
+      "loss": 1.698,
+      "step": 620
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039404398836054543,
+      "loss": 1.6767,
+      "step": 621
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003940212173602517,
+      "loss": 1.5162,
+      "step": 622
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039399840357444654,
+      "loss": 1.478,
+      "step": 623
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003939755470081609,
+      "loss": 1.4401,
+      "step": 624
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003939526476664351,
+      "loss": 1.3641,
+      "step": 625
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003939297055543189,
+      "loss": 1.7617,
+      "step": 626
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003939067206768715,
+      "loss": 1.6983,
+      "step": 627
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003938836930391615,
+      "loss": 1.5936,
+      "step": 628
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039386062264626696,
+      "loss": 1.6239,
+      "step": 629
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003938375095032754,
+      "loss": 1.7184,
+      "step": 630
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039381435361528363,
+      "loss": 1.5136,
+      "step": 631
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000393791154987398,
+      "loss": 1.6812,
+      "step": 632
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039376791362473435,
+      "loss": 1.5531,
+      "step": 633
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039374462953241775,
+      "loss": 1.7029,
+      "step": 634
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003937213027155829,
+      "loss": 1.3812,
+      "step": 635
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039369793317937376,
+      "loss": 1.6323,
+      "step": 636
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003936745209289438,
+      "loss": 1.6777,
+      "step": 637
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003936510659694559,
+      "loss": 1.5442,
+      "step": 638
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003936275683060823,
+      "loss": 1.549,
+      "step": 639
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039360402794400473,
+      "loss": 1.4644,
+      "step": 640
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039358044488841423,
+      "loss": 1.6107,
+      "step": 641
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003935568191445115,
+      "loss": 1.5302,
+      "step": 642
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00039353315071750635,
+      "loss": 1.749,
+      "step": 643
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003935094396126181,
+      "loss": 1.5518,
+      "step": 644
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003934856858350757,
+      "loss": 1.672,
+      "step": 645
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039346188939011706,
+      "loss": 1.4871,
+      "step": 646
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039343805028299003,
+      "loss": 1.5179,
+      "step": 647
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003934141685189514,
+      "loss": 1.4641,
+      "step": 648
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003933902441032677,
+      "loss": 1.4858,
+      "step": 649
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003933662770412147,
+      "loss": 1.4716,
+      "step": 650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039334226733807765,
+      "loss": 1.5734,
+      "step": 651
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039331821499915106,
+      "loss": 1.4555,
+      "step": 652
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000393294120029739,
+      "loss": 1.6127,
+      "step": 653
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039326998243515495,
+      "loss": 1.498,
+      "step": 654
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039324580222072157,
+      "loss": 1.3582,
+      "step": 655
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039322157939177124,
+      "loss": 1.5743,
+      "step": 656
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003931973139536455,
+      "loss": 1.6453,
+      "step": 657
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039317300591169537,
+      "loss": 1.6836,
+      "step": 658
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003931486552712812,
+      "loss": 1.2926,
+      "step": 659
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003931242620377729,
+      "loss": 1.5307,
+      "step": 660
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039309982621654955,
+      "loss": 1.4516,
+      "step": 661
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003930753478129998,
+      "loss": 1.6667,
+      "step": 662
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003930508268325216,
+      "loss": 1.529,
+      "step": 663
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039302626328052237,
+      "loss": 1.4201,
+      "step": 664
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039300165716241864,
+      "loss": 1.3482,
+      "step": 665
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039297700848363684,
+      "loss": 1.6872,
+      "step": 666
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003929523172496123,
+      "loss": 1.4205,
+      "step": 667
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039292758346579,
+      "loss": 1.4811,
+      "step": 668
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003929028071376242,
+      "loss": 1.5839,
+      "step": 669
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003928779882705785,
+      "loss": 1.693,
+      "step": 670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003928531268701261,
+      "loss": 1.535,
+      "step": 671
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039282822294174933,
+      "loss": 1.542,
+      "step": 672
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039280327649094006,
+      "loss": 1.5841,
+      "step": 673
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003927782875231994,
+      "loss": 1.5136,
+      "step": 674
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000392753256044038,
+      "loss": 1.5974,
+      "step": 675
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003927281820589757,
+      "loss": 1.5085,
+      "step": 676
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003927030655735418,
+      "loss": 1.6081,
+      "step": 677
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003926779065932752,
+      "loss": 1.5931,
+      "step": 678
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039265270512372363,
+      "loss": 1.3269,
+      "step": 679
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039262746117044476,
+      "loss": 1.4248,
+      "step": 680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039260217473900525,
+      "loss": 1.5001,
+      "step": 681
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003925768458349813,
+      "loss": 1.5468,
+      "step": 682
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039255147446395845,
+      "loss": 1.5619,
+      "step": 683
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039252606063153156,
+      "loss": 1.6709,
+      "step": 684
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003925006043433049,
+      "loss": 1.4961,
+      "step": 685
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003924751056048921,
+      "loss": 1.4908,
+      "step": 686
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039244956442191606,
+      "loss": 1.5729,
+      "step": 687
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039242398080000917,
+      "loss": 1.3864,
+      "step": 688
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003923983547448132,
+      "loss": 1.6019,
+      "step": 689
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039237268626197904,
+      "loss": 1.5368,
+      "step": 690
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003923469753571672,
+      "loss": 1.5732,
+      "step": 691
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003923212220360474,
+      "loss": 1.7633,
+      "step": 692
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039229542630429885,
+      "loss": 1.6119,
+      "step": 693
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039226958816760993,
+      "loss": 1.7213,
+      "step": 694
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003922437076316785,
+      "loss": 1.5098,
+      "step": 695
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003922177847022116,
+      "loss": 1.609,
+      "step": 696
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003921918193849259,
+      "loss": 1.4105,
+      "step": 697
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003921658116855473,
+      "loss": 1.568,
+      "step": 698
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039213976160981087,
+      "loss": 1.6352,
+      "step": 699
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003921136691634612,
+      "loss": 1.3287,
+      "step": 700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003920875343522523,
+      "loss": 1.4389,
+      "step": 701
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039206135718194724,
+      "loss": 1.3886,
+      "step": 702
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003920351376583187,
+      "loss": 1.5077,
+      "step": 703
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003920088757871487,
+      "loss": 1.6164,
+      "step": 704
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003919825715742283,
+      "loss": 1.5735,
+      "step": 705
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039195622502535826,
+      "loss": 1.7722,
+      "step": 706
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039192983614634835,
+      "loss": 1.7663,
+      "step": 707
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00039190340494301806,
+      "loss": 1.7443,
+      "step": 708
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003918769314211958,
+      "loss": 1.6344,
+      "step": 709
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003918504155867196,
+      "loss": 1.5701,
+      "step": 710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003918238574454367,
+      "loss": 1.7835,
+      "step": 711
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003917972570032037,
+      "loss": 1.5671,
+      "step": 712
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003917706142658865,
+      "loss": 1.5934,
+      "step": 713
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039174392923936043,
+      "loss": 1.7197,
+      "step": 714
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039171720192951,
+      "loss": 1.5164,
+      "step": 715
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039169043234222906,
+      "loss": 1.7022,
+      "step": 716
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003916636204834209,
+      "loss": 1.4995,
+      "step": 717
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003916367663589981,
+      "loss": 1.3458,
+      "step": 718
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003916098699748824,
+      "loss": 1.6065,
+      "step": 719
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039158293133700504,
+      "loss": 1.4139,
+      "step": 720
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003915559504513066,
+      "loss": 1.6033,
+      "step": 721
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003915289273237369,
+      "loss": 1.487,
+      "step": 722
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039150186196025495,
+      "loss": 1.4691,
+      "step": 723
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003914747543668293,
+      "loss": 1.408,
+      "step": 724
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003914476045494376,
+      "loss": 1.5066,
+      "step": 725
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003914204125140671,
+      "loss": 1.6618,
+      "step": 726
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000391393178266714,
+      "loss": 1.5172,
+      "step": 727
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003913659018133841,
+      "loss": 1.6508,
+      "step": 728
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039133858316009235,
+      "loss": 1.4165,
+      "step": 729
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039131122231286303,
+      "loss": 1.6862,
+      "step": 730
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039128381927772983,
+      "loss": 1.5169,
+      "step": 731
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039125637406073555,
+      "loss": 1.7404,
+      "step": 732
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003912288866679325,
+      "loss": 1.6038,
+      "step": 733
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003912013571053822,
+      "loss": 1.5247,
+      "step": 734
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039117378537915535,
+      "loss": 1.5959,
+      "step": 735
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003911461714953322,
+      "loss": 1.3927,
+      "step": 736
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003911185154600021,
+      "loss": 1.6331,
+      "step": 737
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039109081727926366,
+      "loss": 1.347,
+      "step": 738
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039106307695922494,
+      "loss": 1.5514,
+      "step": 739
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003910352945060033,
+      "loss": 1.6877,
+      "step": 740
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039100746992572525,
+      "loss": 1.6682,
+      "step": 741
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003909796032245267,
+      "loss": 1.6385,
+      "step": 742
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039095169440855275,
+      "loss": 1.4982,
+      "step": 743
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003909237434839579,
+      "loss": 1.5717,
+      "step": 744
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003908957504569058,
+      "loss": 1.7718,
+      "step": 745
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003908677153335696,
+      "loss": 1.5193,
+      "step": 746
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039083963812013147,
+      "loss": 1.4813,
+      "step": 747
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003908115188227831,
+      "loss": 1.522,
+      "step": 748
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039078335744772524,
+      "loss": 1.4607,
+      "step": 749
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039075515400116805,
+      "loss": 1.5857,
+      "step": 750
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039072690848933106,
+      "loss": 1.5042,
+      "step": 751
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039069862091844283,
+      "loss": 1.3573,
+      "step": 752
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003906702912947414,
+      "loss": 1.5034,
+      "step": 753
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000390641919624474,
+      "loss": 1.4025,
+      "step": 754
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039061350591389714,
+      "loss": 1.5555,
+      "step": 755
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003905850501692766,
+      "loss": 1.4453,
+      "step": 756
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003905565523968874,
+      "loss": 1.5924,
+      "step": 757
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039052801260301397,
+      "loss": 1.5903,
+      "step": 758
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039049943079394976,
+      "loss": 1.7131,
+      "step": 759
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003904708069759977,
+      "loss": 1.3069,
+      "step": 760
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003904421411554699,
+      "loss": 1.4272,
+      "step": 761
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039041343333868773,
+      "loss": 1.6243,
+      "step": 762
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039038468353198185,
+      "loss": 1.5816,
+      "step": 763
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039035589174169215,
+      "loss": 1.6701,
+      "step": 764
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003903270579741678,
+      "loss": 1.8287,
+      "step": 765
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039029818223576714,
+      "loss": 1.6725,
+      "step": 766
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003902692645328579,
+      "loss": 1.4517,
+      "step": 767
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039024030487181706,
+      "loss": 1.5178,
+      "step": 768
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039021130325903074,
+      "loss": 1.4759,
+      "step": 769
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039018225970089435,
+      "loss": 1.6269,
+      "step": 770
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039015317420381263,
+      "loss": 1.524,
+      "step": 771
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003901240467741994,
+      "loss": 1.4536,
+      "step": 772
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039009487741847794,
+      "loss": 1.6024,
+      "step": 773
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003900656661430806,
+      "loss": 1.4983,
+      "step": 774
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00039003641295444914,
+      "loss": 1.4915,
+      "step": 775
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003900071178590343,
+      "loss": 1.429,
+      "step": 776
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00038997778086329645,
+      "loss": 1.5219,
+      "step": 777
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003899484019737048,
+      "loss": 1.5938,
+      "step": 778
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00038991898119673796,
+      "loss": 1.5891,
+      "step": 779
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003898895185388839,
+      "loss": 1.4167,
+      "step": 780
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003898600140066397,
+      "loss": 1.3507,
+      "step": 781
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038983046760651175,
+      "loss": 1.6376,
+      "step": 782
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003898008793450154,
+      "loss": 1.4203,
+      "step": 783
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003897712492286757,
+      "loss": 1.3416,
+      "step": 784
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038974157726402644,
+      "loss": 1.3448,
+      "step": 785
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038971186345761107,
+      "loss": 1.2945,
+      "step": 786
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038968210781598197,
+      "loss": 1.4655,
+      "step": 787
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038965231034570084,
+      "loss": 1.4959,
+      "step": 788
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003896224710533387,
+      "loss": 1.5594,
+      "step": 789
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038959258994547563,
+      "loss": 1.5035,
+      "step": 790
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000389562667028701,
+      "loss": 1.5386,
+      "step": 791
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003895327023096135,
+      "loss": 1.663,
+      "step": 792
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038950269579482075,
+      "loss": 1.5536,
+      "step": 793
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038947264749094,
+      "loss": 1.5191,
+      "step": 794
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038944255740459735,
+      "loss": 1.5997,
+      "step": 795
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003894124255424283,
+      "loss": 1.5596,
+      "step": 796
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003893822519110775,
+      "loss": 1.5056,
+      "step": 797
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003893520365171989,
+      "loss": 1.5979,
+      "step": 798
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003893217793674555,
+      "loss": 1.5391,
+      "step": 799
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003892914804685197,
+      "loss": 1.4412,
+      "step": 800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038926113982707284,
+      "loss": 1.4203,
+      "step": 801
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038923075744980585,
+      "loss": 1.6794,
+      "step": 802
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003892003333434185,
+      "loss": 1.4631,
+      "step": 803
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003891698675146199,
+      "loss": 1.7884,
+      "step": 804
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003891393599701284,
+      "loss": 1.6744,
+      "step": 805
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003891088107166716,
+      "loss": 1.6197,
+      "step": 806
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003890782197609861,
+      "loss": 1.604,
+      "step": 807
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003890475871098178,
+      "loss": 1.6152,
+      "step": 808
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038901691276992193,
+      "loss": 1.4371,
+      "step": 809
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038898619674806274,
+      "loss": 1.5547,
+      "step": 810
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003889554390510136,
+      "loss": 1.5128,
+      "step": 811
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038892463968555726,
+      "loss": 1.6002,
+      "step": 812
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003888937986584857,
+      "loss": 1.6069,
+      "step": 813
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003888629159765998,
+      "loss": 1.6332,
+      "step": 814
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038883199164670997,
+      "loss": 1.434,
+      "step": 815
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003888010256756355,
+      "loss": 1.3753,
+      "step": 816
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038877001807020503,
+      "loss": 1.3606,
+      "step": 817
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003887389688372564,
+      "loss": 1.4982,
+      "step": 818
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038870787798363655,
+      "loss": 1.6248,
+      "step": 819
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038867674551620163,
+      "loss": 1.5323,
+      "step": 820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038864557144181695,
+      "loss": 1.5982,
+      "step": 821
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038861435576735705,
+      "loss": 1.4366,
+      "step": 822
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038858309849970553,
+      "loss": 1.6845,
+      "step": 823
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038855179964575527,
+      "loss": 1.7337,
+      "step": 824
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038852045921240833,
+      "loss": 1.5676,
+      "step": 825
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003884890772065759,
+      "loss": 1.6004,
+      "step": 826
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038845765363517814,
+      "loss": 1.5673,
+      "step": 827
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038842618850514485,
+      "loss": 1.6041,
+      "step": 828
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003883946818234145,
+      "loss": 1.5175,
+      "step": 829
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003883631335969351,
+      "loss": 1.3728,
+      "step": 830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003883315438326635,
+      "loss": 1.6565,
+      "step": 831
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038829991253756595,
+      "loss": 1.4714,
+      "step": 832
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003882682397186178,
+      "loss": 1.5541,
+      "step": 833
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003882365253828035,
+      "loss": 1.6213,
+      "step": 834
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003882047695371166,
+      "loss": 1.7237,
+      "step": 835
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038817297218856007,
+      "loss": 1.7124,
+      "step": 836
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003881411333441458,
+      "loss": 1.3732,
+      "step": 837
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003881092530108947,
+      "loss": 1.5546,
+      "step": 838
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038807733119583727,
+      "loss": 1.4159,
+      "step": 839
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038804536790601273,
+      "loss": 1.3383,
+      "step": 840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003880133631484697,
+      "loss": 1.4674,
+      "step": 841
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038798131693026586,
+      "loss": 1.4851,
+      "step": 842
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000387949229258468,
+      "loss": 1.4487,
+      "step": 843
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038791710014015213,
+      "loss": 1.5378,
+      "step": 844
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038788492958240326,
+      "loss": 1.5824,
+      "step": 845
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003878527175923157,
+      "loss": 1.7137,
+      "step": 846
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003878204641769929,
+      "loss": 1.5453,
+      "step": 847
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00038778816934354724,
+      "loss": 1.3204,
+      "step": 848
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038775583309910044,
+      "loss": 1.6232,
+      "step": 849
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038772345545078324,
+      "loss": 1.5974,
+      "step": 850
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003876910364057356,
+      "loss": 1.5142,
+      "step": 851
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038765857597110656,
+      "loss": 1.5733,
+      "step": 852
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003876260741540543,
+      "loss": 1.5103,
+      "step": 853
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000387593530961746,
+      "loss": 1.7963,
+      "step": 854
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003875609464013582,
+      "loss": 1.5783,
+      "step": 855
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003875283204800764,
+      "loss": 1.6528,
+      "step": 856
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003874956532050953,
+      "loss": 1.4332,
+      "step": 857
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038746294458361856,
+      "loss": 1.4903,
+      "step": 858
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003874301946228592,
+      "loss": 1.4795,
+      "step": 859
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038739740333003917,
+      "loss": 1.5938,
+      "step": 860
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038736457071238963,
+      "loss": 1.4803,
+      "step": 861
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038733169677715077,
+      "loss": 1.5879,
+      "step": 862
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000387298781531572,
+      "loss": 1.5241,
+      "step": 863
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003872658249829118,
+      "loss": 1.548,
+      "step": 864
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038723282713843765,
+      "loss": 1.4887,
+      "step": 865
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003871997880054263,
+      "loss": 1.5601,
+      "step": 866
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003871667075911635,
+      "loss": 1.3767,
+      "step": 867
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038713358590294417,
+      "loss": 1.5917,
+      "step": 868
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003871004229480722,
+      "loss": 1.5008,
+      "step": 869
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038706721873386083,
+      "loss": 1.5287,
+      "step": 870
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038703397326763213,
+      "loss": 1.4225,
+      "step": 871
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003870006865567174,
+      "loss": 1.5645,
+      "step": 872
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000386967358608457,
+      "loss": 1.6548,
+      "step": 873
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038693398943020053,
+      "loss": 1.584,
+      "step": 874
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038690057902930635,
+      "loss": 1.5301,
+      "step": 875
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003868671274131423,
+      "loss": 1.3553,
+      "step": 876
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000386833634589085,
+      "loss": 1.5386,
+      "step": 877
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038680010056452033,
+      "loss": 1.395,
+      "step": 878
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003867665253468432,
+      "loss": 1.3854,
+      "step": 879
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003867329089434576,
+      "loss": 1.5661,
+      "step": 880
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003866992513617766,
+      "loss": 1.7218,
+      "step": 881
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003866655526092224,
+      "loss": 1.5151,
+      "step": 882
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003866318126932262,
+      "loss": 1.5566,
+      "step": 883
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038659803162122836,
+      "loss": 1.4699,
+      "step": 884
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003865642094006782,
+      "loss": 1.5088,
+      "step": 885
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038653034603903434,
+      "loss": 1.6382,
+      "step": 886
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003864964415437641,
+      "loss": 1.7748,
+      "step": 887
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038646249592234433,
+      "loss": 1.5273,
+      "step": 888
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038642850918226045,
+      "loss": 1.6275,
+      "step": 889
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038639448133100747,
+      "loss": 1.4923,
+      "step": 890
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038636041237608904,
+      "loss": 1.6912,
+      "step": 891
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038632630232501803,
+      "loss": 1.3766,
+      "step": 892
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038629215118531643,
+      "loss": 1.555,
+      "step": 893
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038625795896451526,
+      "loss": 1.3095,
+      "step": 894
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003862237256701546,
+      "loss": 1.4821,
+      "step": 895
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038618945130978343,
+      "loss": 1.3768,
+      "step": 896
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003861551358909601,
+      "loss": 1.7245,
+      "step": 897
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038612077942125177,
+      "loss": 1.5492,
+      "step": 898
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003860863819082347,
+      "loss": 1.1945,
+      "step": 899
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038605194335949414,
+      "loss": 1.3454,
+      "step": 900
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038601746378262466,
+      "loss": 1.5728,
+      "step": 901
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038598294318522956,
+      "loss": 1.6285,
+      "step": 902
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038594838157492134,
+      "loss": 1.4612,
+      "step": 903
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038591377895932146,
+      "loss": 1.7561,
+      "step": 904
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003858791353460606,
+      "loss": 1.6683,
+      "step": 905
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003858444507427783,
+      "loss": 1.6568,
+      "step": 906
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003858097251571232,
+      "loss": 1.4502,
+      "step": 907
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038577495859675294,
+      "loss": 1.3934,
+      "step": 908
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003857401510693343,
+      "loss": 1.4471,
+      "step": 909
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038570530258254303,
+      "loss": 1.4266,
+      "step": 910
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038567041314406384,
+      "loss": 1.3095,
+      "step": 911
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003856354827615905,
+      "loss": 1.8519,
+      "step": 912
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038560051144282607,
+      "loss": 1.5989,
+      "step": 913
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003855654991954822,
+      "loss": 1.7732,
+      "step": 914
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003855304460272799,
+      "loss": 1.7127,
+      "step": 915
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00038549535194594897,
+      "loss": 1.506,
+      "step": 916
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038546021695922854,
+      "loss": 1.7999,
+      "step": 917
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038542504107486633,
+      "loss": 1.3227,
+      "step": 918
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003853898243006195,
+      "loss": 1.5,
+      "step": 919
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000385354566644254,
+      "loss": 1.5511,
+      "step": 920
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003853192681135448,
+      "loss": 1.7017,
+      "step": 921
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003852839287162759,
+      "loss": 1.5159,
+      "step": 922
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003852485484602404,
+      "loss": 1.3174,
+      "step": 923
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038521312735324034,
+      "loss": 1.3017,
+      "step": 924
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003851776654030868,
+      "loss": 1.543,
+      "step": 925
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038514216261759977,
+      "loss": 1.3168,
+      "step": 926
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038510661900460834,
+      "loss": 1.6602,
+      "step": 927
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038507103457195057,
+      "loss": 1.496,
+      "step": 928
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003850354093274735,
+      "loss": 1.763,
+      "step": 929
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038499974327903334,
+      "loss": 1.6829,
+      "step": 930
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038496403643449497,
+      "loss": 1.6362,
+      "step": 931
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003849282888017327,
+      "loss": 1.749,
+      "step": 932
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038489250038862925,
+      "loss": 1.4259,
+      "step": 933
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000384856671203077,
+      "loss": 1.678,
+      "step": 934
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003848208012529768,
+      "loss": 1.6668,
+      "step": 935
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038478489054623874,
+      "loss": 1.5049,
+      "step": 936
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038474893909078183,
+      "loss": 1.5108,
+      "step": 937
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003847129468945341,
+      "loss": 1.5255,
+      "step": 938
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003846769139654325,
+      "loss": 1.5504,
+      "step": 939
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038464084031142304,
+      "loss": 1.7083,
+      "step": 940
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038460472594046066,
+      "loss": 1.6054,
+      "step": 941
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038456857086050926,
+      "loss": 1.7159,
+      "step": 942
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003845323750795418,
+      "loss": 1.6767,
+      "step": 943
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038449613860554014,
+      "loss": 1.5226,
+      "step": 944
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038445986144649516,
+      "loss": 1.5709,
+      "step": 945
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038442354361040664,
+      "loss": 1.4384,
+      "step": 946
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038438718510528345,
+      "loss": 1.636,
+      "step": 947
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003843507859391433,
+      "loss": 1.7387,
+      "step": 948
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038431434612001293,
+      "loss": 1.6422,
+      "step": 949
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003842778656559281,
+      "loss": 1.3665,
+      "step": 950
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003842413445549334,
+      "loss": 1.6488,
+      "step": 951
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038420478282508256,
+      "loss": 1.3922,
+      "step": 952
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000384168180474438,
+      "loss": 1.5908,
+      "step": 953
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003841315375110713,
+      "loss": 1.6259,
+      "step": 954
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003840948539430631,
+      "loss": 1.5613,
+      "step": 955
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003840581297785028,
+      "loss": 1.6568,
+      "step": 956
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003840213650254887,
+      "loss": 1.6794,
+      "step": 957
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003839845596921282,
+      "loss": 1.5772,
+      "step": 958
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003839477137865376,
+      "loss": 1.5506,
+      "step": 959
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003839108273168422,
+      "loss": 1.7379,
+      "step": 960
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003838739002911762,
+      "loss": 1.5639,
+      "step": 961
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038383693271768263,
+      "loss": 1.6277,
+      "step": 962
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003837999246045137,
+      "loss": 1.4198,
+      "step": 963
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003837628759598304,
+      "loss": 1.3602,
+      "step": 964
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003837257867918026,
+      "loss": 1.5063,
+      "step": 965
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003836886571086093,
+      "loss": 1.4172,
+      "step": 966
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003836514869184383,
+      "loss": 1.4566,
+      "step": 967
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003836142762294863,
+      "loss": 1.6195,
+      "step": 968
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038357702504995903,
+      "loss": 1.6053,
+      "step": 969
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003835397333880711,
+      "loss": 1.5727,
+      "step": 970
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038350240125204616,
+      "loss": 1.4643,
+      "step": 971
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038346502865011656,
+      "loss": 1.3032,
+      "step": 972
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038342761559052375,
+      "loss": 1.4967,
+      "step": 973
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038339016208151797,
+      "loss": 1.4863,
+      "step": 974
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003833526681313586,
+      "loss": 1.5244,
+      "step": 975
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003833151337483137,
+      "loss": 1.4189,
+      "step": 976
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038327755894066037,
+      "loss": 1.5719,
+      "step": 977
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038323994371668456,
+      "loss": 1.4684,
+      "step": 978
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003832022880846813,
+      "loss": 1.4409,
+      "step": 979
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038316459205295415,
+      "loss": 1.5544,
+      "step": 980
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003831268556298161,
+      "loss": 1.4396,
+      "step": 981
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038308907882358864,
+      "loss": 1.5764,
+      "step": 982
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003830512616426023,
+      "loss": 1.5976,
+      "step": 983
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00038301340409519657,
+      "loss": 1.4563,
+      "step": 984
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038297550618971967,
+      "loss": 1.6701,
+      "step": 985
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000382937567934529,
+      "loss": 1.7124,
+      "step": 986
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003828995893379906,
+      "loss": 1.4785,
+      "step": 987
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038286157040847953,
+      "loss": 1.3263,
+      "step": 988
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003828235111543797,
+      "loss": 1.4567,
+      "step": 989
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003827854115840839,
+      "loss": 1.6926,
+      "step": 990
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038274727170599387,
+      "loss": 1.4308,
+      "step": 991
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003827090915285202,
+      "loss": 1.4724,
+      "step": 992
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038267087106008235,
+      "loss": 1.5799,
+      "step": 993
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003826326103091088,
+      "loss": 1.7164,
+      "step": 994
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038259430928403663,
+      "loss": 1.4576,
+      "step": 995
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003825559679933121,
+      "loss": 1.5785,
+      "step": 996
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003825175864453902,
+      "loss": 1.5316,
+      "step": 997
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038247916464873477,
+      "loss": 1.4658,
+      "step": 998
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003824407026118186,
+      "loss": 1.6844,
+      "step": 999
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003824022003431234,
+      "loss": 1.5687,
+      "step": 1000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003823636578511396,
+      "loss": 1.6936,
+      "step": 1001
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038232507514436657,
+      "loss": 1.5446,
+      "step": 1002
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038228645223131253,
+      "loss": 1.4437,
+      "step": 1003
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038224778912049475,
+      "loss": 1.7204,
+      "step": 1004
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003822090858204391,
+      "loss": 1.6179,
+      "step": 1005
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038217034233968035,
+      "loss": 1.617,
+      "step": 1006
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038213155868676234,
+      "loss": 1.6622,
+      "step": 1007
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003820927348702376,
+      "loss": 1.4657,
+      "step": 1008
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003820538708986675,
+      "loss": 1.3275,
+      "step": 1009
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038201496678062227,
+      "loss": 1.5408,
+      "step": 1010
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003819760225246812,
+      "loss": 1.721,
+      "step": 1011
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003819370381394322,
+      "loss": 1.5027,
+      "step": 1012
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038189801363347194,
+      "loss": 1.55,
+      "step": 1013
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038185894901540636,
+      "loss": 1.4115,
+      "step": 1014
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038181984429384977,
+      "loss": 1.6737,
+      "step": 1015
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003817806994774256,
+      "loss": 1.6877,
+      "step": 1016
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038174151457476616,
+      "loss": 1.6392,
+      "step": 1017
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003817022895945123,
+      "loss": 1.4671,
+      "step": 1018
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038166302454531405,
+      "loss": 1.4058,
+      "step": 1019
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038162371943583007,
+      "loss": 1.5628,
+      "step": 1020
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038158437427472793,
+      "loss": 1.6404,
+      "step": 1021
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000381544989070684,
+      "loss": 1.4427,
+      "step": 1022
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003815055638323835,
+      "loss": 1.5979,
+      "step": 1023
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003814660985685205,
+      "loss": 1.4039,
+      "step": 1024
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003814265932877978,
+      "loss": 1.472,
+      "step": 1025
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003813870479989273,
+      "loss": 1.7841,
+      "step": 1026
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003813474627106292,
+      "loss": 1.6543,
+      "step": 1027
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003813078374316331,
+      "loss": 1.6881,
+      "step": 1028
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038126817217067695,
+      "loss": 1.5648,
+      "step": 1029
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003812284669365079,
+      "loss": 1.6453,
+      "step": 1030
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038118872173788167,
+      "loss": 1.8837,
+      "step": 1031
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038114893658356287,
+      "loss": 1.6244,
+      "step": 1032
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038110911148232486,
+      "loss": 1.3797,
+      "step": 1033
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038106924644294997,
+      "loss": 1.5755,
+      "step": 1034
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003810293414742291,
+      "loss": 1.4411,
+      "step": 1035
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038098939658496217,
+      "loss": 1.6374,
+      "step": 1036
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038094941178395784,
+      "loss": 1.486,
+      "step": 1037
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003809093870800334,
+      "loss": 1.5887,
+      "step": 1038
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038086932248201526,
+      "loss": 1.5433,
+      "step": 1039
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038082921799873843,
+      "loss": 1.4956,
+      "step": 1040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003807890736390466,
+      "loss": 1.7396,
+      "step": 1041
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038074888941179257,
+      "loss": 1.7505,
+      "step": 1042
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038070866532583766,
+      "loss": 1.6796,
+      "step": 1043
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003806684013900521,
+      "loss": 1.456,
+      "step": 1044
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038062809761331477,
+      "loss": 1.3975,
+      "step": 1045
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003805877540045137,
+      "loss": 1.6633,
+      "step": 1046
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038054737057254527,
+      "loss": 1.6674,
+      "step": 1047
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00038050694732631485,
+      "loss": 1.4519,
+      "step": 1048
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003804664842747365,
+      "loss": 1.6183,
+      "step": 1049
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003804259814267333,
+      "loss": 1.6019,
+      "step": 1050
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003803854387912368,
+      "loss": 1.6785,
+      "step": 1051
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003803448563771875,
+      "loss": 1.6016,
+      "step": 1052
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00038030423419353456,
+      "loss": 1.5112,
+      "step": 1053
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003802635722492361,
+      "loss": 1.5265,
+      "step": 1054
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003802228705532587,
+      "loss": 1.3899,
+      "step": 1055
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000380182129114578,
+      "loss": 1.5856,
+      "step": 1056
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00038014134794217834,
+      "loss": 1.655,
+      "step": 1057
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003801005270450527,
+      "loss": 1.5073,
+      "step": 1058
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00038005966643220285,
+      "loss": 1.3617,
+      "step": 1059
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003800187661126395,
+      "loss": 1.4509,
+      "step": 1060
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037997782609538186,
+      "loss": 1.4197,
+      "step": 1061
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003799368463894581,
+      "loss": 1.46,
+      "step": 1062
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003798958270039049,
+      "loss": 1.4538,
+      "step": 1063
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037985476794776804,
+      "loss": 1.6198,
+      "step": 1064
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037981366923010173,
+      "loss": 1.4197,
+      "step": 1065
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037977253085996906,
+      "loss": 1.6124,
+      "step": 1066
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037973135284644197,
+      "loss": 1.4463,
+      "step": 1067
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037969013519860087,
+      "loss": 1.6527,
+      "step": 1068
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003796488779255351,
+      "loss": 1.5935,
+      "step": 1069
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037960758103634275,
+      "loss": 1.5764,
+      "step": 1070
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003795662445401306,
+      "loss": 1.4519,
+      "step": 1071
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037952486844601416,
+      "loss": 1.4869,
+      "step": 1072
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003794834527631176,
+      "loss": 1.5615,
+      "step": 1073
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037944199750057406,
+      "loss": 1.6667,
+      "step": 1074
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037940050266752507,
+      "loss": 1.5409,
+      "step": 1075
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037935896827312113,
+      "loss": 1.5145,
+      "step": 1076
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003793173943265214,
+      "loss": 1.6259,
+      "step": 1077
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003792757808368938,
+      "loss": 1.502,
+      "step": 1078
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037923412781341485,
+      "loss": 1.4892,
+      "step": 1079
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003791924352652699,
+      "loss": 1.6301,
+      "step": 1080
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003791507032016529,
+      "loss": 1.7832,
+      "step": 1081
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037910893163176676,
+      "loss": 1.5723,
+      "step": 1082
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003790671205648228,
+      "loss": 1.5799,
+      "step": 1083
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003790252700100413,
+      "loss": 1.54,
+      "step": 1084
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000378983379976651,
+      "loss": 1.6316,
+      "step": 1085
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003789414504738896,
+      "loss": 1.6247,
+      "step": 1086
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037889948151100325,
+      "loss": 1.623,
+      "step": 1087
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003788574730972471,
+      "loss": 1.69,
+      "step": 1088
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037881542524188474,
+      "loss": 1.4626,
+      "step": 1089
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003787733379541885,
+      "loss": 1.487,
+      "step": 1090
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003787312112434395,
+      "loss": 1.4927,
+      "step": 1091
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037868904511892767,
+      "loss": 1.6588,
+      "step": 1092
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003786468395899513,
+      "loss": 1.6543,
+      "step": 1093
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003786045946658175,
+      "loss": 1.387,
+      "step": 1094
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003785623103558422,
+      "loss": 1.5381,
+      "step": 1095
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037851998666934997,
+      "loss": 1.4175,
+      "step": 1096
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037847762361567397,
+      "loss": 1.5178,
+      "step": 1097
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037843522120415606,
+      "loss": 1.3961,
+      "step": 1098
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037839277944414683,
+      "loss": 1.7324,
+      "step": 1099
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003783502983450055,
+      "loss": 1.7115,
+      "step": 1100
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037830777791610006,
+      "loss": 1.5103,
+      "step": 1101
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037826521816680713,
+      "loss": 1.3496,
+      "step": 1102
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003782226191065118,
+      "loss": 1.4635,
+      "step": 1103
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003781799807446082,
+      "loss": 1.5867,
+      "step": 1104
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037813730309049884,
+      "loss": 1.495,
+      "step": 1105
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037809458615359494,
+      "loss": 1.4524,
+      "step": 1106
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037805182994331656,
+      "loss": 1.6328,
+      "step": 1107
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037800903446909214,
+      "loss": 1.5589,
+      "step": 1108
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000377966199740359,
+      "loss": 1.6192,
+      "step": 1109
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003779233257665631,
+      "loss": 1.4888,
+      "step": 1110
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003778804125571589,
+      "loss": 1.7047,
+      "step": 1111
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037783746012160966,
+      "loss": 1.4217,
+      "step": 1112
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037779446846938723,
+      "loss": 1.5148,
+      "step": 1113
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037775143760997214,
+      "loss": 1.5127,
+      "step": 1114
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003777083675528535,
+      "loss": 1.5641,
+      "step": 1115
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003776652583075292,
+      "loss": 1.5974,
+      "step": 1116
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00037762210988350554,
+      "loss": 1.5172,
+      "step": 1117
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003775789222902977,
+      "loss": 1.7894,
+      "step": 1118
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003775356955374294,
+      "loss": 1.5603,
+      "step": 1119
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003774924296344329,
+      "loss": 1.5561,
+      "step": 1120
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037744912459084927,
+      "loss": 1.5745,
+      "step": 1121
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003774057804162281,
+      "loss": 1.5125,
+      "step": 1122
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037736239712012764,
+      "loss": 1.5103,
+      "step": 1123
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037731897471211486,
+      "loss": 1.3298,
+      "step": 1124
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037727551320176515,
+      "loss": 1.6215,
+      "step": 1125
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037723201259866253,
+      "loss": 1.2994,
+      "step": 1126
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003771884729124,
+      "loss": 1.3822,
+      "step": 1127
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003771448941525787,
+      "loss": 1.4761,
+      "step": 1128
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003771012763288087,
+      "loss": 1.6481,
+      "step": 1129
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037705761945070864,
+      "loss": 1.4232,
+      "step": 1130
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037701392352790556,
+      "loss": 1.3731,
+      "step": 1131
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037697018857003547,
+      "loss": 1.2586,
+      "step": 1132
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003769264145867426,
+      "loss": 1.5202,
+      "step": 1133
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003768826015876802,
+      "loss": 1.9012,
+      "step": 1134
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003768387495825098,
+      "loss": 1.6034,
+      "step": 1135
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037679485858090154,
+      "loss": 1.5443,
+      "step": 1136
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003767509285925344,
+      "loss": 1.4899,
+      "step": 1137
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037670695962709565,
+      "loss": 1.5964,
+      "step": 1138
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003766629516942814,
+      "loss": 1.3956,
+      "step": 1139
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037661890480379634,
+      "loss": 1.559,
+      "step": 1140
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037657481896535356,
+      "loss": 1.5645,
+      "step": 1141
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037653069418867494,
+      "loss": 1.4451,
+      "step": 1142
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037648653048349083,
+      "loss": 1.5922,
+      "step": 1143
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037644232785954024,
+      "loss": 1.6469,
+      "step": 1144
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037639808632657063,
+      "loss": 1.4667,
+      "step": 1145
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037635380589433825,
+      "loss": 1.6246,
+      "step": 1146
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037630948657260774,
+      "loss": 1.4398,
+      "step": 1147
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037626512837115233,
+      "loss": 1.6503,
+      "step": 1148
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000376220731299754,
+      "loss": 1.7182,
+      "step": 1149
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037617629536820313,
+      "loss": 1.6444,
+      "step": 1150
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037613182058629865,
+      "loss": 1.4967,
+      "step": 1151
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037608730696384827,
+      "loss": 1.5361,
+      "step": 1152
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000376042754510668,
+      "loss": 1.512,
+      "step": 1153
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003759981632365826,
+      "loss": 1.4509,
+      "step": 1154
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003759535331514254,
+      "loss": 1.5315,
+      "step": 1155
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037590886426503806,
+      "loss": 1.4581,
+      "step": 1156
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000375864156587271,
+      "loss": 1.5873,
+      "step": 1157
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003758194101279831,
+      "loss": 1.5601,
+      "step": 1158
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037577462489704204,
+      "loss": 1.3178,
+      "step": 1159
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037572980090432365,
+      "loss": 1.4627,
+      "step": 1160
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037568493815971256,
+      "loss": 1.4805,
+      "step": 1161
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037564003667310195,
+      "loss": 1.6381,
+      "step": 1162
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037559509645439335,
+      "loss": 1.6207,
+      "step": 1163
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037555011751349714,
+      "loss": 1.4092,
+      "step": 1164
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003755050998603319,
+      "loss": 1.5416,
+      "step": 1165
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000375460043504825,
+      "loss": 1.6204,
+      "step": 1166
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037541494845691226,
+      "loss": 1.5336,
+      "step": 1167
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000375369814726538,
+      "loss": 1.6359,
+      "step": 1168
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003753246423236551,
+      "loss": 1.7406,
+      "step": 1169
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000375279431258225,
+      "loss": 1.478,
+      "step": 1170
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037523418154021765,
+      "loss": 1.5098,
+      "step": 1171
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003751888931796114,
+      "loss": 1.6203,
+      "step": 1172
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003751435661863933,
+      "loss": 1.5906,
+      "step": 1173
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003750982005705589,
+      "loss": 1.4676,
+      "step": 1174
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003750527963421121,
+      "loss": 1.4966,
+      "step": 1175
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037500735351106554,
+      "loss": 1.4652,
+      "step": 1176
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003749618720874402,
+      "loss": 1.5825,
+      "step": 1177
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003749163520812656,
+      "loss": 1.3793,
+      "step": 1178
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037487079350257985,
+      "loss": 1.5097,
+      "step": 1179
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003748251963614296,
+      "loss": 1.4421,
+      "step": 1180
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003747795606678697,
+      "loss": 1.4073,
+      "step": 1181
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037473388643196394,
+      "loss": 1.447,
+      "step": 1182
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003746881736637843,
+      "loss": 1.6815,
+      "step": 1183
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003746424223734113,
+      "loss": 1.5083,
+      "step": 1184
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037459663257093405,
+      "loss": 1.6192,
+      "step": 1185
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003745508042664502,
+      "loss": 1.7711,
+      "step": 1186
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003745049374700656,
+      "loss": 1.5947,
+      "step": 1187
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00037445903219189493,
+      "loss": 1.4428,
+      "step": 1188
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003744130884420612,
+      "loss": 1.6238,
+      "step": 1189
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037436710623069586,
+      "loss": 1.6346,
+      "step": 1190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037432108556793896,
+      "loss": 1.597,
+      "step": 1191
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037427502646393887,
+      "loss": 1.6644,
+      "step": 1192
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037422892892885267,
+      "loss": 1.3509,
+      "step": 1193
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003741827929728456,
+      "loss": 1.5146,
+      "step": 1194
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003741366186060917,
+      "loss": 1.4539,
+      "step": 1195
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037409040583877333,
+      "loss": 1.4886,
+      "step": 1196
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037404415468108123,
+      "loss": 1.6293,
+      "step": 1197
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003739978651432147,
+      "loss": 1.7132,
+      "step": 1198
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037395153723538153,
+      "loss": 1.3669,
+      "step": 1199
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000373905170967798,
+      "loss": 1.595,
+      "step": 1200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037385876635068866,
+      "loss": 1.4819,
+      "step": 1201
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037381232339428667,
+      "loss": 1.5423,
+      "step": 1202
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037376584210883375,
+      "loss": 1.524,
+      "step": 1203
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003737193225045798,
+      "loss": 1.6132,
+      "step": 1204
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037367276459178336,
+      "loss": 1.7726,
+      "step": 1205
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003736261683807114,
+      "loss": 1.4751,
+      "step": 1206
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003735795338816393,
+      "loss": 1.544,
+      "step": 1207
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037353286110485077,
+      "loss": 1.7398,
+      "step": 1208
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037348615006063825,
+      "loss": 1.6588,
+      "step": 1209
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037343940075930236,
+      "loss": 1.7667,
+      "step": 1210
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037339261321115227,
+      "loss": 1.7481,
+      "step": 1211
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037334578742650553,
+      "loss": 1.766,
+      "step": 1212
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037329892341568814,
+      "loss": 1.4553,
+      "step": 1213
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037325202118903465,
+      "loss": 1.729,
+      "step": 1214
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037320508075688776,
+      "loss": 1.6388,
+      "step": 1215
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037315810212959884,
+      "loss": 1.5526,
+      "step": 1216
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037311108531752765,
+      "loss": 1.5791,
+      "step": 1217
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003730640303310423,
+      "loss": 1.5116,
+      "step": 1218
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037301693718051933,
+      "loss": 1.4654,
+      "step": 1219
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003729698058763437,
+      "loss": 1.4998,
+      "step": 1220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003729226364289088,
+      "loss": 1.4402,
+      "step": 1221
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037287542884861645,
+      "loss": 1.5967,
+      "step": 1222
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003728281831458768,
+      "loss": 1.421,
+      "step": 1223
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037278089933110846,
+      "loss": 1.5213,
+      "step": 1224
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037273357741473857,
+      "loss": 1.4941,
+      "step": 1225
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037268621740720243,
+      "loss": 1.3927,
+      "step": 1226
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003726388193189439,
+      "loss": 1.5854,
+      "step": 1227
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037259138316041515,
+      "loss": 1.5014,
+      "step": 1228
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003725439089420768,
+      "loss": 1.5268,
+      "step": 1229
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003724963966743979,
+      "loss": 1.6163,
+      "step": 1230
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003724488463678558,
+      "loss": 1.7086,
+      "step": 1231
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037240125803293634,
+      "loss": 1.4929,
+      "step": 1232
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003723536316801336,
+      "loss": 1.8659,
+      "step": 1233
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037230596731995024,
+      "loss": 1.6829,
+      "step": 1234
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037225826496289704,
+      "loss": 1.556,
+      "step": 1235
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003722105246194935,
+      "loss": 1.6121,
+      "step": 1236
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003721627463002672,
+      "loss": 1.388,
+      "step": 1237
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037211493001575413,
+      "loss": 1.4163,
+      "step": 1238
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037206707577649886,
+      "loss": 1.5268,
+      "step": 1239
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003720191835930542,
+      "loss": 1.6322,
+      "step": 1240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003719712534759812,
+      "loss": 1.6626,
+      "step": 1241
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003719232854358495,
+      "loss": 1.6476,
+      "step": 1242
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037187527948323693,
+      "loss": 1.6349,
+      "step": 1243
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037182723562872983,
+      "loss": 1.5598,
+      "step": 1244
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037177915388292277,
+      "loss": 1.72,
+      "step": 1245
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037173103425641876,
+      "loss": 1.4893,
+      "step": 1246
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037168287675982904,
+      "loss": 1.7027,
+      "step": 1247
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037163468140377333,
+      "loss": 1.5054,
+      "step": 1248
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037158644819887973,
+      "loss": 1.6118,
+      "step": 1249
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003715381771557846,
+      "loss": 1.4623,
+      "step": 1250
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037148986828513256,
+      "loss": 1.3675,
+      "step": 1251
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003714415215975767,
+      "loss": 1.4767,
+      "step": 1252
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003713931371037785,
+      "loss": 1.3789,
+      "step": 1253
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037134471481440766,
+      "loss": 1.516,
+      "step": 1254
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00037129625474014225,
+      "loss": 1.5761,
+      "step": 1255
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003712477568916686,
+      "loss": 1.7002,
+      "step": 1256
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00037119922127968154,
+      "loss": 1.6042,
+      "step": 1257
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003711506479148841,
+      "loss": 1.7997,
+      "step": 1258
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00037110203680798765,
+      "loss": 1.706,
+      "step": 1259
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00037105338796971195,
+      "loss": 1.6641,
+      "step": 1260
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00037100470141078495,
+      "loss": 1.8492,
+      "step": 1261
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00037095597714194306,
+      "loss": 1.6639,
+      "step": 1262
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00037090721517393094,
+      "loss": 1.4025,
+      "step": 1263
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00037085841551750154,
+      "loss": 1.5787,
+      "step": 1264
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003708095781834162,
+      "loss": 1.7868,
+      "step": 1265
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003707607031824444,
+      "loss": 1.6314,
+      "step": 1266
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00037071179052536416,
+      "loss": 1.4819,
+      "step": 1267
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00037066284022296164,
+      "loss": 1.7012,
+      "step": 1268
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00037061385228603135,
+      "loss": 1.7116,
+      "step": 1269
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003705648267253762,
+      "loss": 1.6364,
+      "step": 1270
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00037051576355180704,
+      "loss": 1.5436,
+      "step": 1271
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00037046666277614355,
+      "loss": 1.6093,
+      "step": 1272
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00037041752440921327,
+      "loss": 1.5463,
+      "step": 1273
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003703683484618522,
+      "loss": 1.3867,
+      "step": 1274
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003703191349449046,
+      "loss": 1.6792,
+      "step": 1275
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00037026988386922304,
+      "loss": 1.4267,
+      "step": 1276
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003702205952456684,
+      "loss": 1.408,
+      "step": 1277
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003701712690851097,
+      "loss": 1.3557,
+      "step": 1278
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003701219053984244,
+      "loss": 1.548,
+      "step": 1279
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00037007250419649813,
+      "loss": 1.3556,
+      "step": 1280
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00037002306549022493,
+      "loss": 1.6956,
+      "step": 1281
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00036997358929050686,
+      "loss": 1.3649,
+      "step": 1282
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003699240756082545,
+      "loss": 1.3267,
+      "step": 1283
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003698745244543866,
+      "loss": 1.5276,
+      "step": 1284
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00036982493583983004,
+      "loss": 1.5458,
+      "step": 1285
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00036977530977552026,
+      "loss": 1.8774,
+      "step": 1286
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003697256462724007,
+      "loss": 1.5676,
+      "step": 1287
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003696759453414232,
+      "loss": 1.5357,
+      "step": 1288
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003696262069935477,
+      "loss": 1.434,
+      "step": 1289
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003695764312397426,
+      "loss": 1.2697,
+      "step": 1290
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00036952661809098433,
+      "loss": 1.4826,
+      "step": 1291
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003694767675582577,
+      "loss": 1.4385,
+      "step": 1292
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003694268796525559,
+      "loss": 1.5539,
+      "step": 1293
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00036937695438487986,
+      "loss": 1.4703,
+      "step": 1294
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003693269917662394,
+      "loss": 1.674,
+      "step": 1295
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003692769918076522,
+      "loss": 1.3136,
+      "step": 1296
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003692269545201441,
+      "loss": 1.4451,
+      "step": 1297
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00036917687991474944,
+      "loss": 1.6257,
+      "step": 1298
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003691267680025106,
+      "loss": 1.6366,
+      "step": 1299
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00036907661879447825,
+      "loss": 1.6149,
+      "step": 1300
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00036902643230171136,
+      "loss": 1.592,
+      "step": 1301
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00036897620853527696,
+      "loss": 1.7654,
+      "step": 1302
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003689259475062503,
+      "loss": 1.4664,
+      "step": 1303
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003688756492257152,
+      "loss": 1.6115,
+      "step": 1304
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00036882531370476316,
+      "loss": 1.4684,
+      "step": 1305
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00036877494095449427,
+      "loss": 1.4256,
+      "step": 1306
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003687245309860167,
+      "loss": 1.7153,
+      "step": 1307
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003686740838104469,
+      "loss": 1.5289,
+      "step": 1308
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003686235994389093,
+      "loss": 1.3527,
+      "step": 1309
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003685730778825369,
+      "loss": 1.4315,
+      "step": 1310
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003685225191524706,
+      "loss": 1.6948,
+      "step": 1311
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003684719232598596,
+      "loss": 1.5163,
+      "step": 1312
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00036842129021586134,
+      "loss": 1.4228,
+      "step": 1313
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003683706200316414,
+      "loss": 1.6637,
+      "step": 1314
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003683199127183735,
+      "loss": 1.5528,
+      "step": 1315
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003682691682872396,
+      "loss": 1.4112,
+      "step": 1316
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003682183867494299,
+      "loss": 1.4752,
+      "step": 1317
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00036816756811614275,
+      "loss": 1.3504,
+      "step": 1318
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003681167123985846,
+      "loss": 1.7084,
+      "step": 1319
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00036806581960797025,
+      "loss": 1.3865,
+      "step": 1320
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003680148897555224,
+      "loss": 1.5428,
+      "step": 1321
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003679639228524722,
+      "loss": 1.6284,
+      "step": 1322
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00036791291891005885,
+      "loss": 1.7597,
+      "step": 1323
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003678618779395297,
+      "loss": 1.7886,
+      "step": 1324
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036781079995214027,
+      "loss": 1.5232,
+      "step": 1325
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036775968495915434,
+      "loss": 1.5284,
+      "step": 1326
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036770853297184375,
+      "loss": 1.6027,
+      "step": 1327
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036765734400148845,
+      "loss": 1.5724,
+      "step": 1328
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036760611805937676,
+      "loss": 1.4172,
+      "step": 1329
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003675548551568049,
+      "loss": 1.5565,
+      "step": 1330
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036750355530507735,
+      "loss": 1.3935,
+      "step": 1331
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003674522185155068,
+      "loss": 1.6847,
+      "step": 1332
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000367400844799414,
+      "loss": 1.5088,
+      "step": 1333
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003673494341681279,
+      "loss": 1.3365,
+      "step": 1334
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003672979866329854,
+      "loss": 1.6474,
+      "step": 1335
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036724650220533195,
+      "loss": 1.5749,
+      "step": 1336
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036719498089652074,
+      "loss": 1.4968,
+      "step": 1337
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036714342271791325,
+      "loss": 1.6377,
+      "step": 1338
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036709182768087913,
+      "loss": 1.5236,
+      "step": 1339
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000367040195796796,
+      "loss": 1.6009,
+      "step": 1340
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036698852707704976,
+      "loss": 1.5603,
+      "step": 1341
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003669368215330344,
+      "loss": 1.5035,
+      "step": 1342
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000366885079176152,
+      "loss": 1.4445,
+      "step": 1343
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036683330001781284,
+      "loss": 1.4949,
+      "step": 1344
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036678148406943514,
+      "loss": 1.5313,
+      "step": 1345
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003667296313424454,
+      "loss": 1.4307,
+      "step": 1346
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036667774184827816,
+      "loss": 1.3752,
+      "step": 1347
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036662581559837603,
+      "loss": 1.5763,
+      "step": 1348
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036657385260418986,
+      "loss": 1.4529,
+      "step": 1349
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036652185287717846,
+      "loss": 1.3905,
+      "step": 1350
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036646981642880885,
+      "loss": 1.5014,
+      "step": 1351
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000366417743270556,
+      "loss": 1.5917,
+      "step": 1352
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036636563341390315,
+      "loss": 1.5419,
+      "step": 1353
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036631348687034157,
+      "loss": 1.3758,
+      "step": 1354
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036626130365137054,
+      "loss": 1.596,
+      "step": 1355
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003662090837684975,
+      "loss": 1.4872,
+      "step": 1356
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000366156827233238,
+      "loss": 1.5407,
+      "step": 1357
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003661045340571156,
+      "loss": 1.4841,
+      "step": 1358
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000366052204251662,
+      "loss": 1.474,
+      "step": 1359
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036599983782841693,
+      "loss": 1.3717,
+      "step": 1360
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003659474347989283,
+      "loss": 1.4846,
+      "step": 1361
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003658949951747519,
+      "loss": 1.7095,
+      "step": 1362
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003658425189674518,
+      "loss": 1.591,
+      "step": 1363
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036579000618860004,
+      "loss": 1.6589,
+      "step": 1364
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036573745684977664,
+      "loss": 1.5751,
+      "step": 1365
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036568487096256977,
+      "loss": 1.4897,
+      "step": 1366
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036563224853857575,
+      "loss": 1.4634,
+      "step": 1367
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003655795895893988,
+      "loss": 1.5914,
+      "step": 1368
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036552689412665127,
+      "loss": 1.5381,
+      "step": 1369
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036547416216195353,
+      "loss": 1.5515,
+      "step": 1370
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003654213937069341,
+      "loss": 1.6621,
+      "step": 1371
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036536858877322936,
+      "loss": 1.5009,
+      "step": 1372
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036531574737248395,
+      "loss": 1.453,
+      "step": 1373
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003652628695163503,
+      "loss": 1.4643,
+      "step": 1374
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036520995521648915,
+      "loss": 1.7492,
+      "step": 1375
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036515700448456914,
+      "loss": 1.6593,
+      "step": 1376
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003651040173322669,
+      "loss": 1.4553,
+      "step": 1377
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003650509937712671,
+      "loss": 1.2812,
+      "step": 1378
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003649979338132626,
+      "loss": 1.5073,
+      "step": 1379
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036494483746995414,
+      "loss": 1.4487,
+      "step": 1380
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003648917047530504,
+      "loss": 1.6077,
+      "step": 1381
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003648385356742683,
+      "loss": 1.5845,
+      "step": 1382
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003647853302453327,
+      "loss": 1.6772,
+      "step": 1383
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003647320884779764,
+      "loss": 1.5248,
+      "step": 1384
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003646788103839402,
+      "loss": 1.5761,
+      "step": 1385
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003646254959749731,
+      "loss": 1.4917,
+      "step": 1386
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036457214526283184,
+      "loss": 1.6257,
+      "step": 1387
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036451875825928146,
+      "loss": 1.4558,
+      "step": 1388
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036446533497609473,
+      "loss": 1.5365,
+      "step": 1389
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003644118754250526,
+      "loss": 1.5197,
+      "step": 1390
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036435837961794386,
+      "loss": 1.4971,
+      "step": 1391
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003643048475665655,
+      "loss": 1.3175,
+      "step": 1392
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003642512792827224,
+      "loss": 1.6886,
+      "step": 1393
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036419767477822735,
+      "loss": 1.3147,
+      "step": 1394
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036414403406490124,
+      "loss": 1.6847,
+      "step": 1395
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003640903571545729,
+      "loss": 1.498,
+      "step": 1396
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003640366440590792,
+      "loss": 1.206,
+      "step": 1397
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003639828947902649,
+      "loss": 1.6808,
+      "step": 1398
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036392910935998274,
+      "loss": 1.4894,
+      "step": 1399
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003638752877800935,
+      "loss": 1.5661,
+      "step": 1400
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036382143006246585,
+      "loss": 1.451,
+      "step": 1401
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003637675362189766,
+      "loss": 1.5288,
+      "step": 1402
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036371360626151033,
+      "loss": 1.54,
+      "step": 1403
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036365964020195965,
+      "loss": 1.3027,
+      "step": 1404
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003636056380522252,
+      "loss": 1.6833,
+      "step": 1405
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003635515998242154,
+      "loss": 1.5676,
+      "step": 1406
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036349752552984686,
+      "loss": 1.6928,
+      "step": 1407
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003634434151810441,
+      "loss": 1.324,
+      "step": 1408
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003633892687897393,
+      "loss": 1.3747,
+      "step": 1409
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000363335086367873,
+      "loss": 1.6444,
+      "step": 1410
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036328086792739337,
+      "loss": 1.5322,
+      "step": 1411
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003632266134802567,
+      "loss": 1.4855,
+      "step": 1412
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003631723230384272,
+      "loss": 1.4397,
+      "step": 1413
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036311799661387697,
+      "loss": 1.4325,
+      "step": 1414
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000363063634218586,
+      "loss": 1.5131,
+      "step": 1415
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036300923586454233,
+      "loss": 1.5603,
+      "step": 1416
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036295480156374184,
+      "loss": 1.5425,
+      "step": 1417
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003629003313281884,
+      "loss": 1.4944,
+      "step": 1418
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003628458251698938,
+      "loss": 1.6622,
+      "step": 1419
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003627912831008777,
+      "loss": 1.6588,
+      "step": 1420
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036273670513316757,
+      "loss": 1.7931,
+      "step": 1421
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036268209127879916,
+      "loss": 1.4835,
+      "step": 1422
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003626274415498157,
+      "loss": 1.628,
+      "step": 1423
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003625727559582687,
+      "loss": 1.4252,
+      "step": 1424
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003625180345162173,
+      "loss": 1.5651,
+      "step": 1425
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003624632772357287,
+      "loss": 1.7534,
+      "step": 1426
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000362408484128878,
+      "loss": 1.4569,
+      "step": 1427
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003623536552077481,
+      "loss": 1.4717,
+      "step": 1428
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036229879048442986,
+      "loss": 1.6428,
+      "step": 1429
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003622438899710221,
+      "loss": 1.7727,
+      "step": 1430
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036218895367963136,
+      "loss": 1.478,
+      "step": 1431
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003621339816223723,
+      "loss": 1.3538,
+      "step": 1432
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036207897381136726,
+      "loss": 1.6538,
+      "step": 1433
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003620239302587466,
+      "loss": 1.4319,
+      "step": 1434
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003619688509766485,
+      "loss": 1.5597,
+      "step": 1435
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000361913735977219,
+      "loss": 1.6915,
+      "step": 1436
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036185858527261206,
+      "loss": 1.5267,
+      "step": 1437
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003618033988749895,
+      "loss": 1.3813,
+      "step": 1438
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036174817679652106,
+      "loss": 1.6244,
+      "step": 1439
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003616929190493842,
+      "loss": 1.441,
+      "step": 1440
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003616376256457644,
+      "loss": 1.6571,
+      "step": 1441
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036158229659785494,
+      "loss": 1.253,
+      "step": 1442
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036152693191785705,
+      "loss": 1.6449,
+      "step": 1443
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003614715316179796,
+      "loss": 1.4165,
+      "step": 1444
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003614160957104395,
+      "loss": 1.6168,
+      "step": 1445
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036136062420746146,
+      "loss": 1.6726,
+      "step": 1446
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036130511712127814,
+      "loss": 1.6476,
+      "step": 1447
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003612495744641298,
+      "loss": 1.6158,
+      "step": 1448
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003611939962482648,
+      "loss": 1.2528,
+      "step": 1449
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003611383824859392,
+      "loss": 1.7357,
+      "step": 1450
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003610827331894168,
+      "loss": 1.7235,
+      "step": 1451
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036102704837096965,
+      "loss": 1.6628,
+      "step": 1452
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003609713280428771,
+      "loss": 1.5961,
+      "step": 1453
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036091557221742677,
+      "loss": 1.497,
+      "step": 1454
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003608597809069138,
+      "loss": 1.6637,
+      "step": 1455
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003608039541236412,
+      "loss": 1.5318,
+      "step": 1456
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003607480918799201,
+      "loss": 1.4441,
+      "step": 1457
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003606921941880691,
+      "loss": 1.5102,
+      "step": 1458
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003606362610604148,
+      "loss": 1.5917,
+      "step": 1459
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003605802925092914,
+      "loss": 1.5366,
+      "step": 1460
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00036052428854704133,
+      "loss": 1.1791,
+      "step": 1461
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00036046824918601434,
+      "loss": 1.4302,
+      "step": 1462
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003604121744385683,
+      "loss": 1.4127,
+      "step": 1463
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003603560643170688,
+      "loss": 1.4103,
+      "step": 1464
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00036029991883388936,
+      "loss": 1.5212,
+      "step": 1465
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00036024373800141096,
+      "loss": 1.4421,
+      "step": 1466
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003601875218320227,
+      "loss": 1.5084,
+      "step": 1467
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00036013127033812127,
+      "loss": 1.5023,
+      "step": 1468
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003600749835321113,
+      "loss": 1.328,
+      "step": 1469
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00036001866142640516,
+      "loss": 1.6686,
+      "step": 1470
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035996230403342297,
+      "loss": 1.7626,
+      "step": 1471
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035990591136559263,
+      "loss": 1.7084,
+      "step": 1472
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035984948343534983,
+      "loss": 1.6233,
+      "step": 1473
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003597930202551381,
+      "loss": 1.5975,
+      "step": 1474
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003597365218374086,
+      "loss": 1.5874,
+      "step": 1475
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035967998819462044,
+      "loss": 1.4551,
+      "step": 1476
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035962341933924025,
+      "loss": 1.7407,
+      "step": 1477
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003595668152837428,
+      "loss": 1.3905,
+      "step": 1478
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035951017604061026,
+      "loss": 1.214,
+      "step": 1479
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035945350162233273,
+      "loss": 1.9179,
+      "step": 1480
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035939679204140805,
+      "loss": 1.6771,
+      "step": 1481
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035934004731034177,
+      "loss": 1.3751,
+      "step": 1482
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003592832674416473,
+      "loss": 1.5973,
+      "step": 1483
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003592264524478456,
+      "loss": 1.6484,
+      "step": 1484
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003591696023414656,
+      "loss": 1.6412,
+      "step": 1485
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035911271713504387,
+      "loss": 1.5407,
+      "step": 1486
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003590557968411247,
+      "loss": 1.7071,
+      "step": 1487
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035899884147226006,
+      "loss": 1.1601,
+      "step": 1488
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035894185104100984,
+      "loss": 1.7678,
+      "step": 1489
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035888482555994155,
+      "loss": 1.5909,
+      "step": 1490
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003588277650416304,
+      "loss": 1.6193,
+      "step": 1491
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003587706694986594,
+      "loss": 1.6676,
+      "step": 1492
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003587135389436192,
+      "loss": 1.3086,
+      "step": 1493
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003586563733891082,
+      "loss": 1.7121,
+      "step": 1494
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003585991728477326,
+      "loss": 1.5212,
+      "step": 1495
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003585419373321062,
+      "loss": 1.4815,
+      "step": 1496
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035848466685485053,
+      "loss": 1.6279,
+      "step": 1497
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035842736142859493,
+      "loss": 1.5513,
+      "step": 1498
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003583700210659764,
+      "loss": 1.6999,
+      "step": 1499
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035831264577963957,
+      "loss": 1.4316,
+      "step": 1500
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003582552355822368,
+      "loss": 1.4251,
+      "step": 1501
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035819779048642825,
+      "loss": 1.6289,
+      "step": 1502
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003581403105048816,
+      "loss": 1.2918,
+      "step": 1503
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035808279565027236,
+      "loss": 1.3252,
+      "step": 1504
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003580252459352838,
+      "loss": 1.4776,
+      "step": 1505
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003579676613726066,
+      "loss": 1.4631,
+      "step": 1506
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003579100419749394,
+      "loss": 1.3759,
+      "step": 1507
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003578523877549884,
+      "loss": 1.5311,
+      "step": 1508
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035779469872546744,
+      "loss": 1.3931,
+      "step": 1509
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003577369748990982,
+      "loss": 1.4935,
+      "step": 1510
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003576792162886098,
+      "loss": 1.5862,
+      "step": 1511
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003576214229067392,
+      "loss": 1.5305,
+      "step": 1512
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035756359476623104,
+      "loss": 1.4863,
+      "step": 1513
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003575057318798375,
+      "loss": 1.5922,
+      "step": 1514
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003574478342603185,
+      "loss": 1.7416,
+      "step": 1515
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003573899019204416,
+      "loss": 1.4416,
+      "step": 1516
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003573319348729821,
+      "loss": 1.4921,
+      "step": 1517
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003572739331307227,
+      "loss": 1.6682,
+      "step": 1518
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003572158967064542,
+      "loss": 1.4448,
+      "step": 1519
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035715782561297456,
+      "loss": 1.6473,
+      "step": 1520
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003570997198630897,
+      "loss": 1.6352,
+      "step": 1521
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000357041579469613,
+      "loss": 1.4473,
+      "step": 1522
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003569834044453657,
+      "loss": 1.4846,
+      "step": 1523
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035692519480317637,
+      "loss": 1.7282,
+      "step": 1524
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035686695055588163,
+      "loss": 1.4984,
+      "step": 1525
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035680867171632527,
+      "loss": 1.701,
+      "step": 1526
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00035675035829735893,
+      "loss": 1.5605,
+      "step": 1527
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000356692010311842,
+      "loss": 1.5343,
+      "step": 1528
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003566336277726413,
+      "loss": 1.3996,
+      "step": 1529
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003565752106926313,
+      "loss": 1.585,
+      "step": 1530
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003565167590846942,
+      "loss": 1.6988,
+      "step": 1531
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035645827296171967,
+      "loss": 1.7247,
+      "step": 1532
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003563997523366051,
+      "loss": 1.5498,
+      "step": 1533
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003563411972222554,
+      "loss": 1.3093,
+      "step": 1534
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003562826076315831,
+      "loss": 1.6277,
+      "step": 1535
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003562239835775084,
+      "loss": 1.4609,
+      "step": 1536
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035616532507295914,
+      "loss": 1.5788,
+      "step": 1537
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035610663213087057,
+      "loss": 1.646,
+      "step": 1538
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035604790476418573,
+      "loss": 1.6064,
+      "step": 1539
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035598914298585506,
+      "loss": 1.6924,
+      "step": 1540
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003559303468088368,
+      "loss": 1.6157,
+      "step": 1541
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003558715162460966,
+      "loss": 1.4356,
+      "step": 1542
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035581265131060777,
+      "loss": 1.4179,
+      "step": 1543
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035575375201535124,
+      "loss": 1.2381,
+      "step": 1544
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035569481837331534,
+      "loss": 1.5064,
+      "step": 1545
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035563585039749627,
+      "loss": 1.6213,
+      "step": 1546
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035557684810089753,
+      "loss": 1.8705,
+      "step": 1547
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003555178114965303,
+      "loss": 1.3392,
+      "step": 1548
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035545874059741343,
+      "loss": 1.5244,
+      "step": 1549
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035539963541657304,
+      "loss": 1.6012,
+      "step": 1550
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003553404959670431,
+      "loss": 1.6012,
+      "step": 1551
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035528132226186497,
+      "loss": 1.416,
+      "step": 1552
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035522211431408764,
+      "loss": 1.442,
+      "step": 1553
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003551628721367677,
+      "loss": 1.5589,
+      "step": 1554
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003551035957429692,
+      "loss": 1.3046,
+      "step": 1555
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035504428514576366,
+      "loss": 1.5733,
+      "step": 1556
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003549849403582303,
+      "loss": 1.411,
+      "step": 1557
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035492556139345583,
+      "loss": 1.646,
+      "step": 1558
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003548661482645345,
+      "loss": 1.5568,
+      "step": 1559
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000354806700984568,
+      "loss": 1.6618,
+      "step": 1560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003547472195666657,
+      "loss": 1.5664,
+      "step": 1561
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003546877040239444,
+      "loss": 1.5612,
+      "step": 1562
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003546281543695285,
+      "loss": 1.6628,
+      "step": 1563
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035456857061654987,
+      "loss": 1.4588,
+      "step": 1564
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035450895277814784,
+      "loss": 1.4426,
+      "step": 1565
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003544493008674693,
+      "loss": 1.4924,
+      "step": 1566
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035438961489766874,
+      "loss": 1.4002,
+      "step": 1567
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003543298948819081,
+      "loss": 1.3243,
+      "step": 1568
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003542701408333568,
+      "loss": 1.4432,
+      "step": 1569
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003542103527651918,
+      "loss": 1.5187,
+      "step": 1570
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003541505306905976,
+      "loss": 1.3144,
+      "step": 1571
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035409067462276606,
+      "loss": 1.5468,
+      "step": 1572
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035403078457489667,
+      "loss": 1.4624,
+      "step": 1573
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035397086056019624,
+      "loss": 1.7317,
+      "step": 1574
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035391090259187947,
+      "loss": 1.3581,
+      "step": 1575
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003538509106831681,
+      "loss": 1.5833,
+      "step": 1576
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003537908848472915,
+      "loss": 1.6044,
+      "step": 1577
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003537308250974866,
+      "loss": 1.4853,
+      "step": 1578
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003536707314469978,
+      "loss": 1.6634,
+      "step": 1579
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003536106039090769,
+      "loss": 1.5139,
+      "step": 1580
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035355044249698323,
+      "loss": 1.5716,
+      "step": 1581
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035349024722398356,
+      "loss": 1.4208,
+      "step": 1582
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003534300181033521,
+      "loss": 1.7618,
+      "step": 1583
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003533697551483706,
+      "loss": 1.681,
+      "step": 1584
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035330945837232827,
+      "loss": 1.6482,
+      "step": 1585
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035324912778852166,
+      "loss": 1.493,
+      "step": 1586
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003531887634102549,
+      "loss": 1.4973,
+      "step": 1587
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003531283652508395,
+      "loss": 1.6617,
+      "step": 1588
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003530679333235945,
+      "loss": 1.4895,
+      "step": 1589
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035300746764184627,
+      "loss": 1.5347,
+      "step": 1590
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035294696821892873,
+      "loss": 1.5876,
+      "step": 1591
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003528864350681832,
+      "loss": 1.183,
+      "step": 1592
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00035282586820295844,
+      "loss": 1.5403,
+      "step": 1593
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003527652676366106,
+      "loss": 1.3928,
+      "step": 1594
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003527046333825033,
+      "loss": 1.2938,
+      "step": 1595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003526439654540076,
+      "loss": 1.443,
+      "step": 1596
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00035258326386450203,
+      "loss": 1.5921,
+      "step": 1597
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003525225286273725,
+      "loss": 1.522,
+      "step": 1598
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00035246175975601216,
+      "loss": 1.623,
+      "step": 1599
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000352400957263822,
+      "loss": 1.6044,
+      "step": 1600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00035234012116420994,
+      "loss": 1.5946,
+      "step": 1601
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003522792514705917,
+      "loss": 1.5455,
+      "step": 1602
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003522183481963901,
+      "loss": 1.4014,
+      "step": 1603
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003521574113550357,
+      "loss": 1.5627,
+      "step": 1604
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003520964409599661,
+      "loss": 1.5244,
+      "step": 1605
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00035203543702462666,
+      "loss": 1.4357,
+      "step": 1606
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00035197439956246977,
+      "loss": 1.4674,
+      "step": 1607
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00035191332858695547,
+      "loss": 1.2138,
+      "step": 1608
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003518522241115512,
+      "loss": 1.6656,
+      "step": 1609
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003517910861497316,
+      "loss": 1.3703,
+      "step": 1610
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003517299147149788,
+      "loss": 1.5371,
+      "step": 1611
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003516687098207824,
+      "loss": 1.4421,
+      "step": 1612
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003516074714806392,
+      "loss": 1.4229,
+      "step": 1613
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003515461997080534,
+      "loss": 1.6234,
+      "step": 1614
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003514848945165368,
+      "loss": 1.5507,
+      "step": 1615
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003514235559196084,
+      "loss": 1.4377,
+      "step": 1616
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003513621839307944,
+      "loss": 1.5035,
+      "step": 1617
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003513007785636287,
+      "loss": 1.3674,
+      "step": 1618
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00035123933983165223,
+      "loss": 1.5867,
+      "step": 1619
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00035117786774841364,
+      "loss": 1.5246,
+      "step": 1620
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00035111636232746865,
+      "loss": 1.6199,
+      "step": 1621
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003510548235823804,
+      "loss": 1.5454,
+      "step": 1622
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00035099325152671945,
+      "loss": 1.6928,
+      "step": 1623
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003509316461740636,
+      "loss": 1.7949,
+      "step": 1624
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00035087000753799804,
+      "loss": 1.382,
+      "step": 1625
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003508083356321154,
+      "loss": 1.5548,
+      "step": 1626
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003507466304700154,
+      "loss": 1.6625,
+      "step": 1627
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003506848920653054,
+      "loss": 1.618,
+      "step": 1628
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00035062312043159985,
+      "loss": 1.4813,
+      "step": 1629
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00035056131558252067,
+      "loss": 1.4544,
+      "step": 1630
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000350499477531697,
+      "loss": 1.2933,
+      "step": 1631
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003504376062927654,
+      "loss": 1.4137,
+      "step": 1632
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00035037570187936967,
+      "loss": 1.4273,
+      "step": 1633
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00035031376430516093,
+      "loss": 1.6605,
+      "step": 1634
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00035025179358379773,
+      "loss": 1.6065,
+      "step": 1635
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003501897897289457,
+      "loss": 1.5249,
+      "step": 1636
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000350127752754278,
+      "loss": 1.2874,
+      "step": 1637
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00035006568267347503,
+      "loss": 1.5029,
+      "step": 1638
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003500035795002245,
+      "loss": 1.4532,
+      "step": 1639
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003499414432482212,
+      "loss": 1.3737,
+      "step": 1640
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003498792739311676,
+      "loss": 1.5215,
+      "step": 1641
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00034981707156277317,
+      "loss": 1.5084,
+      "step": 1642
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00034975483615675476,
+      "loss": 1.4566,
+      "step": 1643
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003496925677268366,
+      "loss": 1.4416,
+      "step": 1644
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00034963026628675,
+      "loss": 1.5368,
+      "step": 1645
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00034956793185023365,
+      "loss": 1.4982,
+      "step": 1646
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00034950556443103366,
+      "loss": 1.409,
+      "step": 1647
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00034944316404290313,
+      "loss": 1.5985,
+      "step": 1648
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00034938073069960265,
+      "loss": 1.4939,
+      "step": 1649
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00034931826441489996,
+      "loss": 1.5296,
+      "step": 1650
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003492557652025702,
+      "loss": 1.5045,
+      "step": 1651
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003491932330763955,
+      "loss": 1.543,
+      "step": 1652
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00034913066805016566,
+      "loss": 1.4482,
+      "step": 1653
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003490680701376774,
+      "loss": 1.5277,
+      "step": 1654
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00034900543935273475,
+      "loss": 1.5821,
+      "step": 1655
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00034894277570914913,
+      "loss": 1.5155,
+      "step": 1656
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000348880079220739,
+      "loss": 1.4621,
+      "step": 1657
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003488173499013302,
+      "loss": 1.464,
+      "step": 1658
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003487545877647559,
+      "loss": 1.4326,
+      "step": 1659
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00034869179282485626,
+      "loss": 1.6773,
+      "step": 1660
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00034862896509547887,
+      "loss": 1.5418,
+      "step": 1661
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00034856610459047845,
+      "loss": 1.5901,
+      "step": 1662
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00034850321132371705,
+      "loss": 1.5318,
+      "step": 1663
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034844028530906375,
+      "loss": 1.6494,
+      "step": 1664
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003483773265603952,
+      "loss": 1.4763,
+      "step": 1665
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034831433509159473,
+      "loss": 1.3801,
+      "step": 1666
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034825131091655346,
+      "loss": 1.7758,
+      "step": 1667
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003481882540491694,
+      "loss": 1.399,
+      "step": 1668
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034812516450334776,
+      "loss": 1.3298,
+      "step": 1669
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003480620422930012,
+      "loss": 1.5803,
+      "step": 1670
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003479988874320492,
+      "loss": 1.6093,
+      "step": 1671
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003479356999344188,
+      "loss": 1.5617,
+      "step": 1672
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034787247981404407,
+      "loss": 1.7235,
+      "step": 1673
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003478092270848663,
+      "loss": 1.2456,
+      "step": 1674
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034774594176083387,
+      "loss": 1.4445,
+      "step": 1675
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003476826238559026,
+      "loss": 1.3905,
+      "step": 1676
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034761927338403516,
+      "loss": 1.3644,
+      "step": 1677
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003475558903592018,
+      "loss": 1.5775,
+      "step": 1678
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034749247479537947,
+      "loss": 1.3307,
+      "step": 1679
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003474290267065528,
+      "loss": 1.5949,
+      "step": 1680
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034736554610671316,
+      "loss": 1.5934,
+      "step": 1681
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003473020330098594,
+      "loss": 1.4341,
+      "step": 1682
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034723848742999734,
+      "loss": 1.5465,
+      "step": 1683
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034717490938114006,
+      "loss": 1.6415,
+      "step": 1684
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003471112988773078,
+      "loss": 1.4616,
+      "step": 1685
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003470476559325279,
+      "loss": 1.3289,
+      "step": 1686
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034698398056083493,
+      "loss": 1.3665,
+      "step": 1687
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034692027277627056,
+      "loss": 1.4824,
+      "step": 1688
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003468565325928835,
+      "loss": 1.534,
+      "step": 1689
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034679276002472985,
+      "loss": 1.4131,
+      "step": 1690
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034672895508587267,
+      "loss": 1.6499,
+      "step": 1691
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034666511779038223,
+      "loss": 1.7237,
+      "step": 1692
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003466012481523359,
+      "loss": 1.691,
+      "step": 1693
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034653734618581824,
+      "loss": 1.3484,
+      "step": 1694
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003464734119049208,
+      "loss": 1.5892,
+      "step": 1695
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003464094453237425,
+      "loss": 1.5904,
+      "step": 1696
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000346345446456389,
+      "loss": 1.6487,
+      "step": 1697
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003462814153169736,
+      "loss": 1.5618,
+      "step": 1698
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034621735191961623,
+      "loss": 1.7691,
+      "step": 1699
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003461532562784442,
+      "loss": 1.6181,
+      "step": 1700
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034608912840759185,
+      "loss": 1.3831,
+      "step": 1701
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003460249683212006,
+      "loss": 1.5588,
+      "step": 1702
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034596077603341925,
+      "loss": 1.6367,
+      "step": 1703
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003458965515584031,
+      "loss": 1.4548,
+      "step": 1704
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003458322949103152,
+      "loss": 1.6076,
+      "step": 1705
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003457680061033253,
+      "loss": 1.5576,
+      "step": 1706
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034570368515161035,
+      "loss": 1.3367,
+      "step": 1707
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003456393320693544,
+      "loss": 1.5485,
+      "step": 1708
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003455749468707486,
+      "loss": 1.638,
+      "step": 1709
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003455105295699911,
+      "loss": 1.4263,
+      "step": 1710
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034544608018128724,
+      "loss": 1.5742,
+      "step": 1711
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003453815987188494,
+      "loss": 1.7887,
+      "step": 1712
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000345317085196897,
+      "loss": 1.6948,
+      "step": 1713
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003452525396296565,
+      "loss": 1.511,
+      "step": 1714
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034518796203136153,
+      "loss": 1.6722,
+      "step": 1715
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034512335241625267,
+      "loss": 1.7141,
+      "step": 1716
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034505871079857763,
+      "loss": 1.6575,
+      "step": 1717
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034499403719259123,
+      "loss": 1.479,
+      "step": 1718
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003449293316125552,
+      "loss": 1.5659,
+      "step": 1719
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034486459407273844,
+      "loss": 1.4676,
+      "step": 1720
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034479982458741687,
+      "loss": 1.4995,
+      "step": 1721
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003447350231708734,
+      "loss": 1.3403,
+      "step": 1722
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000344670189837398,
+      "loss": 1.646,
+      "step": 1723
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003446053246012878,
+      "loss": 1.5029,
+      "step": 1724
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003445404274768468,
+      "loss": 1.6225,
+      "step": 1725
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003444754984783861,
+      "loss": 1.3906,
+      "step": 1726
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034441053762022375,
+      "loss": 1.5828,
+      "step": 1727
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003443455449166851,
+      "loss": 1.6213,
+      "step": 1728
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034428052038210213,
+      "loss": 1.5415,
+      "step": 1729
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00034421546403081414,
+      "loss": 1.5807,
+      "step": 1730
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003441503758771673,
+      "loss": 1.4129,
+      "step": 1731
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034408525593551486,
+      "loss": 1.506,
+      "step": 1732
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000344020104220217,
+      "loss": 1.5018,
+      "step": 1733
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000343954920745641,
+      "loss": 1.7144,
+      "step": 1734
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003438897055261612,
+      "loss": 1.8894,
+      "step": 1735
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034382445857615866,
+      "loss": 1.4131,
+      "step": 1736
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034375917991002174,
+      "loss": 1.431,
+      "step": 1737
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003436938695421456,
+      "loss": 1.7208,
+      "step": 1738
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003436285274869326,
+      "loss": 1.5374,
+      "step": 1739
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034356315375879185,
+      "loss": 1.7084,
+      "step": 1740
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003434977483721395,
+      "loss": 1.2436,
+      "step": 1741
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034343231134139886,
+      "loss": 1.6205,
+      "step": 1742
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034336684268099996,
+      "loss": 1.6436,
+      "step": 1743
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034330134240538,
+      "loss": 1.4344,
+      "step": 1744
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003432358105289831,
+      "loss": 1.665,
+      "step": 1745
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034317024706626036,
+      "loss": 1.6432,
+      "step": 1746
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003431046520316697,
+      "loss": 1.3568,
+      "step": 1747
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034303902543967625,
+      "loss": 1.401,
+      "step": 1748
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003429733673047519,
+      "loss": 1.7272,
+      "step": 1749
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034290767764137556,
+      "loss": 1.674,
+      "step": 1750
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003428419564640331,
+      "loss": 1.542,
+      "step": 1751
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003427762037872174,
+      "loss": 1.3389,
+      "step": 1752
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034271041962542813,
+      "loss": 1.5301,
+      "step": 1753
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003426446039931721,
+      "loss": 1.4746,
+      "step": 1754
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003425787569049629,
+      "loss": 1.4486,
+      "step": 1755
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003425128783753211,
+      "loss": 1.52,
+      "step": 1756
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034244696841877426,
+      "loss": 1.455,
+      "step": 1757
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003423810270498567,
+      "loss": 1.5948,
+      "step": 1758
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034231505428311006,
+      "loss": 1.468,
+      "step": 1759
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034224905013308235,
+      "loss": 1.6282,
+      "step": 1760
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000342183014614329,
+      "loss": 1.55,
+      "step": 1761
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000342116947741412,
+      "loss": 1.6792,
+      "step": 1762
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034205084952890045,
+      "loss": 1.4841,
+      "step": 1763
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003419847199913703,
+      "loss": 1.5773,
+      "step": 1764
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034191855914340457,
+      "loss": 1.601,
+      "step": 1765
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003418523669995928,
+      "loss": 1.6164,
+      "step": 1766
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003417861435745318,
+      "loss": 1.583,
+      "step": 1767
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034171988888282513,
+      "loss": 1.6592,
+      "step": 1768
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003416536029390832,
+      "loss": 1.5393,
+      "step": 1769
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003415872857579235,
+      "loss": 1.4544,
+      "step": 1770
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003415209373539701,
+      "loss": 1.2569,
+      "step": 1771
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034145455774185426,
+      "loss": 1.3097,
+      "step": 1772
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034138814693621393,
+      "loss": 1.5111,
+      "step": 1773
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003413217049516941,
+      "loss": 1.539,
+      "step": 1774
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034125523180294646,
+      "loss": 1.5024,
+      "step": 1775
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034118872750462965,
+      "loss": 1.5458,
+      "step": 1776
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003411221920714092,
+      "loss": 1.4306,
+      "step": 1777
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003410556255179575,
+      "loss": 1.5413,
+      "step": 1778
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003409890278589538,
+      "loss": 1.534,
+      "step": 1779
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003409223991090841,
+      "loss": 1.5671,
+      "step": 1780
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034085573928304157,
+      "loss": 1.4948,
+      "step": 1781
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034078904839552574,
+      "loss": 1.4676,
+      "step": 1782
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003407223264612435,
+      "loss": 1.3504,
+      "step": 1783
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034065557349490823,
+      "loss": 1.6263,
+      "step": 1784
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003405887895112403,
+      "loss": 1.4077,
+      "step": 1785
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034052197452496695,
+      "loss": 1.5883,
+      "step": 1786
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034045512855082215,
+      "loss": 1.4701,
+      "step": 1787
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003403882516035467,
+      "loss": 1.6245,
+      "step": 1788
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003403213436978885,
+      "loss": 1.4398,
+      "step": 1789
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003402544048486018,
+      "loss": 1.4369,
+      "step": 1790
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00034018743507044813,
+      "loss": 1.6134,
+      "step": 1791
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003401204343781956,
+      "loss": 1.5836,
+      "step": 1792
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003400534027866191,
+      "loss": 1.5446,
+      "step": 1793
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00033998634031050056,
+      "loss": 1.5113,
+      "step": 1794
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003399192469646285,
+      "loss": 1.2509,
+      "step": 1795
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003398521227637983,
+      "loss": 1.6881,
+      "step": 1796
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00033978496772281224,
+      "loss": 1.7496,
+      "step": 1797
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003397177818564793,
+      "loss": 1.5066,
+      "step": 1798
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003396505651796153,
+      "loss": 1.6852,
+      "step": 1799
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003395833177070428,
+      "loss": 1.3723,
+      "step": 1800
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003395160394535913,
+      "loss": 1.4536,
+      "step": 1801
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003394487304340969,
+      "loss": 1.4792,
+      "step": 1802
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033938139066340254,
+      "loss": 1.4575,
+      "step": 1803
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003393140201563581,
+      "loss": 1.5704,
+      "step": 1804
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033924661892782,
+      "loss": 1.5181,
+      "step": 1805
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033917918699265156,
+      "loss": 1.6646,
+      "step": 1806
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033911172436572285,
+      "loss": 1.7014,
+      "step": 1807
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003390442310619107,
+      "loss": 1.5953,
+      "step": 1808
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033897670709609874,
+      "loss": 1.3817,
+      "step": 1809
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033890915248317743,
+      "loss": 1.5697,
+      "step": 1810
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003388415672380437,
+      "loss": 1.4501,
+      "step": 1811
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003387739513756016,
+      "loss": 1.3583,
+      "step": 1812
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003387063049107617,
+      "loss": 1.3326,
+      "step": 1813
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003386386278584414,
+      "loss": 1.5437,
+      "step": 1814
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033857092023356476,
+      "loss": 1.6454,
+      "step": 1815
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033850318205106274,
+      "loss": 1.3881,
+      "step": 1816
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033843541332587287,
+      "loss": 1.4246,
+      "step": 1817
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003383676140729396,
+      "loss": 1.6625,
+      "step": 1818
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003382997843072139,
+      "loss": 1.6203,
+      "step": 1819
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033823192404365353,
+      "loss": 1.6151,
+      "step": 1820
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003381640332972232,
+      "loss": 1.3215,
+      "step": 1821
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000338096112082894,
+      "loss": 1.6349,
+      "step": 1822
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033802816041564397,
+      "loss": 1.6535,
+      "step": 1823
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003379601783104578,
+      "loss": 1.5193,
+      "step": 1824
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033789216578232684,
+      "loss": 1.5747,
+      "step": 1825
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033782412284624926,
+      "loss": 1.5247,
+      "step": 1826
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003377560495172298,
+      "loss": 1.3304,
+      "step": 1827
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003376879458102801,
+      "loss": 1.5982,
+      "step": 1828
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003376198117404182,
+      "loss": 1.6003,
+      "step": 1829
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003375516473226691,
+      "loss": 1.5091,
+      "step": 1830
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033748345257206437,
+      "loss": 1.3962,
+      "step": 1831
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033741522750364237,
+      "loss": 1.2721,
+      "step": 1832
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033734697213244803,
+      "loss": 1.6126,
+      "step": 1833
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033727868647353296,
+      "loss": 1.6561,
+      "step": 1834
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003372103705419556,
+      "loss": 1.5148,
+      "step": 1835
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003371420243527808,
+      "loss": 1.5349,
+      "step": 1836
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003370736479210805,
+      "loss": 1.556,
+      "step": 1837
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003370052412619328,
+      "loss": 1.5455,
+      "step": 1838
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033693680439042284,
+      "loss": 1.5192,
+      "step": 1839
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033686833732164234,
+      "loss": 1.5965,
+      "step": 1840
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003367998400706895,
+      "loss": 1.6311,
+      "step": 1841
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033673131265266943,
+      "loss": 1.7112,
+      "step": 1842
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033666275508269385,
+      "loss": 1.6788,
+      "step": 1843
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033659416737588085,
+      "loss": 1.4735,
+      "step": 1844
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003365255495473555,
+      "loss": 1.503,
+      "step": 1845
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003364569016122494,
+      "loss": 1.4881,
+      "step": 1846
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003363882235857007,
+      "loss": 1.62,
+      "step": 1847
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003363195154828544,
+      "loss": 1.3868,
+      "step": 1848
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033625077731886185,
+      "loss": 1.5897,
+      "step": 1849
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003361820091088812,
+      "loss": 1.652,
+      "step": 1850
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003361132108680773,
+      "loss": 1.4455,
+      "step": 1851
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003360443826116214,
+      "loss": 1.6701,
+      "step": 1852
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003359755243546915,
+      "loss": 1.5456,
+      "step": 1853
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033590663611247234,
+      "loss": 1.3695,
+      "step": 1854
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000335837717900155,
+      "loss": 1.6413,
+      "step": 1855
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033576876973293734,
+      "loss": 1.6846,
+      "step": 1856
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003356997916260238,
+      "loss": 1.4427,
+      "step": 1857
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003356307835946254,
+      "loss": 1.5308,
+      "step": 1858
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003355617456539598,
+      "loss": 1.6319,
+      "step": 1859
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003354926778192512,
+      "loss": 1.4211,
+      "step": 1860
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003354235801057305,
+      "loss": 1.4225,
+      "step": 1861
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033535445252863506,
+      "loss": 1.6391,
+      "step": 1862
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033528529510320885,
+      "loss": 1.4105,
+      "step": 1863
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033521610784470246,
+      "loss": 1.5133,
+      "step": 1864
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033514689076837305,
+      "loss": 1.6023,
+      "step": 1865
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003350776438894844,
+      "loss": 1.4276,
+      "step": 1866
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00033500836722330675,
+      "loss": 1.6119,
+      "step": 1867
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000334939060785117,
+      "loss": 1.3704,
+      "step": 1868
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033486972459019856,
+      "loss": 1.6189,
+      "step": 1869
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003348003586538415,
+      "loss": 1.5933,
+      "step": 1870
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003347309629913423,
+      "loss": 1.106,
+      "step": 1871
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003346615376180041,
+      "loss": 1.5141,
+      "step": 1872
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033459208254913667,
+      "loss": 1.6069,
+      "step": 1873
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000334522597800056,
+      "loss": 1.5525,
+      "step": 1874
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033445308338608506,
+      "loss": 1.5774,
+      "step": 1875
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033438353932255303,
+      "loss": 1.3831,
+      "step": 1876
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003343139656247958,
+      "loss": 1.5823,
+      "step": 1877
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033424436230815584,
+      "loss": 1.5112,
+      "step": 1878
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033417472938798183,
+      "loss": 1.6225,
+      "step": 1879
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003341050668796294,
+      "loss": 1.2689,
+      "step": 1880
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003340353747984604,
+      "loss": 1.6141,
+      "step": 1881
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033396565315984336,
+      "loss": 1.6571,
+      "step": 1882
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033389590197915333,
+      "loss": 1.8531,
+      "step": 1883
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033382612127177166,
+      "loss": 1.3452,
+      "step": 1884
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033375631105308655,
+      "loss": 1.5616,
+      "step": 1885
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003336864713384924,
+      "loss": 1.4738,
+      "step": 1886
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033361660214339034,
+      "loss": 1.5185,
+      "step": 1887
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003335467034831878,
+      "loss": 1.5006,
+      "step": 1888
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000333476775373299,
+      "loss": 1.4968,
+      "step": 1889
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003334068178291443,
+      "loss": 1.5314,
+      "step": 1890
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033333683086615083,
+      "loss": 1.5624,
+      "step": 1891
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033326681449975196,
+      "loss": 1.4159,
+      "step": 1892
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033319676874538784,
+      "loss": 1.5174,
+      "step": 1893
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003331266936185048,
+      "loss": 1.4717,
+      "step": 1894
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033305658913455595,
+      "loss": 1.4379,
+      "step": 1895
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003329864553090006,
+      "loss": 1.465,
+      "step": 1896
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033291629215730464,
+      "loss": 1.2932,
+      "step": 1897
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003328460996949405,
+      "loss": 1.5493,
+      "step": 1898
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003327758779373869,
+      "loss": 1.4994,
+      "step": 1899
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033270562690012923,
+      "loss": 1.5171,
+      "step": 1900
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003326353465986592,
+      "loss": 1.4099,
+      "step": 1901
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003325650370484749,
+      "loss": 1.6249,
+      "step": 1902
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003324946982650812,
+      "loss": 1.6828,
+      "step": 1903
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000332424330263989,
+      "loss": 1.6698,
+      "step": 1904
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003323539330607159,
+      "loss": 1.5492,
+      "step": 1905
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003322835066707859,
+      "loss": 1.451,
+      "step": 1906
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033221305110972934,
+      "loss": 1.4611,
+      "step": 1907
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003321425663930831,
+      "loss": 1.4244,
+      "step": 1908
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003320720525363905,
+      "loss": 1.6112,
+      "step": 1909
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033200150955520116,
+      "loss": 1.622,
+      "step": 1910
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003319309374650712,
+      "loss": 1.5114,
+      "step": 1911
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003318603362815632,
+      "loss": 1.5238,
+      "step": 1912
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003317897060202462,
+      "loss": 1.6671,
+      "step": 1913
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003317190466966954,
+      "loss": 1.518,
+      "step": 1914
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003316483583264927,
+      "loss": 1.4391,
+      "step": 1915
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033157764092522625,
+      "loss": 1.5168,
+      "step": 1916
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003315068945084906,
+      "loss": 1.5339,
+      "step": 1917
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033143611909188675,
+      "loss": 1.5525,
+      "step": 1918
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033136531469102207,
+      "loss": 1.3056,
+      "step": 1919
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003312944813215103,
+      "loss": 1.5997,
+      "step": 1920
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003312236189989717,
+      "loss": 1.7574,
+      "step": 1921
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033115272773903275,
+      "loss": 1.5668,
+      "step": 1922
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033108180755732635,
+      "loss": 1.473,
+      "step": 1923
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003310108584694919,
+      "loss": 1.6924,
+      "step": 1924
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033093988049117485,
+      "loss": 1.5389,
+      "step": 1925
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033086887363802754,
+      "loss": 1.4801,
+      "step": 1926
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003307978379257081,
+      "loss": 1.7996,
+      "step": 1927
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003307267733698816,
+      "loss": 1.5307,
+      "step": 1928
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000330655679986219,
+      "loss": 1.326,
+      "step": 1929
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003305845577903977,
+      "loss": 1.1977,
+      "step": 1930
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033051340679810177,
+      "loss": 1.6035,
+      "step": 1931
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003304422270250213,
+      "loss": 1.4315,
+      "step": 1932
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003303710184868528,
+      "loss": 1.539,
+      "step": 1933
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033029978119929925,
+      "loss": 1.5311,
+      "step": 1934
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00033022851517806987,
+      "loss": 1.3792,
+      "step": 1935
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003301572204388801,
+      "loss": 1.4335,
+      "step": 1936
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00033008589699745205,
+      "loss": 1.4464,
+      "step": 1937
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003300145448695137,
+      "loss": 1.5223,
+      "step": 1938
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032994316407079983,
+      "loss": 1.2805,
+      "step": 1939
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032987175461705117,
+      "loss": 1.4169,
+      "step": 1940
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000329800316524015,
+      "loss": 1.4812,
+      "step": 1941
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032972884980744474,
+      "loss": 1.3449,
+      "step": 1942
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003296573544831002,
+      "loss": 1.7321,
+      "step": 1943
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003295858305667477,
+      "loss": 1.6111,
+      "step": 1944
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032951427807415955,
+      "loss": 1.4845,
+      "step": 1945
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032944269702111443,
+      "loss": 1.6439,
+      "step": 1946
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032937108742339753,
+      "loss": 1.4245,
+      "step": 1947
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032929944929680003,
+      "loss": 1.5163,
+      "step": 1948
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003292277826571196,
+      "loss": 1.5479,
+      "step": 1949
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032915608752016016,
+      "loss": 1.7732,
+      "step": 1950
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032908436390173193,
+      "loss": 1.573,
+      "step": 1951
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032901261181765134,
+      "loss": 1.6828,
+      "step": 1952
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003289408312837412,
+      "loss": 1.6977,
+      "step": 1953
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032886902231583047,
+      "loss": 1.4149,
+      "step": 1954
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003287971849297545,
+      "loss": 1.7386,
+      "step": 1955
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003287253191413548,
+      "loss": 1.2628,
+      "step": 1956
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003286534249664792,
+      "loss": 1.6759,
+      "step": 1957
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032858150242098187,
+      "loss": 1.3769,
+      "step": 1958
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000328509551520723,
+      "loss": 1.3409,
+      "step": 1959
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032843757228156936,
+      "loss": 1.6713,
+      "step": 1960
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003283655647193937,
+      "loss": 1.5391,
+      "step": 1961
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032829352885007507,
+      "loss": 1.3437,
+      "step": 1962
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003282214646894989,
+      "loss": 1.6157,
+      "step": 1963
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003281493722535567,
+      "loss": 1.6936,
+      "step": 1964
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003280772515581463,
+      "loss": 1.7196,
+      "step": 1965
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032800510261917176,
+      "loss": 1.6345,
+      "step": 1966
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003279329254525433,
+      "loss": 1.6095,
+      "step": 1967
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032786072007417734,
+      "loss": 1.4927,
+      "step": 1968
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003277884864999968,
+      "loss": 1.4369,
+      "step": 1969
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003277162247459304,
+      "loss": 1.4614,
+      "step": 1970
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003276439348279134,
+      "loss": 1.5591,
+      "step": 1971
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003275716167618871,
+      "loss": 1.5774,
+      "step": 1972
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003274992705637992,
+      "loss": 1.5112,
+      "step": 1973
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032742689624960325,
+      "loss": 1.6582,
+      "step": 1974
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032735449383525933,
+      "loss": 1.5631,
+      "step": 1975
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032728206333673356,
+      "loss": 1.6436,
+      "step": 1976
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003272096047699984,
+      "loss": 1.5109,
+      "step": 1977
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003271371181510323,
+      "loss": 1.5519,
+      "step": 1978
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032706460349582,
+      "loss": 1.4462,
+      "step": 1979
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003269920608203525,
+      "loss": 1.7354,
+      "step": 1980
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032691949014062677,
+      "loss": 1.3683,
+      "step": 1981
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032684689147264613,
+      "loss": 1.3323,
+      "step": 1982
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003267742648324201,
+      "loss": 1.2685,
+      "step": 1983
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032670161023596407,
+      "loss": 1.4312,
+      "step": 1984
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032662892769930006,
+      "loss": 1.3586,
+      "step": 1985
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032655621723845583,
+      "loss": 1.5895,
+      "step": 1986
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032648347886946564,
+      "loss": 1.5359,
+      "step": 1987
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032641071260836957,
+      "loss": 1.712,
+      "step": 1988
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003263379184712141,
+      "loss": 1.4663,
+      "step": 1989
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003262650964740518,
+      "loss": 1.4356,
+      "step": 1990
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003261922466329413,
+      "loss": 1.5134,
+      "step": 1991
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032611936896394743,
+      "loss": 1.5289,
+      "step": 1992
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003260464634831412,
+      "loss": 1.6237,
+      "step": 1993
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032597353020659967,
+      "loss": 1.4875,
+      "step": 1994
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003259005691504061,
+      "loss": 1.7081,
+      "step": 1995
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032582758033064985,
+      "loss": 1.3447,
+      "step": 1996
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003257545637634264,
+      "loss": 1.4183,
+      "step": 1997
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003256815194648373,
+      "loss": 1.5849,
+      "step": 1998
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00032560844745099023,
+      "loss": 1.6191,
+      "step": 1999
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003255353477379991,
+      "loss": 1.4454,
+      "step": 2000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003254622203419838,
+      "loss": 1.603,
+      "step": 2001
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003253890652790704,
+      "loss": 1.3977,
+      "step": 2002
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000325315882565391,
+      "loss": 1.473,
+      "step": 2003
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032524267221708376,
+      "loss": 1.3753,
+      "step": 2004
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003251694342502932,
+      "loss": 1.5265,
+      "step": 2005
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003250961686811696,
+      "loss": 1.3411,
+      "step": 2006
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003250228755258695,
+      "loss": 1.2354,
+      "step": 2007
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003249495548005554,
+      "loss": 1.5504,
+      "step": 2008
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032487620652139616,
+      "loss": 1.5988,
+      "step": 2009
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032480283070456645,
+      "loss": 1.5573,
+      "step": 2010
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003247294273662469,
+      "loss": 1.4763,
+      "step": 2011
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003246559965226246,
+      "loss": 1.7562,
+      "step": 2012
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003245825381898925,
+      "loss": 1.6251,
+      "step": 2013
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003245090523842496,
+      "loss": 1.6166,
+      "step": 2014
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003244355391219009,
+      "loss": 1.3551,
+      "step": 2015
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003243619984190575,
+      "loss": 1.5131,
+      "step": 2016
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032428843029193673,
+      "loss": 1.6308,
+      "step": 2017
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003242148347567617,
+      "loss": 1.6082,
+      "step": 2018
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003241412118297617,
+      "loss": 1.4506,
+      "step": 2019
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032406756152717197,
+      "loss": 1.3692,
+      "step": 2020
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000323993883865234,
+      "loss": 1.5878,
+      "step": 2021
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032392017886019516,
+      "loss": 1.6522,
+      "step": 2022
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003238464465283087,
+      "loss": 1.708,
+      "step": 2023
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032377268688583424,
+      "loss": 1.689,
+      "step": 2024
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003236988999490371,
+      "loss": 1.6841,
+      "step": 2025
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003236250857341888,
+      "loss": 1.4976,
+      "step": 2026
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032355124425756696,
+      "loss": 1.5996,
+      "step": 2027
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032347737553545486,
+      "loss": 1.4894,
+      "step": 2028
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032340347958414214,
+      "loss": 1.4708,
+      "step": 2029
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003233295564199243,
+      "loss": 1.6562,
+      "step": 2030
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003232556060591029,
+      "loss": 1.6137,
+      "step": 2031
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032318162851798535,
+      "loss": 1.3795,
+      "step": 2032
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032310762381288524,
+      "loss": 1.4064,
+      "step": 2033
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003230335919601221,
+      "loss": 1.5622,
+      "step": 2034
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003229595329760214,
+      "loss": 1.5003,
+      "step": 2035
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032288544687691446,
+      "loss": 1.4169,
+      "step": 2036
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003228113336791389,
+      "loss": 1.4235,
+      "step": 2037
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003227371933990381,
+      "loss": 1.6628,
+      "step": 2038
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032266302605296143,
+      "loss": 1.6251,
+      "step": 2039
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032258883165726433,
+      "loss": 1.3539,
+      "step": 2040
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000322514610228308,
+      "loss": 1.3329,
+      "step": 2041
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032244036178245985,
+      "loss": 1.7033,
+      "step": 2042
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032236608633609304,
+      "loss": 1.6954,
+      "step": 2043
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003222917839055868,
+      "loss": 1.3218,
+      "step": 2044
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032221745450732633,
+      "loss": 1.3517,
+      "step": 2045
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003221430981577027,
+      "loss": 1.3907,
+      "step": 2046
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032206871487311297,
+      "loss": 1.6568,
+      "step": 2047
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032199430466996014,
+      "loss": 1.667,
+      "step": 2048
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032191986756465303,
+      "loss": 1.5948,
+      "step": 2049
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003218454035736066,
+      "loss": 1.5012,
+      "step": 2050
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032177091271324155,
+      "loss": 1.4494,
+      "step": 2051
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032169639499998466,
+      "loss": 1.6441,
+      "step": 2052
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003216218504502685,
+      "loss": 1.6031,
+      "step": 2053
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003215472790805317,
+      "loss": 1.3038,
+      "step": 2054
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032147268090721863,
+      "loss": 1.6119,
+      "step": 2055
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032139805594677963,
+      "loss": 1.5654,
+      "step": 2056
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003213234042156711,
+      "loss": 1.5851,
+      "step": 2057
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032124872573035515,
+      "loss": 1.751,
+      "step": 2058
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003211740205072999,
+      "loss": 1.427,
+      "step": 2059
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032109928856297927,
+      "loss": 1.4434,
+      "step": 2060
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032102452991387314,
+      "loss": 1.3392,
+      "step": 2061
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032094974457646733,
+      "loss": 1.4512,
+      "step": 2062
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003208749325672534,
+      "loss": 1.4103,
+      "step": 2063
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032080009390272895,
+      "loss": 1.5684,
+      "step": 2064
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00032072522859939734,
+      "loss": 1.3734,
+      "step": 2065
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003206503366737679,
+      "loss": 1.6504,
+      "step": 2066
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003205754181423557,
+      "loss": 1.3231,
+      "step": 2067
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003205004730216818,
+      "loss": 1.5593,
+      "step": 2068
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003204255013282731,
+      "loss": 1.5828,
+      "step": 2069
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003203505030786623,
+      "loss": 1.5053,
+      "step": 2070
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00032027547828938804,
+      "loss": 1.6788,
+      "step": 2071
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003202004269769947,
+      "loss": 1.6237,
+      "step": 2072
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003201253491580326,
+      "loss": 1.3642,
+      "step": 2073
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00032005024484905797,
+      "loss": 1.2902,
+      "step": 2074
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031997511406663267,
+      "loss": 1.4843,
+      "step": 2075
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031989995682732463,
+      "loss": 1.5066,
+      "step": 2076
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003198247731477074,
+      "loss": 1.5495,
+      "step": 2077
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031974956304436057,
+      "loss": 1.4473,
+      "step": 2078
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003196743265338693,
+      "loss": 1.6385,
+      "step": 2079
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003195990636328249,
+      "loss": 1.5221,
+      "step": 2080
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003195237743578242,
+      "loss": 1.3544,
+      "step": 2081
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003194484587254701,
+      "loss": 1.5859,
+      "step": 2082
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031937311675237114,
+      "loss": 1.4149,
+      "step": 2083
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003192977484551416,
+      "loss": 1.3341,
+      "step": 2084
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031922235385040175,
+      "loss": 1.5948,
+      "step": 2085
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003191469329547777,
+      "loss": 1.3656,
+      "step": 2086
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003190714857849011,
+      "loss": 1.5917,
+      "step": 2087
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031899601235740963,
+      "loss": 1.4097,
+      "step": 2088
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031892051268894657,
+      "loss": 1.4178,
+      "step": 2089
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031884498679616115,
+      "loss": 1.5764,
+      "step": 2090
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031876943469570835,
+      "loss": 1.5756,
+      "step": 2091
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003186938564042489,
+      "loss": 1.5744,
+      "step": 2092
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003186182519384493,
+      "loss": 1.5441,
+      "step": 2093
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003185426213149818,
+      "loss": 1.4905,
+      "step": 2094
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031846696455052435,
+      "loss": 1.6935,
+      "step": 2095
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000318391281661761,
+      "loss": 1.4936,
+      "step": 2096
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003183155726653811,
+      "loss": 1.5043,
+      "step": 2097
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003182398375780801,
+      "loss": 1.5413,
+      "step": 2098
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000318164076416559,
+      "loss": 1.7472,
+      "step": 2099
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031808828919752477,
+      "loss": 1.5987,
+      "step": 2100
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031801247593768987,
+      "loss": 1.53,
+      "step": 2101
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031793663665377267,
+      "loss": 1.4011,
+      "step": 2102
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003178607713624972,
+      "loss": 1.5092,
+      "step": 2103
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003177848800805933,
+      "loss": 1.5554,
+      "step": 2104
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031770896282479644,
+      "loss": 1.4052,
+      "step": 2105
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003176330196118479,
+      "loss": 1.5562,
+      "step": 2106
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031755705045849464,
+      "loss": 1.5067,
+      "step": 2107
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003174810553814894,
+      "loss": 1.4707,
+      "step": 2108
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003174050343975906,
+      "loss": 1.9285,
+      "step": 2109
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031732898752356235,
+      "loss": 1.5314,
+      "step": 2110
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003172529147761744,
+      "loss": 1.4053,
+      "step": 2111
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003171768161722025,
+      "loss": 1.512,
+      "step": 2112
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031710069172842764,
+      "loss": 1.403,
+      "step": 2113
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031702454146163703,
+      "loss": 1.5535,
+      "step": 2114
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031694836538862306,
+      "loss": 1.4084,
+      "step": 2115
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003168721635261842,
+      "loss": 1.4329,
+      "step": 2116
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031679593589112445,
+      "loss": 1.6899,
+      "step": 2117
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031671968250025355,
+      "loss": 1.1082,
+      "step": 2118
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003166434033703868,
+      "loss": 1.3326,
+      "step": 2119
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031656709851834525,
+      "loss": 1.3954,
+      "step": 2120
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003164907679609557,
+      "loss": 1.58,
+      "step": 2121
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003164144117150505,
+      "loss": 1.4451,
+      "step": 2122
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031633802979746765,
+      "loss": 1.5545,
+      "step": 2123
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000316261622225051,
+      "loss": 1.7524,
+      "step": 2124
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003161851890146499,
+      "loss": 1.4001,
+      "step": 2125
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031610873018311927,
+      "loss": 1.5964,
+      "step": 2126
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031603224574731995,
+      "loss": 1.7643,
+      "step": 2127
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003159557357241181,
+      "loss": 1.2444,
+      "step": 2128
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031587920013038587,
+      "loss": 1.5525,
+      "step": 2129
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003158026389830007,
+      "loss": 1.3515,
+      "step": 2130
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000315726052298846,
+      "loss": 1.4727,
+      "step": 2131
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031564944009481057,
+      "loss": 1.5004,
+      "step": 2132
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031557280238778883,
+      "loss": 1.4906,
+      "step": 2133
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000315496139194681,
+      "loss": 1.4864,
+      "step": 2134
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003154194505323929,
+      "loss": 1.4382,
+      "step": 2135
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031534273641783576,
+      "loss": 1.5968,
+      "step": 2136
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00031526599686792665,
+      "loss": 1.4937,
+      "step": 2137
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003151892318995881,
+      "loss": 1.5079,
+      "step": 2138
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031511244152974824,
+      "loss": 1.3373,
+      "step": 2139
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000315035625775341,
+      "loss": 1.5716,
+      "step": 2140
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031495878465330577,
+      "loss": 1.4495,
+      "step": 2141
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003148819181805874,
+      "loss": 1.6901,
+      "step": 2142
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031480502637413663,
+      "loss": 1.4761,
+      "step": 2143
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031472810925090944,
+      "loss": 1.7003,
+      "step": 2144
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003146511668278677,
+      "loss": 1.5052,
+      "step": 2145
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031457419912197875,
+      "loss": 1.4834,
+      "step": 2146
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003144972061502154,
+      "loss": 1.5681,
+      "step": 2147
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003144972061502154,
+      "loss": 1.4511,
+      "step": 2148
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003144201879295562,
+      "loss": 1.5254,
+      "step": 2149
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031434314447698514,
+      "loss": 1.4782,
+      "step": 2150
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003142660758094919,
+      "loss": 1.7858,
+      "step": 2151
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003141889819440715,
+      "loss": 1.5168,
+      "step": 2152
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003141118628977248,
+      "loss": 1.2965,
+      "step": 2153
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000314034718687458,
+      "loss": 1.6234,
+      "step": 2154
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031395754933028294,
+      "loss": 1.5827,
+      "step": 2155
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003138803548432169,
+      "loss": 1.6718,
+      "step": 2156
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000313803135243283,
+      "loss": 1.5637,
+      "step": 2157
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031372589054750954,
+      "loss": 1.4651,
+      "step": 2158
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031364862077293054,
+      "loss": 1.3266,
+      "step": 2159
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031357132593658546,
+      "loss": 1.6721,
+      "step": 2160
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003134940060555194,
+      "loss": 1.4916,
+      "step": 2161
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003134166611467829,
+      "loss": 1.454,
+      "step": 2162
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031333929122743197,
+      "loss": 1.5788,
+      "step": 2163
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031326189631452836,
+      "loss": 1.5562,
+      "step": 2164
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000313184476425139,
+      "loss": 1.3938,
+      "step": 2165
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003131070315763367,
+      "loss": 1.4993,
+      "step": 2166
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031302956178519946,
+      "loss": 1.4929,
+      "step": 2167
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003129520670688109,
+      "loss": 1.4302,
+      "step": 2168
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003128745474442601,
+      "loss": 1.6303,
+      "step": 2169
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003127970029286418,
+      "loss": 1.4557,
+      "step": 2170
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031271943353905593,
+      "loss": 1.3236,
+      "step": 2171
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031264183929260824,
+      "loss": 1.5851,
+      "step": 2172
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003125642202064096,
+      "loss": 1.4012,
+      "step": 2173
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003124865762975768,
+      "loss": 1.5358,
+      "step": 2174
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003124089075832317,
+      "loss": 1.3195,
+      "step": 2175
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031233121408050175,
+      "loss": 1.3756,
+      "step": 2176
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031225349580652,
+      "loss": 1.6071,
+      "step": 2177
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003121757527784248,
+      "loss": 1.4902,
+      "step": 2178
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031209798501336005,
+      "loss": 1.4932,
+      "step": 2179
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000312020192528475,
+      "loss": 1.6622,
+      "step": 2180
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031194237534092463,
+      "loss": 1.6962,
+      "step": 2181
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000311864533467869,
+      "loss": 1.5594,
+      "step": 2182
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031178666692647383,
+      "loss": 1.5355,
+      "step": 2183
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003117087757339102,
+      "loss": 1.5591,
+      "step": 2184
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003116308599073547,
+      "loss": 1.6026,
+      "step": 2185
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003115529194639893,
+      "loss": 1.4231,
+      "step": 2186
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003114749544210014,
+      "loss": 1.5947,
+      "step": 2187
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031139696479558384,
+      "loss": 1.624,
+      "step": 2188
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003113189506049348,
+      "loss": 1.4124,
+      "step": 2189
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003112409118662581,
+      "loss": 1.1582,
+      "step": 2190
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031116284859676266,
+      "loss": 1.5302,
+      "step": 2191
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003110847608136631,
+      "loss": 1.4591,
+      "step": 2192
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003110066485341792,
+      "loss": 1.5601,
+      "step": 2193
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003109285117755364,
+      "loss": 1.4927,
+      "step": 2194
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031085035055496537,
+      "loss": 1.6208,
+      "step": 2195
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031077216488970206,
+      "loss": 1.4464,
+      "step": 2196
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031069395479698807,
+      "loss": 1.4674,
+      "step": 2197
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003106157202940703,
+      "loss": 1.4069,
+      "step": 2198
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031053746139820095,
+      "loss": 1.7044,
+      "step": 2199
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00031045917812663764,
+      "loss": 1.3827,
+      "step": 2200
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003103808704966434,
+      "loss": 1.6477,
+      "step": 2201
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003103025385254866,
+      "loss": 1.5811,
+      "step": 2202
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003102241822304411,
+      "loss": 1.3914,
+      "step": 2203
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003101458016287858,
+      "loss": 1.6075,
+      "step": 2204
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003100673967378053,
+      "loss": 1.5151,
+      "step": 2205
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003099889675747894,
+      "loss": 1.5138,
+      "step": 2206
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003099105141570333,
+      "loss": 1.4911,
+      "step": 2207
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003098320365018375,
+      "loss": 1.4401,
+      "step": 2208
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003097535346265079,
+      "loss": 1.4382,
+      "step": 2209
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030967500854835574,
+      "loss": 1.1917,
+      "step": 2210
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003095964582846975,
+      "loss": 1.5046,
+      "step": 2211
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030951788385285504,
+      "loss": 1.6125,
+      "step": 2212
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003094392852701557,
+      "loss": 1.6434,
+      "step": 2213
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030936066255393197,
+      "loss": 1.4356,
+      "step": 2214
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003092820157215216,
+      "loss": 1.3685,
+      "step": 2215
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003092033447902679,
+      "loss": 1.4102,
+      "step": 2216
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030912464977751947,
+      "loss": 1.447,
+      "step": 2217
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030904593070062977,
+      "loss": 1.3449,
+      "step": 2218
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030896718757695825,
+      "loss": 1.5513,
+      "step": 2219
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003088884204238692,
+      "loss": 1.3549,
+      "step": 2220
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030880962925873226,
+      "loss": 1.4394,
+      "step": 2221
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003087308140989225,
+      "loss": 1.5096,
+      "step": 2222
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003086519749618202,
+      "loss": 1.4672,
+      "step": 2223
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003085731118648111,
+      "loss": 1.3754,
+      "step": 2224
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003084942248252858,
+      "loss": 1.5723,
+      "step": 2225
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003084153138606407,
+      "loss": 1.5,
+      "step": 2226
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030833637898827705,
+      "loss": 1.502,
+      "step": 2227
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030825742022560164,
+      "loss": 1.5716,
+      "step": 2228
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003081784375900264,
+      "loss": 1.3123,
+      "step": 2229
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003080994310989685,
+      "loss": 1.5783,
+      "step": 2230
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003080204007698506,
+      "loss": 1.612,
+      "step": 2231
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003079413466201004,
+      "loss": 1.7377,
+      "step": 2232
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030786226866715076,
+      "loss": 1.7135,
+      "step": 2233
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003077831669284401,
+      "loss": 1.4777,
+      "step": 2234
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030770404142141177,
+      "loss": 1.4628,
+      "step": 2235
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003076248921635146,
+      "loss": 1.5371,
+      "step": 2236
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003075457191722026,
+      "loss": 1.4201,
+      "step": 2237
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003074665224649349,
+      "loss": 1.6616,
+      "step": 2238
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030738730205917596,
+      "loss": 1.3769,
+      "step": 2239
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030730805797239544,
+      "loss": 1.5881,
+      "step": 2240
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003072287902220683,
+      "loss": 1.6685,
+      "step": 2241
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003071494988256745,
+      "loss": 1.5891,
+      "step": 2242
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030707018380069943,
+      "loss": 1.736,
+      "step": 2243
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030699084516463374,
+      "loss": 1.4268,
+      "step": 2244
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003069114829349729,
+      "loss": 1.3627,
+      "step": 2245
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030683209712921807,
+      "loss": 1.5069,
+      "step": 2246
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030675268776487536,
+      "loss": 1.6798,
+      "step": 2247
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003066732548594561,
+      "loss": 1.7199,
+      "step": 2248
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030659379843047665,
+      "loss": 1.4227,
+      "step": 2249
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030651431849545893,
+      "loss": 1.6368,
+      "step": 2250
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003064348150719298,
+      "loss": 1.5206,
+      "step": 2251
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030635528817742123,
+      "loss": 1.4751,
+      "step": 2252
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030627573782947065,
+      "loss": 1.3312,
+      "step": 2253
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003061961640456202,
+      "loss": 1.3179,
+      "step": 2254
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003061165668434178,
+      "loss": 1.5926,
+      "step": 2255
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003060369462404159,
+      "loss": 1.4785,
+      "step": 2256
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030595730225417274,
+      "loss": 1.4776,
+      "step": 2257
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003058776349022511,
+      "loss": 1.606,
+      "step": 2258
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003057979442022193,
+      "loss": 1.4439,
+      "step": 2259
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003057182301716508,
+      "loss": 1.5091,
+      "step": 2260
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000305638492828124,
+      "loss": 1.4095,
+      "step": 2261
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003055587321892227,
+      "loss": 1.5059,
+      "step": 2262
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003054789482725355,
+      "loss": 1.6574,
+      "step": 2263
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030539914109565646,
+      "loss": 1.7457,
+      "step": 2264
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030531931067618467,
+      "loss": 1.6806,
+      "step": 2265
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003052394570317242,
+      "loss": 1.6636,
+      "step": 2266
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030515958017988444,
+      "loss": 1.4719,
+      "step": 2267
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003050796801382798,
+      "loss": 1.5299,
+      "step": 2268
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030499975692452986,
+      "loss": 1.5115,
+      "step": 2269
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003049198105562591,
+      "loss": 1.5331,
+      "step": 2270
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003048398410510975,
+      "loss": 1.6145,
+      "step": 2271
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003047598484266798,
+      "loss": 1.4509,
+      "step": 2272
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030467983270064593,
+      "loss": 1.4959,
+      "step": 2273
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00030459979389064096,
+      "loss": 1.4253,
+      "step": 2274
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030451973201431505,
+      "loss": 1.4822,
+      "step": 2275
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003044396470893234,
+      "loss": 1.4122,
+      "step": 2276
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030435953913332634,
+      "loss": 1.6513,
+      "step": 2277
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003042794081639893,
+      "loss": 1.6873,
+      "step": 2278
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003041992541989826,
+      "loss": 1.5009,
+      "step": 2279
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030411907725598193,
+      "loss": 1.5664,
+      "step": 2280
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003040388773526678,
+      "loss": 1.5517,
+      "step": 2281
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003039586545067259,
+      "loss": 1.6624,
+      "step": 2282
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030387840873584687,
+      "loss": 1.6086,
+      "step": 2283
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003037981400577267,
+      "loss": 1.5562,
+      "step": 2284
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030371784849006603,
+      "loss": 1.4437,
+      "step": 2285
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030363753405057074,
+      "loss": 1.5218,
+      "step": 2286
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003035571967569519,
+      "loss": 1.6605,
+      "step": 2287
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003034768366269253,
+      "loss": 1.4013,
+      "step": 2288
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000303396453678212,
+      "loss": 1.3849,
+      "step": 2289
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030331604792853807,
+      "loss": 1.4494,
+      "step": 2290
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030323561939563456,
+      "loss": 1.6489,
+      "step": 2291
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003031551680972375,
+      "loss": 1.4347,
+      "step": 2292
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000303074694051088,
+      "loss": 1.4726,
+      "step": 2293
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030299419727493224,
+      "loss": 1.7304,
+      "step": 2294
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030299419727493224,
+      "loss": 1.5444,
+      "step": 2295
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030291367778652117,
+      "loss": 1.7775,
+      "step": 2296
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030283313560361115,
+      "loss": 1.4137,
+      "step": 2297
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003027525707439632,
+      "loss": 1.6202,
+      "step": 2298
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030267198322534355,
+      "loss": 1.5534,
+      "step": 2299
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030259137306552325,
+      "loss": 1.4859,
+      "step": 2300
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003025107402822785,
+      "loss": 1.3892,
+      "step": 2301
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003024300848933903,
+      "loss": 1.6247,
+      "step": 2302
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003023494069166449,
+      "loss": 1.4017,
+      "step": 2303
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030226870636983335,
+      "loss": 1.2723,
+      "step": 2304
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003021879832707517,
+      "loss": 1.4788,
+      "step": 2305
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000302107237637201,
+      "loss": 1.6187,
+      "step": 2306
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003020264694869872,
+      "loss": 1.5199,
+      "step": 2307
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003019456788379213,
+      "loss": 1.447,
+      "step": 2308
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030186486570781934,
+      "loss": 1.6042,
+      "step": 2309
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000301784030114502,
+      "loss": 1.763,
+      "step": 2310
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030170317207579535,
+      "loss": 1.5367,
+      "step": 2311
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030162229160953,
+      "loss": 1.5148,
+      "step": 2312
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003015413887335419,
+      "loss": 1.6551,
+      "step": 2313
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030146046346567147,
+      "loss": 1.479,
+      "step": 2314
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003013795158237645,
+      "loss": 1.5289,
+      "step": 2315
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003012985458256715,
+      "loss": 1.563,
+      "step": 2316
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000301217553489248,
+      "loss": 1.5238,
+      "step": 2317
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030113653883235436,
+      "loss": 1.4766,
+      "step": 2318
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003010555018728559,
+      "loss": 1.5415,
+      "step": 2319
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003009744426286229,
+      "loss": 1.4943,
+      "step": 2320
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030089336111753057,
+      "loss": 1.4161,
+      "step": 2321
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030081225735745893,
+      "loss": 1.6246,
+      "step": 2322
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030073113136629294,
+      "loss": 1.4892,
+      "step": 2323
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003006499831619225,
+      "loss": 1.5485,
+      "step": 2324
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003005688127622424,
+      "loss": 1.5049,
+      "step": 2325
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030048762018515236,
+      "loss": 1.3711,
+      "step": 2326
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030040640544855686,
+      "loss": 1.397,
+      "step": 2327
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030032516857036545,
+      "loss": 1.5181,
+      "step": 2328
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003002439095684924,
+      "loss": 1.7501,
+      "step": 2329
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030016262846085695,
+      "loss": 1.5274,
+      "step": 2330
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003000813252653832,
+      "loss": 1.6014,
+      "step": 2331
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00030000000000000003,
+      "loss": 1.5863,
+      "step": 2332
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00029991865268264137,
+      "loss": 1.4341,
+      "step": 2333
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00029983728333124584,
+      "loss": 1.486,
+      "step": 2334
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00029975589196375705,
+      "loss": 1.6085,
+      "step": 2335
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00029967447859812336,
+      "loss": 1.7431,
+      "step": 2336
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002995930432522981,
+      "loss": 1.6025,
+      "step": 2337
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002995115859442392,
+      "loss": 1.4088,
+      "step": 2338
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002994301066919098,
+      "loss": 1.4305,
+      "step": 2339
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002993486055132776,
+      "loss": 1.4672,
+      "step": 2340
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002992670824263152,
+      "loss": 1.3983,
+      "step": 2341
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002991855374490001,
+      "loss": 1.5942,
+      "step": 2342
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029910397059931457,
+      "loss": 1.4921,
+      "step": 2343
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029902238189524565,
+      "loss": 1.6773,
+      "step": 2344
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002989407713547853,
+      "loss": 1.38,
+      "step": 2345
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002988591389959303,
+      "loss": 1.5477,
+      "step": 2346
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029877748483668216,
+      "loss": 1.4641,
+      "step": 2347
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029869580889504727,
+      "loss": 1.6478,
+      "step": 2348
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029861411118903665,
+      "loss": 1.473,
+      "step": 2349
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029853239173666644,
+      "loss": 1.4006,
+      "step": 2350
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029845065055595727,
+      "loss": 1.33,
+      "step": 2351
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029836888766493475,
+      "loss": 1.5083,
+      "step": 2352
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029828710308162917,
+      "loss": 1.4423,
+      "step": 2353
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029820529682407574,
+      "loss": 1.47,
+      "step": 2354
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029812346891031424,
+      "loss": 1.3443,
+      "step": 2355
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029804161935838937,
+      "loss": 1.4673,
+      "step": 2356
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029795974818635064,
+      "loss": 1.3311,
+      "step": 2357
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029787785541225215,
+      "loss": 1.736,
+      "step": 2358
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000297795941054153,
+      "loss": 1.5232,
+      "step": 2359
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029771400513011685,
+      "loss": 1.5319,
+      "step": 2360
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029763204765821216,
+      "loss": 1.5825,
+      "step": 2361
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002975500686565123,
+      "loss": 1.3937,
+      "step": 2362
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029746806814309507,
+      "loss": 1.3847,
+      "step": 2363
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029738604613604335,
+      "loss": 1.6024,
+      "step": 2364
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002973040026534445,
+      "loss": 1.4688,
+      "step": 2365
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002972219377133909,
+      "loss": 1.8117,
+      "step": 2366
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002971398513339793,
+      "loss": 1.278,
+      "step": 2367
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029705774353331157,
+      "loss": 1.3163,
+      "step": 2368
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029697561432949385,
+      "loss": 1.3999,
+      "step": 2369
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002968934637406375,
+      "loss": 1.5372,
+      "step": 2370
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002968112917848581,
+      "loss": 1.4746,
+      "step": 2371
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002967290984802765,
+      "loss": 1.7613,
+      "step": 2372
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002966468838450177,
+      "loss": 1.5342,
+      "step": 2373
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029656464789721166,
+      "loss": 1.5181,
+      "step": 2374
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029648239065499315,
+      "loss": 1.4005,
+      "step": 2375
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029640011213650147,
+      "loss": 1.4587,
+      "step": 2376
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002963178123598807,
+      "loss": 1.6324,
+      "step": 2377
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029623549134327943,
+      "loss": 1.3577,
+      "step": 2378
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002961531491048512,
+      "loss": 1.5543,
+      "step": 2379
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029607078566275404,
+      "loss": 1.5994,
+      "step": 2380
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029598840103515074,
+      "loss": 1.4307,
+      "step": 2381
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002959059952402088,
+      "loss": 1.5111,
+      "step": 2382
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029582356829610014,
+      "loss": 1.5827,
+      "step": 2383
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029574112022100175,
+      "loss": 1.7005,
+      "step": 2384
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002956586510330949,
+      "loss": 1.4302,
+      "step": 2385
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002955761607505657,
+      "loss": 1.6469,
+      "step": 2386
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029549364939160497,
+      "loss": 1.4539,
+      "step": 2387
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029541111697440806,
+      "loss": 1.3371,
+      "step": 2388
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029532856351717496,
+      "loss": 1.4311,
+      "step": 2389
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029524598903811033,
+      "loss": 1.6396,
+      "step": 2390
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002951633935554236,
+      "loss": 1.4966,
+      "step": 2391
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002950807770873285,
+      "loss": 1.6439,
+      "step": 2392
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002949981396520438,
+      "loss": 1.8061,
+      "step": 2393
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002949154812677925,
+      "loss": 1.7254,
+      "step": 2394
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002948328019528026,
+      "loss": 1.4805,
+      "step": 2395
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029475010172530633,
+      "loss": 1.42,
+      "step": 2396
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029466738060354085,
+      "loss": 1.5008,
+      "step": 2397
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002945846386057478,
+      "loss": 1.3824,
+      "step": 2398
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002945018757501733,
+      "loss": 1.6322,
+      "step": 2399
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029441909205506835,
+      "loss": 1.5017,
+      "step": 2400
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029433628753868836,
+      "loss": 1.474,
+      "step": 2401
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002942534622192933,
+      "loss": 1.3641,
+      "step": 2402
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029417061611514773,
+      "loss": 1.5781,
+      "step": 2403
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029408774924452103,
+      "loss": 1.3302,
+      "step": 2404
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002940048616256868,
+      "loss": 1.59,
+      "step": 2405
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029392195327692365,
+      "loss": 1.4428,
+      "step": 2406
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002938390242165142,
+      "loss": 1.2823,
+      "step": 2407
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00029375607446274613,
+      "loss": 1.5121,
+      "step": 2408
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002936731040339115,
+      "loss": 1.2688,
+      "step": 2409
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002935901129483069,
+      "loss": 1.7354,
+      "step": 2410
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029350710122423346,
+      "loss": 1.5494,
+      "step": 2411
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002934240688799969,
+      "loss": 1.576,
+      "step": 2412
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002933410159339076,
+      "loss": 1.1983,
+      "step": 2413
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002932579424042803,
+      "loss": 1.3135,
+      "step": 2414
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002931748483094343,
+      "loss": 1.4291,
+      "step": 2415
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029309173366769364,
+      "loss": 1.5709,
+      "step": 2416
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002930085984973866,
+      "loss": 1.7591,
+      "step": 2417
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029292544281684627,
+      "loss": 1.5974,
+      "step": 2418
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002928422666444099,
+      "loss": 1.4479,
+      "step": 2419
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002927590699984196,
+      "loss": 1.5737,
+      "step": 2420
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002926758528972219,
+      "loss": 1.3459,
+      "step": 2421
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002925926153591679,
+      "loss": 1.508,
+      "step": 2422
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029250935740261297,
+      "loss": 1.4463,
+      "step": 2423
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029242607904591715,
+      "loss": 1.5676,
+      "step": 2424
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029234278030744503,
+      "loss": 1.4873,
+      "step": 2425
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029225946120556555,
+      "loss": 1.4342,
+      "step": 2426
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029217612175865237,
+      "loss": 1.5243,
+      "step": 2427
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002920927619850833,
+      "loss": 1.6278,
+      "step": 2428
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002920093819032409,
+      "loss": 1.3729,
+      "step": 2429
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029192598153151205,
+      "loss": 1.3402,
+      "step": 2430
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029184256088828834,
+      "loss": 1.3156,
+      "step": 2431
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002917591199919656,
+      "loss": 1.3156,
+      "step": 2432
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002916756588609441,
+      "loss": 1.4898,
+      "step": 2433
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029159217751362877,
+      "loss": 1.5613,
+      "step": 2434
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029150867596842886,
+      "loss": 1.2421,
+      "step": 2435
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002914251542437581,
+      "loss": 1.6308,
+      "step": 2436
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002913416123580347,
+      "loss": 1.3829,
+      "step": 2437
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029125805032968136,
+      "loss": 1.3955,
+      "step": 2438
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000291174468177125,
+      "loss": 1.7338,
+      "step": 2439
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029109086591879726,
+      "loss": 1.5109,
+      "step": 2440
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002910072435731341,
+      "loss": 1.3852,
+      "step": 2441
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029092360115857583,
+      "loss": 1.7356,
+      "step": 2442
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002908399386935673,
+      "loss": 1.5349,
+      "step": 2443
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002907562561965577,
+      "loss": 1.5205,
+      "step": 2444
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002906725536860007,
+      "loss": 1.5954,
+      "step": 2445
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029058883118035435,
+      "loss": 1.3615,
+      "step": 2446
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002905050886980811,
+      "loss": 1.5078,
+      "step": 2447
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029042132625764787,
+      "loss": 1.6185,
+      "step": 2448
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029033754387752586,
+      "loss": 1.7026,
+      "step": 2449
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029025374157619084,
+      "loss": 1.7131,
+      "step": 2450
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029016991937212276,
+      "loss": 1.4399,
+      "step": 2451
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00029008607728380615,
+      "loss": 1.54,
+      "step": 2452
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002900022153297298,
+      "loss": 1.5439,
+      "step": 2453
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00028991833352838694,
+      "loss": 1.4294,
+      "step": 2454
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002898344318982752,
+      "loss": 1.441,
+      "step": 2455
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002897505104578965,
+      "loss": 1.4576,
+      "step": 2456
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00028966656922575724,
+      "loss": 1.5748,
+      "step": 2457
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00028958260822036804,
+      "loss": 1.5387,
+      "step": 2458
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00028949862746024404,
+      "loss": 1.1022,
+      "step": 2459
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00028941462696390456,
+      "loss": 1.3808,
+      "step": 2460
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00028933060674987346,
+      "loss": 1.5776,
+      "step": 2461
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002892465668366788,
+      "loss": 1.5113,
+      "step": 2462
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000289162507242853,
+      "loss": 1.1891,
+      "step": 2463
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00028907842798693307,
+      "loss": 1.6181,
+      "step": 2464
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002889943290874599,
+      "loss": 1.4799,
+      "step": 2465
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002889102105629791,
+      "loss": 1.4167,
+      "step": 2466
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00028882607243204035,
+      "loss": 1.3898,
+      "step": 2467
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002887419147131979,
+      "loss": 1.6703,
+      "step": 2468
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002886577374250101,
+      "loss": 1.6146,
+      "step": 2469
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002885735405860398,
+      "loss": 1.7205,
+      "step": 2470
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000288489324214854,
+      "loss": 1.5312,
+      "step": 2471
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00028840508833002395,
+      "loss": 1.6053,
+      "step": 2472
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002883208329501256,
+      "loss": 1.5126,
+      "step": 2473
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00028823655809373874,
+      "loss": 1.6649,
+      "step": 2474
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002881522637794477,
+      "loss": 1.6248,
+      "step": 2475
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00028806795002584096,
+      "loss": 1.4772,
+      "step": 2476
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002879836168515115,
+      "loss": 1.6692,
+      "step": 2477
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002878992642750563,
+      "loss": 1.5597,
+      "step": 2478
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002878148923150769,
+      "loss": 1.5621,
+      "step": 2479
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000287730500990179,
+      "loss": 1.6634,
+      "step": 2480
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028764609031897237,
+      "loss": 1.3623,
+      "step": 2481
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028756166032007144,
+      "loss": 1.4824,
+      "step": 2482
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002874772110120946,
+      "loss": 1.472,
+      "step": 2483
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028739274241366454,
+      "loss": 1.6516,
+      "step": 2484
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028730825454340833,
+      "loss": 1.2968,
+      "step": 2485
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002872237474199572,
+      "loss": 1.388,
+      "step": 2486
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028713922106194664,
+      "loss": 1.4824,
+      "step": 2487
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002870546754880164,
+      "loss": 1.5735,
+      "step": 2488
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028697011071681034,
+      "loss": 1.488,
+      "step": 2489
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028688552676697684,
+      "loss": 1.5172,
+      "step": 2490
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002868009236571682,
+      "loss": 1.4412,
+      "step": 2491
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002867163014060411,
+      "loss": 1.6386,
+      "step": 2492
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028663166003225647,
+      "loss": 1.484,
+      "step": 2493
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002865469995544793,
+      "loss": 1.4138,
+      "step": 2494
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028646231999137907,
+      "loss": 1.6241,
+      "step": 2495
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028637762136162913,
+      "loss": 1.7344,
+      "step": 2496
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002862929036839073,
+      "loss": 1.6023,
+      "step": 2497
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002862081669768954,
+      "loss": 1.505,
+      "step": 2498
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028612341125927967,
+      "loss": 1.564,
+      "step": 2499
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028603863654975037,
+      "loss": 1.6483,
+      "step": 2500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002859538428670019,
+      "loss": 1.4305,
+      "step": 2501
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028586903022973317,
+      "loss": 1.3188,
+      "step": 2502
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002857841986566468,
+      "loss": 1.6356,
+      "step": 2503
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028569934816645003,
+      "loss": 1.5847,
+      "step": 2504
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002856144787778539,
+      "loss": 1.7294,
+      "step": 2505
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002855295905095738,
+      "loss": 1.3651,
+      "step": 2506
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028544468338032947,
+      "loss": 1.5273,
+      "step": 2507
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002853597574088444,
+      "loss": 1.4477,
+      "step": 2508
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028527481261384657,
+      "loss": 1.6621,
+      "step": 2509
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028518984901406785,
+      "loss": 1.5498,
+      "step": 2510
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002851048666282446,
+      "loss": 1.5196,
+      "step": 2511
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002850198654751169,
+      "loss": 1.6687,
+      "step": 2512
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028493484557342923,
+      "loss": 1.3898,
+      "step": 2513
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028484980694193026,
+      "loss": 1.5456,
+      "step": 2514
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028476474959937255,
+      "loss": 1.5464,
+      "step": 2515
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002846796735645131,
+      "loss": 1.3977,
+      "step": 2516
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002845945788561127,
+      "loss": 1.3938,
+      "step": 2517
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002845094654929365,
+      "loss": 1.5621,
+      "step": 2518
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028442433349375364,
+      "loss": 1.4181,
+      "step": 2519
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002843391828773374,
+      "loss": 1.4432,
+      "step": 2520
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002842540136624652,
+      "loss": 1.4452,
+      "step": 2521
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002841688258679185,
+      "loss": 1.2613,
+      "step": 2522
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028408361951248295,
+      "loss": 1.4646,
+      "step": 2523
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002839983946149482,
+      "loss": 1.477,
+      "step": 2524
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000283913151194108,
+      "loss": 1.3961,
+      "step": 2525
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028382788926876025,
+      "loss": 1.4218,
+      "step": 2526
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002837426088577069,
+      "loss": 1.5261,
+      "step": 2527
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002836573099797539,
+      "loss": 1.4221,
+      "step": 2528
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002835719926537114,
+      "loss": 1.3951,
+      "step": 2529
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028348665689839354,
+      "loss": 1.5949,
+      "step": 2530
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028340130273261845,
+      "loss": 1.4919,
+      "step": 2531
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002833159301752086,
+      "loss": 1.6357,
+      "step": 2532
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028323053924499017,
+      "loss": 1.5588,
+      "step": 2533
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028314512996079365,
+      "loss": 1.4364,
+      "step": 2534
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028305970234145334,
+      "loss": 1.4334,
+      "step": 2535
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002829742564058078,
+      "loss": 1.6371,
+      "step": 2536
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028288879217269953,
+      "loss": 1.4131,
+      "step": 2537
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002828033096609751,
+      "loss": 1.5459,
+      "step": 2538
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028271780888948505,
+      "loss": 1.4366,
+      "step": 2539
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028263228987708414,
+      "loss": 1.4225,
+      "step": 2540
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028254675264263084,
+      "loss": 1.739,
+      "step": 2541
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002824611972049878,
+      "loss": 1.6734,
+      "step": 2542
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00028237562358302183,
+      "loss": 1.5067,
+      "step": 2543
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002822900317956034,
+      "loss": 1.6574,
+      "step": 2544
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002822044218616073,
+      "loss": 1.3748,
+      "step": 2545
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002821187937999123,
+      "loss": 1.3674,
+      "step": 2546
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002820331476294009,
+      "loss": 1.3027,
+      "step": 2547
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002819474833689599,
+      "loss": 1.2885,
+      "step": 2548
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00028186180103747995,
+      "loss": 1.4986,
+      "step": 2549
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002817761006538556,
+      "loss": 1.5887,
+      "step": 2550
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002816903822369856,
+      "loss": 1.4527,
+      "step": 2551
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00028160464580577243,
+      "loss": 1.3555,
+      "step": 2552
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00028151889137912283,
+      "loss": 1.7434,
+      "step": 2553
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00028143311897594714,
+      "loss": 1.39,
+      "step": 2554
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002813473286151601,
+      "loss": 1.3773,
+      "step": 2555
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002812615203156799,
+      "loss": 1.4904,
+      "step": 2556
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002811756940964293,
+      "loss": 1.5703,
+      "step": 2557
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002810898499763344,
+      "loss": 1.5086,
+      "step": 2558
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00028100398797432566,
+      "loss": 1.2636,
+      "step": 2559
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002809181081093373,
+      "loss": 1.3517,
+      "step": 2560
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00028083221040030754,
+      "loss": 1.5268,
+      "step": 2561
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002807462948661786,
+      "loss": 1.6142,
+      "step": 2562
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00028066036152589636,
+      "loss": 1.5349,
+      "step": 2563
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000280574410398411,
+      "loss": 1.3886,
+      "step": 2564
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002804884415026764,
+      "loss": 1.5938,
+      "step": 2565
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00028040245485765037,
+      "loss": 1.5032,
+      "step": 2566
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002803164504822946,
+      "loss": 1.5284,
+      "step": 2567
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00028023042839557484,
+      "loss": 1.5676,
+      "step": 2568
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002801443886164607,
+      "loss": 1.5518,
+      "step": 2569
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00028005833116392557,
+      "loss": 1.2138,
+      "step": 2570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002799722560569468,
+      "loss": 1.4545,
+      "step": 2571
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027988616331450577,
+      "loss": 1.6086,
+      "step": 2572
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002798000529555875,
+      "loss": 1.224,
+      "step": 2573
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027971392499918105,
+      "loss": 1.2877,
+      "step": 2574
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027962777946427943,
+      "loss": 1.6333,
+      "step": 2575
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027954161636987936,
+      "loss": 1.4689,
+      "step": 2576
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002794554357349815,
+      "loss": 1.504,
+      "step": 2577
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002793692375785904,
+      "loss": 1.595,
+      "step": 2578
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002792830219197144,
+      "loss": 1.5339,
+      "step": 2579
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002791967887773659,
+      "loss": 1.3977,
+      "step": 2580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027911053817056085,
+      "loss": 1.516,
+      "step": 2581
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027902427011831937,
+      "loss": 1.7391,
+      "step": 2582
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002789379846396651,
+      "loss": 1.3659,
+      "step": 2583
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027885168175362577,
+      "loss": 1.4699,
+      "step": 2584
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002787653614792329,
+      "loss": 1.6575,
+      "step": 2585
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027867902383552177,
+      "loss": 1.5931,
+      "step": 2586
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002785926688415316,
+      "loss": 1.2967,
+      "step": 2587
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027850629651630525,
+      "loss": 1.5862,
+      "step": 2588
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027841990687888977,
+      "loss": 1.4884,
+      "step": 2589
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002783334999483354,
+      "loss": 1.4419,
+      "step": 2590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002782470757436969,
+      "loss": 1.2859,
+      "step": 2591
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002781606342840324,
+      "loss": 1.6416,
+      "step": 2592
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027807417558840396,
+      "loss": 1.6231,
+      "step": 2593
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002779876996758774,
+      "loss": 1.3836,
+      "step": 2594
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027790120656552244,
+      "loss": 1.5957,
+      "step": 2595
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002778146962764125,
+      "loss": 1.6744,
+      "step": 2596
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027772816882762466,
+      "loss": 1.4105,
+      "step": 2597
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002776416242382401,
+      "loss": 1.4718,
+      "step": 2598
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027755506252734357,
+      "loss": 1.4505,
+      "step": 2599
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002774684837140237,
+      "loss": 1.5522,
+      "step": 2600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027738188781737267,
+      "loss": 1.7029,
+      "step": 2601
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002772952748564867,
+      "loss": 1.4088,
+      "step": 2602
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027720864485046565,
+      "loss": 1.6674,
+      "step": 2603
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027712199781841304,
+      "loss": 1.8194,
+      "step": 2604
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027703533377943646,
+      "loss": 1.5644,
+      "step": 2605
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002769486527526469,
+      "loss": 1.6764,
+      "step": 2606
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002768619547571592,
+      "loss": 1.582,
+      "step": 2607
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002767752398120921,
+      "loss": 1.4928,
+      "step": 2608
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027668850793656795,
+      "loss": 1.478,
+      "step": 2609
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027660175914971273,
+      "loss": 1.6834,
+      "step": 2610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027651499347065633,
+      "loss": 1.4879,
+      "step": 2611
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027642821091853226,
+      "loss": 1.6324,
+      "step": 2612
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002763414115124779,
+      "loss": 1.6345,
+      "step": 2613
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00027625459527163415,
+      "loss": 1.5319,
+      "step": 2614
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027616776221514567,
+      "loss": 1.3872,
+      "step": 2615
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002760809123621609,
+      "loss": 1.5442,
+      "step": 2616
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027599404573183195,
+      "loss": 1.3617,
+      "step": 2617
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027590716234331467,
+      "loss": 1.3455,
+      "step": 2618
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027582026221576855,
+      "loss": 1.6445,
+      "step": 2619
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002757333453683567,
+      "loss": 1.4017,
+      "step": 2620
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002756464118202461,
+      "loss": 1.5529,
+      "step": 2621
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027555946159060724,
+      "loss": 1.3389,
+      "step": 2622
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027547249469861445,
+      "loss": 1.4108,
+      "step": 2623
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002753855111634455,
+      "loss": 1.3468,
+      "step": 2624
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002752985110042822,
+      "loss": 1.2975,
+      "step": 2625
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027521149424030957,
+      "loss": 1.7112,
+      "step": 2626
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027512446089071667,
+      "loss": 1.4632,
+      "step": 2627
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027503741097469606,
+      "loss": 1.5034,
+      "step": 2628
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027495034451144394,
+      "loss": 1.6292,
+      "step": 2629
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027486326152016026,
+      "loss": 1.5139,
+      "step": 2630
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002747761620200484,
+      "loss": 1.585,
+      "step": 2631
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002746890460303155,
+      "loss": 1.5687,
+      "step": 2632
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002746019135701726,
+      "loss": 1.5212,
+      "step": 2633
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027451476465883387,
+      "loss": 1.312,
+      "step": 2634
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002744275993155175,
+      "loss": 1.6583,
+      "step": 2635
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027434041755944517,
+      "loss": 1.2598,
+      "step": 2636
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027425321940984215,
+      "loss": 1.4949,
+      "step": 2637
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027416600488593736,
+      "loss": 1.4817,
+      "step": 2638
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002740787740069633,
+      "loss": 1.4369,
+      "step": 2639
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027399152679215613,
+      "loss": 1.6167,
+      "step": 2640
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027390426326075553,
+      "loss": 1.4814,
+      "step": 2641
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002738169834320049,
+      "loss": 1.5809,
+      "step": 2642
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002737296873251512,
+      "loss": 1.5743,
+      "step": 2643
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002736423749594449,
+      "loss": 1.5634,
+      "step": 2644
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027355504635414007,
+      "loss": 1.5987,
+      "step": 2645
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002734677015284944,
+      "loss": 1.4211,
+      "step": 2646
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027338034050176913,
+      "loss": 1.4136,
+      "step": 2647
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027329296329322927,
+      "loss": 1.4527,
+      "step": 2648
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002732055699221429,
+      "loss": 1.6497,
+      "step": 2649
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002731181604077823,
+      "loss": 1.5957,
+      "step": 2650
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027303073476942283,
+      "loss": 1.5047,
+      "step": 2651
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002729432930263435,
+      "loss": 1.4405,
+      "step": 2652
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002728558351978272,
+      "loss": 1.6964,
+      "step": 2653
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027276836130315986,
+      "loss": 1.4124,
+      "step": 2654
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027268087136163137,
+      "loss": 1.638,
+      "step": 2655
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002725933653925348,
+      "loss": 1.533,
+      "step": 2656
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002725058434151671,
+      "loss": 1.5882,
+      "step": 2657
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002724183054488285,
+      "loss": 1.6052,
+      "step": 2658
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000272330751512823,
+      "loss": 1.2839,
+      "step": 2659
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002722431816264579,
+      "loss": 1.4584,
+      "step": 2660
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000272155595809044,
+      "loss": 1.5835,
+      "step": 2661
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027206799407989573,
+      "loss": 1.2874,
+      "step": 2662
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002719803764583311,
+      "loss": 1.5926,
+      "step": 2663
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027189274296367147,
+      "loss": 1.4291,
+      "step": 2664
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002718050936152417,
+      "loss": 1.3938,
+      "step": 2665
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002717174284323703,
+      "loss": 1.5382,
+      "step": 2666
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027162974743438913,
+      "loss": 1.2609,
+      "step": 2667
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002715420506406336,
+      "loss": 1.4156,
+      "step": 2668
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002714543380704426,
+      "loss": 1.6345,
+      "step": 2669
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027136660974315835,
+      "loss": 1.5009,
+      "step": 2670
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002712788656781269,
+      "loss": 1.587,
+      "step": 2671
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027119110589469736,
+      "loss": 1.4627,
+      "step": 2672
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027110333041222256,
+      "loss": 1.8777,
+      "step": 2673
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002710155392500588,
+      "loss": 1.53,
+      "step": 2674
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027092773242756566,
+      "loss": 1.4698,
+      "step": 2675
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002708399099641063,
+      "loss": 1.5398,
+      "step": 2676
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027075207187904734,
+      "loss": 1.6513,
+      "step": 2677
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002706642181917588,
+      "loss": 1.5569,
+      "step": 2678
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027057634892161417,
+      "loss": 1.4985,
+      "step": 2679
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027048846408799036,
+      "loss": 1.584,
+      "step": 2680
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002704005637102677,
+      "loss": 1.5651,
+      "step": 2681
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00027031264780783,
+      "loss": 1.573,
+      "step": 2682
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00027022471640006433,
+      "loss": 1.7318,
+      "step": 2683
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00027013676950636144,
+      "loss": 1.2763,
+      "step": 2684
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00027004880714611524,
+      "loss": 1.6066,
+      "step": 2685
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026996082933872326,
+      "loss": 1.5111,
+      "step": 2686
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026987283610358637,
+      "loss": 1.7013,
+      "step": 2687
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002697848274601087,
+      "loss": 1.4031,
+      "step": 2688
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026969680342769795,
+      "loss": 1.547,
+      "step": 2689
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002696087640257652,
+      "loss": 1.5198,
+      "step": 2690
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002695207092737248,
+      "loss": 1.4506,
+      "step": 2691
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002694326391909946,
+      "loss": 1.5763,
+      "step": 2692
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002693445537969958,
+      "loss": 1.4048,
+      "step": 2693
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026925645311115303,
+      "loss": 1.4724,
+      "step": 2694
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002691683371528941,
+      "loss": 1.4396,
+      "step": 2695
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026908020594165044,
+      "loss": 1.4945,
+      "step": 2696
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002689920594968566,
+      "loss": 1.4541,
+      "step": 2697
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002689038978379508,
+      "loss": 1.5915,
+      "step": 2698
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002688157209843742,
+      "loss": 1.6012,
+      "step": 2699
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002687275289555717,
+      "loss": 1.4065,
+      "step": 2700
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002686393217709913,
+      "loss": 1.4273,
+      "step": 2701
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002685510994500845,
+      "loss": 1.4029,
+      "step": 2702
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000268462862012306,
+      "loss": 1.4447,
+      "step": 2703
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026837460947711395,
+      "loss": 1.6631,
+      "step": 2704
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026828634186396976,
+      "loss": 1.4509,
+      "step": 2705
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002681980591923381,
+      "loss": 1.4557,
+      "step": 2706
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026810976148168716,
+      "loss": 1.5418,
+      "step": 2707
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002680214487514883,
+      "loss": 1.4999,
+      "step": 2708
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002679331210212161,
+      "loss": 1.6502,
+      "step": 2709
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026784477831034883,
+      "loss": 1.5322,
+      "step": 2710
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002677564206383675,
+      "loss": 1.4772,
+      "step": 2711
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000267668048024757,
+      "loss": 1.4124,
+      "step": 2712
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026757966048900506,
+      "loss": 1.7345,
+      "step": 2713
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026749125805060293,
+      "loss": 1.642,
+      "step": 2714
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002674028407290451,
+      "loss": 1.5261,
+      "step": 2715
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002673144085438293,
+      "loss": 1.283,
+      "step": 2716
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026722596151445665,
+      "loss": 1.5188,
+      "step": 2717
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026713749966043133,
+      "loss": 1.3755,
+      "step": 2718
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026704902300126116,
+      "loss": 1.3268,
+      "step": 2719
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002669605315564567,
+      "loss": 1.3515,
+      "step": 2720
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002668720253455323,
+      "loss": 1.3937,
+      "step": 2721
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026678350438800514,
+      "loss": 1.507,
+      "step": 2722
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026669496870339597,
+      "loss": 1.4121,
+      "step": 2723
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026660641831122866,
+      "loss": 1.6155,
+      "step": 2724
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002665178532310303,
+      "loss": 1.4592,
+      "step": 2725
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002664292734823311,
+      "loss": 1.614,
+      "step": 2726
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002663406790846649,
+      "loss": 1.5052,
+      "step": 2727
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026625207005756833,
+      "loss": 1.5089,
+      "step": 2728
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002661634464205815,
+      "loss": 1.5977,
+      "step": 2729
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026607480819324766,
+      "loss": 1.5722,
+      "step": 2730
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002659861553951133,
+      "loss": 1.4461,
+      "step": 2731
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026589748804572804,
+      "loss": 1.4422,
+      "step": 2732
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026580880616464494,
+      "loss": 1.4643,
+      "step": 2733
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026572010977141987,
+      "loss": 1.5432,
+      "step": 2734
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002656313988856124,
+      "loss": 1.4908,
+      "step": 2735
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002655426735267848,
+      "loss": 1.5116,
+      "step": 2736
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000265453933714503,
+      "loss": 1.5444,
+      "step": 2737
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026536517946833564,
+      "loss": 1.4084,
+      "step": 2738
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026527641080785496,
+      "loss": 1.5277,
+      "step": 2739
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002651876277526362,
+      "loss": 1.5153,
+      "step": 2740
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002650988303222576,
+      "loss": 1.5749,
+      "step": 2741
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002650100185363009,
+      "loss": 1.6306,
+      "step": 2742
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002649211924143508,
+      "loss": 1.5822,
+      "step": 2743
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002648323519759953,
+      "loss": 1.5454,
+      "step": 2744
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026474349724082535,
+      "loss": 1.5726,
+      "step": 2745
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026465462822843533,
+      "loss": 1.4556,
+      "step": 2746
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002645657449584224,
+      "loss": 1.5019,
+      "step": 2747
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00026447684745038724,
+      "loss": 1.5357,
+      "step": 2748
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002643879357239335,
+      "loss": 1.5838,
+      "step": 2749
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026429900979866786,
+      "loss": 1.2753,
+      "step": 2750
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026421006969420035,
+      "loss": 1.5787,
+      "step": 2751
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026412111543014393,
+      "loss": 1.4022,
+      "step": 2752
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026403214702611495,
+      "loss": 1.5296,
+      "step": 2753
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002639431645017325,
+      "loss": 1.4026,
+      "step": 2754
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000263854167876619,
+      "loss": 1.5486,
+      "step": 2755
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026376515717040013,
+      "loss": 1.4526,
+      "step": 2756
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002636761324027043,
+      "loss": 1.3676,
+      "step": 2757
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026358709359316346,
+      "loss": 1.4442,
+      "step": 2758
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026349804076141223,
+      "loss": 1.4338,
+      "step": 2759
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002634089739270886,
+      "loss": 1.6938,
+      "step": 2760
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002633198931098335,
+      "loss": 1.3292,
+      "step": 2761
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026323079832929117,
+      "loss": 1.6125,
+      "step": 2762
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002631416896051086,
+      "loss": 1.545,
+      "step": 2763
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002630525669569361,
+      "loss": 1.2856,
+      "step": 2764
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026296343040442695,
+      "loss": 1.64,
+      "step": 2765
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026287427996723756,
+      "loss": 1.6693,
+      "step": 2766
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026278511566502733,
+      "loss": 1.461,
+      "step": 2767
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026269593751745876,
+      "loss": 1.4652,
+      "step": 2768
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002626067455441974,
+      "loss": 1.3957,
+      "step": 2769
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026251753976491184,
+      "loss": 1.5668,
+      "step": 2770
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026242832019927366,
+      "loss": 1.4126,
+      "step": 2771
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026233908686695763,
+      "loss": 1.5905,
+      "step": 2772
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002622498397876414,
+      "loss": 1.5971,
+      "step": 2773
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002621605789810058,
+      "loss": 1.4878,
+      "step": 2774
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026207130446673444,
+      "loss": 1.5474,
+      "step": 2775
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026198201626451427,
+      "loss": 1.4707,
+      "step": 2776
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000261892714394035,
+      "loss": 1.5486,
+      "step": 2777
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026180339887498953,
+      "loss": 1.4813,
+      "step": 2778
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002617140697270736,
+      "loss": 1.5958,
+      "step": 2779
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026162472696998623,
+      "loss": 1.6127,
+      "step": 2780
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026153537062342905,
+      "loss": 1.5037,
+      "step": 2781
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000261446000707107,
+      "loss": 1.761,
+      "step": 2782
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026135661724072795,
+      "loss": 1.5525,
+      "step": 2783
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026126722024400265,
+      "loss": 1.3675,
+      "step": 2784
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026117780973664506,
+      "loss": 1.6479,
+      "step": 2785
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026108838573837167,
+      "loss": 1.4746,
+      "step": 2786
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002609989482689025,
+      "loss": 1.5219,
+      "step": 2787
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026090949734796016,
+      "loss": 1.476,
+      "step": 2788
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002608200329952703,
+      "loss": 1.4395,
+      "step": 2789
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026073055523056176,
+      "loss": 1.3812,
+      "step": 2790
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000260641064073566,
+      "loss": 1.3467,
+      "step": 2791
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026055155954401765,
+      "loss": 1.423,
+      "step": 2792
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026046204166165415,
+      "loss": 1.3959,
+      "step": 2793
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00026037251044621604,
+      "loss": 1.47,
+      "step": 2794
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002602829659174467,
+      "loss": 1.4705,
+      "step": 2795
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002601934080950925,
+      "loss": 1.5466,
+      "step": 2796
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002601038369989026,
+      "loss": 1.408,
+      "step": 2797
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002600142526486293,
+      "loss": 1.6283,
+      "step": 2798
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002599246550640278,
+      "loss": 1.4465,
+      "step": 2799
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002598350442648559,
+      "loss": 1.5502,
+      "step": 2800
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00025974542027087477,
+      "loss": 1.4744,
+      "step": 2801
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00025965578310184814,
+      "loss": 1.3366,
+      "step": 2802
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00025956613277754284,
+      "loss": 1.4317,
+      "step": 2803
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00025947646931772847,
+      "loss": 1.2146,
+      "step": 2804
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00025938679274217763,
+      "loss": 1.5558,
+      "step": 2805
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00025929710307066585,
+      "loss": 1.3899,
+      "step": 2806
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00025920740032297135,
+      "loss": 1.6333,
+      "step": 2807
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00025911768451887556,
+      "loss": 1.3715,
+      "step": 2808
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002590279556781623,
+      "loss": 1.4512,
+      "step": 2809
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00025893821382061877,
+      "loss": 1.6156,
+      "step": 2810
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00025884845896603466,
+      "loss": 1.6794,
+      "step": 2811
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00025875869113420287,
+      "loss": 1.3133,
+      "step": 2812
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00025866891034491884,
+      "loss": 1.4726,
+      "step": 2813
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002585791166179811,
+      "loss": 1.5433,
+      "step": 2814
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00025848930997319087,
+      "loss": 1.4702,
+      "step": 2815
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002583994904303523,
+      "loss": 1.3795,
+      "step": 2816
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00025830965800927247,
+      "loss": 1.4205,
+      "step": 2817
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025821981272976106,
+      "loss": 1.6177,
+      "step": 2818
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025812995461163085,
+      "loss": 1.5877,
+      "step": 2819
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002580400836746971,
+      "loss": 1.5863,
+      "step": 2820
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002579501999387785,
+      "loss": 1.3888,
+      "step": 2821
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025786030342369595,
+      "loss": 1.7352,
+      "step": 2822
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002577703941492734,
+      "loss": 1.4447,
+      "step": 2823
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002576804721353377,
+      "loss": 1.4701,
+      "step": 2824
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002575905374017184,
+      "loss": 1.5681,
+      "step": 2825
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002575005899682479,
+      "loss": 1.3401,
+      "step": 2826
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002574106298547613,
+      "loss": 1.6369,
+      "step": 2827
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025732065708109663,
+      "loss": 1.4719,
+      "step": 2828
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002572306716670948,
+      "loss": 1.4069,
+      "step": 2829
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002571406736325992,
+      "loss": 1.7314,
+      "step": 2830
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025705066299745624,
+      "loss": 1.5916,
+      "step": 2831
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025696063978151507,
+      "loss": 1.5672,
+      "step": 2832
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025687060400462754,
+      "loss": 1.6966,
+      "step": 2833
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002567805556866483,
+      "loss": 1.3916,
+      "step": 2834
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025669049484743485,
+      "loss": 1.3723,
+      "step": 2835
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025660042150684733,
+      "loss": 1.425,
+      "step": 2836
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002565103356847487,
+      "loss": 1.4998,
+      "step": 2837
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002564202374010047,
+      "loss": 1.4664,
+      "step": 2838
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002563301266754837,
+      "loss": 1.5878,
+      "step": 2839
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025624000352805705,
+      "loss": 1.4324,
+      "step": 2840
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002561498679785985,
+      "loss": 1.4682,
+      "step": 2841
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002560597200469848,
+      "loss": 1.4051,
+      "step": 2842
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002559695597530954,
+      "loss": 1.5633,
+      "step": 2843
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002558793871168124,
+      "loss": 1.5929,
+      "step": 2844
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025578920215802055,
+      "loss": 1.3421,
+      "step": 2845
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025569900489660755,
+      "loss": 1.3859,
+      "step": 2846
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025560879535246363,
+      "loss": 1.4332,
+      "step": 2847
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002555185735454817,
+      "loss": 1.3254,
+      "step": 2848
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025542833949555757,
+      "loss": 1.3431,
+      "step": 2849
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025533809322258956,
+      "loss": 1.6145,
+      "step": 2850
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002552478347464788,
+      "loss": 1.59,
+      "step": 2851
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000255157564087129,
+      "loss": 1.4986,
+      "step": 2852
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002550672812644467,
+      "loss": 1.6337,
+      "step": 2853
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025497698629834103,
+      "loss": 1.6263,
+      "step": 2854
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025488667920872373,
+      "loss": 1.582,
+      "step": 2855
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002547963600155094,
+      "loss": 1.5368,
+      "step": 2856
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025470602873861516,
+      "loss": 1.5488,
+      "step": 2857
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025461568539796085,
+      "loss": 1.612,
+      "step": 2858
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000254525330013469,
+      "loss": 1.3632,
+      "step": 2859
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025443496260506467,
+      "loss": 1.6533,
+      "step": 2860
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025434458319267573,
+      "loss": 1.4145,
+      "step": 2861
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002542541917962326,
+      "loss": 1.4905,
+      "step": 2862
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025416378843566833,
+      "loss": 1.493,
+      "step": 2863
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002540733731309187,
+      "loss": 1.49,
+      "step": 2864
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002539829459019221,
+      "loss": 1.5801,
+      "step": 2865
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025389250676861945,
+      "loss": 1.4859,
+      "step": 2866
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025380205575095444,
+      "loss": 1.4817,
+      "step": 2867
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025371159286887324,
+      "loss": 1.4706,
+      "step": 2868
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002536211181423247,
+      "loss": 1.2345,
+      "step": 2869
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002535306315912603,
+      "loss": 1.3118,
+      "step": 2870
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002534401332356342,
+      "loss": 1.3877,
+      "step": 2871
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025334962309540295,
+      "loss": 1.3485,
+      "step": 2872
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025325910119052583,
+      "loss": 1.5984,
+      "step": 2873
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025316856754096476,
+      "loss": 1.4415,
+      "step": 2874
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025307802216668424,
+      "loss": 1.3927,
+      "step": 2875
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002529874650876512,
+      "loss": 1.48,
+      "step": 2876
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002528968963238354,
+      "loss": 1.672,
+      "step": 2877
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025280631589520887,
+      "loss": 1.5176,
+      "step": 2878
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025271572382174655,
+      "loss": 1.3632,
+      "step": 2879
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002526251201234256,
+      "loss": 1.3717,
+      "step": 2880
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002525345048202262,
+      "loss": 1.4668,
+      "step": 2881
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025244387793213057,
+      "loss": 1.5058,
+      "step": 2882
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025235323947912384,
+      "loss": 1.4759,
+      "step": 2883
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00025226258948119357,
+      "loss": 1.6033,
+      "step": 2884
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002521719279583298,
+      "loss": 1.7502,
+      "step": 2885
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002520812549305253,
+      "loss": 1.5793,
+      "step": 2886
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00025199057041777514,
+      "loss": 1.5823,
+      "step": 2887
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002518998744400772,
+      "loss": 1.425,
+      "step": 2888
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00025180916701743154,
+      "loss": 1.3789,
+      "step": 2889
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00025171844816984114,
+      "loss": 1.5889,
+      "step": 2890
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002516277179173112,
+      "loss": 1.4614,
+      "step": 2891
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002515369762798495,
+      "loss": 1.506,
+      "step": 2892
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002514462232774664,
+      "loss": 1.4821,
+      "step": 2893
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002513554589301747,
+      "loss": 1.4404,
+      "step": 2894
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00025126468325798986,
+      "loss": 1.4503,
+      "step": 2895
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002511738962809295,
+      "loss": 1.6367,
+      "step": 2896
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00025108309801901416,
+      "loss": 1.7446,
+      "step": 2897
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00025099228849226644,
+      "loss": 1.2629,
+      "step": 2898
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00025090146772071187,
+      "loss": 1.5745,
+      "step": 2899
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000250810635724378,
+      "loss": 1.527,
+      "step": 2900
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00025071979252329525,
+      "loss": 1.4988,
+      "step": 2901
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00025062893813749615,
+      "loss": 1.485,
+      "step": 2902
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00025053807258701606,
+      "loss": 1.4792,
+      "step": 2903
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002504471958918926,
+      "loss": 1.5652,
+      "step": 2904
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00025035630807216584,
+      "loss": 1.5682,
+      "step": 2905
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00025026540914787834,
+      "loss": 1.6872,
+      "step": 2906
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002501744991390751,
+      "loss": 1.3514,
+      "step": 2907
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002500835780658036,
+      "loss": 1.6145,
+      "step": 2908
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002499926459481137,
+      "loss": 1.6298,
+      "step": 2909
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002499017028060578,
+      "loss": 1.6117,
+      "step": 2910
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024981074865969055,
+      "loss": 1.5717,
+      "step": 2911
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024971978352906923,
+      "loss": 1.3773,
+      "step": 2912
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002496288074342534,
+      "loss": 1.6276,
+      "step": 2913
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024953782039530505,
+      "loss": 1.337,
+      "step": 2914
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024944682243228873,
+      "loss": 1.5051,
+      "step": 2915
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002493558135652712,
+      "loss": 1.6789,
+      "step": 2916
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002492647938143218,
+      "loss": 1.3524,
+      "step": 2917
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002491737631995121,
+      "loss": 1.4874,
+      "step": 2918
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002490827217409161,
+      "loss": 1.5069,
+      "step": 2919
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002489916694586104,
+      "loss": 1.3613,
+      "step": 2920
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002489006063726737,
+      "loss": 1.5336,
+      "step": 2921
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002488095325031873,
+      "loss": 1.5403,
+      "step": 2922
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024871844787023465,
+      "loss": 1.6249,
+      "step": 2923
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002486273524939018,
+      "loss": 1.3463,
+      "step": 2924
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024853624639427715,
+      "loss": 1.5874,
+      "step": 2925
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024844512959145126,
+      "loss": 1.4906,
+      "step": 2926
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002483540021055173,
+      "loss": 1.6929,
+      "step": 2927
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002482628639565706,
+      "loss": 1.4503,
+      "step": 2928
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024817171516470897,
+      "loss": 1.5122,
+      "step": 2929
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024808055575003244,
+      "loss": 1.4681,
+      "step": 2930
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002479893857326436,
+      "loss": 1.4064,
+      "step": 2931
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002478982051326471,
+      "loss": 1.6629,
+      "step": 2932
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002478070139701502,
+      "loss": 1.5373,
+      "step": 2933
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024771581226526224,
+      "loss": 1.4987,
+      "step": 2934
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000247624600038095,
+      "loss": 1.3602,
+      "step": 2935
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024753337730876266,
+      "loss": 1.5095,
+      "step": 2936
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024744214409738155,
+      "loss": 1.3807,
+      "step": 2937
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024735090042407045,
+      "loss": 1.4269,
+      "step": 2938
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002472596463089504,
+      "loss": 1.264,
+      "step": 2939
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024716838177214474,
+      "loss": 1.3385,
+      "step": 2940
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024707710683377897,
+      "loss": 1.5646,
+      "step": 2941
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024698582151398116,
+      "loss": 1.4356,
+      "step": 2942
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002468945258328815,
+      "loss": 1.4707,
+      "step": 2943
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002468032198106125,
+      "loss": 1.7866,
+      "step": 2944
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024671190346730893,
+      "loss": 1.5282,
+      "step": 2945
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002466205768231078,
+      "loss": 1.4584,
+      "step": 2946
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002465292398981485,
+      "loss": 1.6537,
+      "step": 2947
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002464378927125725,
+      "loss": 1.4909,
+      "step": 2948
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002463465352865239,
+      "loss": 1.28,
+      "step": 2949
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024625516764014864,
+      "loss": 1.34,
+      "step": 2950
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024616378979359514,
+      "loss": 1.3959,
+      "step": 2951
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000246072401767014,
+      "loss": 1.423,
+      "step": 2952
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00024598100358055804,
+      "loss": 1.443,
+      "step": 2953
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002458895952543825,
+      "loss": 1.529,
+      "step": 2954
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024579817680864464,
+      "loss": 1.7918,
+      "step": 2955
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024570674826350405,
+      "loss": 1.5292,
+      "step": 2956
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002456153096391225,
+      "loss": 1.536,
+      "step": 2957
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024552386095566395,
+      "loss": 1.6146,
+      "step": 2958
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002454324022332948,
+      "loss": 1.4089,
+      "step": 2959
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024534093349218343,
+      "loss": 1.333,
+      "step": 2960
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024524945475250053,
+      "loss": 1.4556,
+      "step": 2961
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002451579660344188,
+      "loss": 1.4299,
+      "step": 2962
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024506646735811364,
+      "loss": 1.5721,
+      "step": 2963
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024497495874376197,
+      "loss": 1.5875,
+      "step": 2964
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002448834402115435,
+      "loss": 1.5921,
+      "step": 2965
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002447919117816397,
+      "loss": 1.3663,
+      "step": 2966
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002447003734742346,
+      "loss": 1.7381,
+      "step": 2967
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000244608825309514,
+      "loss": 1.7021,
+      "step": 2968
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002445172673076661,
+      "loss": 1.6846,
+      "step": 2969
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002444256994888814,
+      "loss": 1.4155,
+      "step": 2970
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024433412187335225,
+      "loss": 1.4561,
+      "step": 2971
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002442425344812735,
+      "loss": 1.636,
+      "step": 2972
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002441509373328418,
+      "loss": 1.6237,
+      "step": 2973
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024405933044825622,
+      "loss": 1.5575,
+      "step": 2974
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024396771384771783,
+      "loss": 1.4129,
+      "step": 2975
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024387608755142997,
+      "loss": 1.0656,
+      "step": 2976
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024378445157959805,
+      "loss": 1.3194,
+      "step": 2977
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024369280595242955,
+      "loss": 1.1741,
+      "step": 2978
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002436011506901342,
+      "loss": 1.5629,
+      "step": 2979
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024350948581292373,
+      "loss": 1.4346,
+      "step": 2980
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024341781134101207,
+      "loss": 1.6236,
+      "step": 2981
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002433261272946153,
+      "loss": 1.6913,
+      "step": 2982
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002432344336939516,
+      "loss": 1.4402,
+      "step": 2983
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024314273055924112,
+      "loss": 1.5704,
+      "step": 2984
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024305101791070624,
+      "loss": 1.4268,
+      "step": 2985
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024295929576857148,
+      "loss": 1.4291,
+      "step": 2986
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024286756415306329,
+      "loss": 1.229,
+      "step": 2987
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024277582308441034,
+      "loss": 1.4693,
+      "step": 2988
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024268407258284334,
+      "loss": 1.4342,
+      "step": 2989
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024259231266859517,
+      "loss": 1.4482,
+      "step": 2990
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024250054336190054,
+      "loss": 1.516,
+      "step": 2991
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024240876468299653,
+      "loss": 1.493,
+      "step": 2992
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024231697665212215,
+      "loss": 1.5149,
+      "step": 2993
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024222517928951836,
+      "loss": 1.5165,
+      "step": 2994
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002421333726154284,
+      "loss": 1.3016,
+      "step": 2995
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024204155665009746,
+      "loss": 1.3659,
+      "step": 2996
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024194973141377268,
+      "loss": 1.4883,
+      "step": 2997
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024185789692670334,
+      "loss": 1.335,
+      "step": 2998
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024176605320914095,
+      "loss": 1.497,
+      "step": 2999
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024167420028133862,
+      "loss": 1.4529,
+      "step": 3000
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024158233816355185,
+      "loss": 1.4569,
+      "step": 3001
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024149046687603812,
+      "loss": 1.5477,
+      "step": 3002
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002413985864390567,
+      "loss": 1.5344,
+      "step": 3003
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024130669687286922,
+      "loss": 1.3418,
+      "step": 3004
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024121479819773898,
+      "loss": 1.3364,
+      "step": 3005
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024112289043393158,
+      "loss": 1.4736,
+      "step": 3006
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002410309736017144,
+      "loss": 1.3519,
+      "step": 3007
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024093904772135697,
+      "loss": 1.4704,
+      "step": 3008
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002408471128131308,
+      "loss": 1.4928,
+      "step": 3009
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024075516889730927,
+      "loss": 1.5044,
+      "step": 3010
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024066321599416787,
+      "loss": 1.6997,
+      "step": 3011
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000240571254123984,
+      "loss": 1.7503,
+      "step": 3012
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024047928330703716,
+      "loss": 1.2193,
+      "step": 3013
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002403873035636086,
+      "loss": 1.5995,
+      "step": 3014
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024029531491398182,
+      "loss": 1.4333,
+      "step": 3015
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024020331737844197,
+      "loss": 1.6454,
+      "step": 3016
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024011131097727645,
+      "loss": 1.4701,
+      "step": 3017
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00024001929573077448,
+      "loss": 1.5253,
+      "step": 3018
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023992727165922717,
+      "loss": 1.458,
+      "step": 3019
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023983523878292774,
+      "loss": 1.4487,
+      "step": 3020
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00023974319712217112,
+      "loss": 1.2825,
+      "step": 3021
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023965114669725445,
+      "loss": 1.6549,
+      "step": 3022
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002395590875284766,
+      "loss": 1.4405,
+      "step": 3023
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023946701963613843,
+      "loss": 1.5857,
+      "step": 3024
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002393749430405428,
+      "loss": 1.7165,
+      "step": 3025
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023928285776199433,
+      "loss": 1.5218,
+      "step": 3026
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002391907638207997,
+      "loss": 1.5721,
+      "step": 3027
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002390986612372674,
+      "loss": 1.7981,
+      "step": 3028
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002390065500317079,
+      "loss": 1.4553,
+      "step": 3029
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023891443022443354,
+      "loss": 1.3479,
+      "step": 3030
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023882230183575848,
+      "loss": 1.3771,
+      "step": 3031
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023873016488599896,
+      "loss": 1.4485,
+      "step": 3032
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023863801939547302,
+      "loss": 1.5249,
+      "step": 3033
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023854586538450044,
+      "loss": 1.4589,
+      "step": 3034
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023845370287340302,
+      "loss": 1.4931,
+      "step": 3035
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023836153188250454,
+      "loss": 1.4388,
+      "step": 3036
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002382693524321304,
+      "loss": 1.5325,
+      "step": 3037
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023817716454260803,
+      "loss": 1.4863,
+      "step": 3038
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002380849682342667,
+      "loss": 1.4181,
+      "step": 3039
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023799276352743745,
+      "loss": 1.4422,
+      "step": 3040
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023790055044245332,
+      "loss": 1.1962,
+      "step": 3041
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023780832899964903,
+      "loss": 1.555,
+      "step": 3042
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023771609921936136,
+      "loss": 1.3105,
+      "step": 3043
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023762386112192866,
+      "loss": 1.5665,
+      "step": 3044
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023753161472769136,
+      "loss": 1.6334,
+      "step": 3045
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023743936005699153,
+      "loss": 1.4255,
+      "step": 3046
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023734709713017313,
+      "loss": 1.7019,
+      "step": 3047
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023725482596758207,
+      "loss": 1.5316,
+      "step": 3048
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002371625465895659,
+      "loss": 1.4862,
+      "step": 3049
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023707025901647405,
+      "loss": 1.583,
+      "step": 3050
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023697796326865774,
+      "loss": 1.4929,
+      "step": 3051
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023688565936647003,
+      "loss": 1.4452,
+      "step": 3052
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023679334733026574,
+      "loss": 1.3471,
+      "step": 3053
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023670102718040153,
+      "loss": 1.5554,
+      "step": 3054
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023660869893723577,
+      "loss": 1.3459,
+      "step": 3055
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002365163626211287,
+      "loss": 1.3198,
+      "step": 3056
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002364240182524422,
+      "loss": 1.4149,
+      "step": 3057
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023633166585154016,
+      "loss": 1.6073,
+      "step": 3058
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023623930543878804,
+      "loss": 1.6924,
+      "step": 3059
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023614693703455318,
+      "loss": 1.4186,
+      "step": 3060
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002360545606592046,
+      "loss": 1.6329,
+      "step": 3061
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023596217633311314,
+      "loss": 1.4042,
+      "step": 3062
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023586978407665136,
+      "loss": 1.3261,
+      "step": 3063
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023577738391019354,
+      "loss": 1.5133,
+      "step": 3064
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023568497585411575,
+      "loss": 1.9005,
+      "step": 3065
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023559255992879586,
+      "loss": 1.6354,
+      "step": 3066
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023550013615461333,
+      "loss": 1.4112,
+      "step": 3067
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023540770455194953,
+      "loss": 1.6054,
+      "step": 3068
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023531526514118735,
+      "loss": 1.2967,
+      "step": 3069
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023522281794271157,
+      "loss": 1.5934,
+      "step": 3070
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023513036297690855,
+      "loss": 1.5701,
+      "step": 3071
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002350379002641665,
+      "loss": 1.5509,
+      "step": 3072
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023494542982487526,
+      "loss": 1.5649,
+      "step": 3073
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023485295167942637,
+      "loss": 1.5281,
+      "step": 3074
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023476046584821315,
+      "loss": 1.4804,
+      "step": 3075
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023466797235163048,
+      "loss": 1.4409,
+      "step": 3076
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023457547121007502,
+      "loss": 1.4703,
+      "step": 3077
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023448296244394512,
+      "loss": 1.7276,
+      "step": 3078
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002343904460736408,
+      "loss": 1.4046,
+      "step": 3079
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023429792211956371,
+      "loss": 1.5086,
+      "step": 3080
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002342053906021172,
+      "loss": 1.5308,
+      "step": 3081
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002341128515417063,
+      "loss": 1.7097,
+      "step": 3082
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023402030495873777,
+      "loss": 1.3213,
+      "step": 3083
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023392775087361997,
+      "loss": 1.1396,
+      "step": 3084
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023383518930676278,
+      "loss": 1.4084,
+      "step": 3085
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023374262027857796,
+      "loss": 1.631,
+      "step": 3086
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023365004380947876,
+      "loss": 1.4478,
+      "step": 3087
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023355745991988016,
+      "loss": 1.444,
+      "step": 3088
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00023346486863019865,
+      "loss": 1.6286,
+      "step": 3089
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023337226996085253,
+      "loss": 1.5801,
+      "step": 3090
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023327966393226163,
+      "loss": 1.6437,
+      "step": 3091
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023318705056484733,
+      "loss": 1.5551,
+      "step": 3092
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023309442987903283,
+      "loss": 1.4584,
+      "step": 3093
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0002330018018952427,
+      "loss": 1.4203,
+      "step": 3094
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0002329091666339033,
+      "loss": 1.4113,
+      "step": 3095
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023281652411544254,
+      "loss": 1.5075,
+      "step": 3096
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023272387436028987,
+      "loss": 1.3837,
+      "step": 3097
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0002326312173888765,
+      "loss": 1.6381,
+      "step": 3098
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023253855322163497,
+      "loss": 1.5944,
+      "step": 3099
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023244588187899972,
+      "loss": 1.4256,
+      "step": 3100
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023235320338140653,
+      "loss": 1.4845,
+      "step": 3101
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023226051774929282,
+      "loss": 1.6219,
+      "step": 3102
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023216782500309763,
+      "loss": 1.4123,
+      "step": 3103
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023207512516326158,
+      "loss": 1.5358,
+      "step": 3104
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023198241825022673,
+      "loss": 1.5049,
+      "step": 3105
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023188970428443688,
+      "loss": 1.517,
+      "step": 3106
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023179698328633724,
+      "loss": 1.3604,
+      "step": 3107
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023170425527637459,
+      "loss": 1.5009,
+      "step": 3108
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023161152027499738,
+      "loss": 1.5948,
+      "step": 3109
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023151877830265537,
+      "loss": 1.5878,
+      "step": 3110
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0002314260293798002,
+      "loss": 1.6131,
+      "step": 3111
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023133327352688464,
+      "loss": 1.4467,
+      "step": 3112
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023124051076436325,
+      "loss": 1.526,
+      "step": 3113
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023114774111269212,
+      "loss": 1.5473,
+      "step": 3114
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023105496459232874,
+      "loss": 1.5548,
+      "step": 3115
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0002309621812237321,
+      "loss": 1.7391,
+      "step": 3116
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0002308693910273629,
+      "loss": 1.6306,
+      "step": 3117
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023077659402368314,
+      "loss": 1.5498,
+      "step": 3118
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023068379023315636,
+      "loss": 1.5011,
+      "step": 3119
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023059097967624773,
+      "loss": 1.6496,
+      "step": 3120
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0002304981623734237,
+      "loss": 1.6473,
+      "step": 3121
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023040533834515237,
+      "loss": 1.2563,
+      "step": 3122
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023031250761190332,
+      "loss": 1.5476,
+      "step": 3123
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023021967019414742,
+      "loss": 1.3671,
+      "step": 3124
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023012682611235735,
+      "loss": 1.6138,
+      "step": 3125
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00023003397538700688,
+      "loss": 1.6092,
+      "step": 3126
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022994111803857154,
+      "loss": 1.5879,
+      "step": 3127
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022984825408752814,
+      "loss": 1.6298,
+      "step": 3128
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022975538355435505,
+      "loss": 1.3748,
+      "step": 3129
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022966250645953213,
+      "loss": 1.4367,
+      "step": 3130
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0002295696228235405,
+      "loss": 1.3233,
+      "step": 3131
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022947673266686289,
+      "loss": 1.4211,
+      "step": 3132
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022938383600998332,
+      "loss": 1.5755,
+      "step": 3133
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022929093287338756,
+      "loss": 1.3472,
+      "step": 3134
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022919802327756234,
+      "loss": 1.4034,
+      "step": 3135
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022910510724299629,
+      "loss": 1.4775,
+      "step": 3136
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022901218479017902,
+      "loss": 1.6916,
+      "step": 3137
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022891925593960193,
+      "loss": 1.1347,
+      "step": 3138
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022882632071175753,
+      "loss": 1.3257,
+      "step": 3139
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022873337912713995,
+      "loss": 1.4052,
+      "step": 3140
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022864043120624463,
+      "loss": 1.685,
+      "step": 3141
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022854747696956845,
+      "loss": 1.66,
+      "step": 3142
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022845451643760964,
+      "loss": 1.3637,
+      "step": 3143
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022836154963086784,
+      "loss": 1.3709,
+      "step": 3144
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022826857656984406,
+      "loss": 1.2202,
+      "step": 3145
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022817559727504065,
+      "loss": 1.4279,
+      "step": 3146
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022808261176696145,
+      "loss": 1.5805,
+      "step": 3147
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022798962006611158,
+      "loss": 1.5588,
+      "step": 3148
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0002278966221929975,
+      "loss": 1.5153,
+      "step": 3149
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022780361816812717,
+      "loss": 1.6208,
+      "step": 3150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022771060801200972,
+      "loss": 1.3498,
+      "step": 3151
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022761759174515583,
+      "loss": 1.4927,
+      "step": 3152
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0002275245693880773,
+      "loss": 1.3718,
+      "step": 3153
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022743154096128754,
+      "loss": 1.4322,
+      "step": 3154
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022733850648530104,
+      "loss": 1.3819,
+      "step": 3155
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022724546598063374,
+      "loss": 1.4085,
+      "step": 3156
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00022715241946780302,
+      "loss": 1.4947,
+      "step": 3157
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022705936696732737,
+      "loss": 1.498,
+      "step": 3158
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022696630849972676,
+      "loss": 1.3292,
+      "step": 3159
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022687324408552235,
+      "loss": 1.5059,
+      "step": 3160
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002267801737452368,
+      "loss": 1.2829,
+      "step": 3161
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002266870974993938,
+      "loss": 1.438,
+      "step": 3162
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022659401536851865,
+      "loss": 1.3945,
+      "step": 3163
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022650092737313773,
+      "loss": 1.3503,
+      "step": 3164
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022640783353377879,
+      "loss": 1.4985,
+      "step": 3165
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022631473387097088,
+      "loss": 1.4992,
+      "step": 3166
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002262216284052442,
+      "loss": 1.4936,
+      "step": 3167
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022612851715713057,
+      "loss": 1.4554,
+      "step": 3168
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022603540014716264,
+      "loss": 1.6564,
+      "step": 3169
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022594227739587472,
+      "loss": 1.6251,
+      "step": 3170
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022584914892380205,
+      "loss": 1.2451,
+      "step": 3171
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022575601475148145,
+      "loss": 1.5377,
+      "step": 3172
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002256628748994508,
+      "loss": 1.5124,
+      "step": 3173
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002255697293882492,
+      "loss": 1.5247,
+      "step": 3174
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022547657823841717,
+      "loss": 1.4694,
+      "step": 3175
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022538342147049631,
+      "loss": 1.3997,
+      "step": 3176
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022529025910502957,
+      "loss": 1.3937,
+      "step": 3177
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022519709116256107,
+      "loss": 1.3958,
+      "step": 3178
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002251039176636362,
+      "loss": 1.3564,
+      "step": 3179
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022501073862880158,
+      "loss": 1.71,
+      "step": 3180
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022491755407860495,
+      "loss": 1.5089,
+      "step": 3181
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022482436403359546,
+      "loss": 1.3593,
+      "step": 3182
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002247311685143232,
+      "loss": 1.424,
+      "step": 3183
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022463796754133974,
+      "loss": 1.4532,
+      "step": 3184
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002245447611351977,
+      "loss": 1.4238,
+      "step": 3185
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000224451549316451,
+      "loss": 1.3488,
+      "step": 3186
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022435833210565455,
+      "loss": 1.3641,
+      "step": 3187
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022426510952336474,
+      "loss": 1.521,
+      "step": 3188
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002241718815901389,
+      "loss": 1.5656,
+      "step": 3189
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022407864832653558,
+      "loss": 1.5815,
+      "step": 3190
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022398540975311472,
+      "loss": 1.2533,
+      "step": 3191
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022389216589043708,
+      "loss": 1.2384,
+      "step": 3192
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022379891675906492,
+      "loss": 1.452,
+      "step": 3193
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022370566237956142,
+      "loss": 1.3144,
+      "step": 3194
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022361240277249107,
+      "loss": 1.448,
+      "step": 3195
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022351913795841947,
+      "loss": 1.6691,
+      "step": 3196
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002234258679579133,
+      "loss": 1.4887,
+      "step": 3197
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002233325927915405,
+      "loss": 1.4186,
+      "step": 3198
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022323931247986996,
+      "loss": 1.4299,
+      "step": 3199
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022314602704347203,
+      "loss": 1.3309,
+      "step": 3200
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022305273650291777,
+      "loss": 1.5214,
+      "step": 3201
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022295944087877975,
+      "loss": 1.4141,
+      "step": 3202
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022286614019163145,
+      "loss": 1.6604,
+      "step": 3203
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022277283446204744,
+      "loss": 1.6776,
+      "step": 3204
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022267952371060362,
+      "loss": 1.5363,
+      "step": 3205
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002225862079578768,
+      "loss": 1.5241,
+      "step": 3206
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002224928872244449,
+      "loss": 1.326,
+      "step": 3207
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022239956153088702,
+      "loss": 1.4701,
+      "step": 3208
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022230623089778328,
+      "loss": 1.4015,
+      "step": 3209
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000222212895345715,
+      "loss": 1.5758,
+      "step": 3210
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022211955489526443,
+      "loss": 1.5139,
+      "step": 3211
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002220262095670151,
+      "loss": 1.4665,
+      "step": 3212
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022193285938155137,
+      "loss": 1.6276,
+      "step": 3213
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022183950435945886,
+      "loss": 1.8528,
+      "step": 3214
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002217461445213242,
+      "loss": 1.4858,
+      "step": 3215
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022165277988773512,
+      "loss": 1.5668,
+      "step": 3216
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022155941047928023,
+      "loss": 1.5537,
+      "step": 3217
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022146603631654946,
+      "loss": 1.487,
+      "step": 3218
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022137265742013358,
+      "loss": 1.6238,
+      "step": 3219
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002212792738106246,
+      "loss": 1.373,
+      "step": 3220
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002211858855086153,
+      "loss": 1.6258,
+      "step": 3221
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022109249253469967,
+      "loss": 1.5688,
+      "step": 3222
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022099909490947285,
+      "loss": 1.3963,
+      "step": 3223
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022090569265353072,
+      "loss": 1.6596,
+      "step": 3224
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00022081228578747038,
+      "loss": 1.3374,
+      "step": 3225
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002207188743318898,
+      "loss": 1.5858,
+      "step": 3226
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002206254583073882,
+      "loss": 1.3095,
+      "step": 3227
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022053203773456556,
+      "loss": 1.3889,
+      "step": 3228
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022043861263402296,
+      "loss": 1.4067,
+      "step": 3229
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022034518302636253,
+      "loss": 1.5934,
+      "step": 3230
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022025174893218727,
+      "loss": 1.439,
+      "step": 3231
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022015831037210134,
+      "loss": 1.4352,
+      "step": 3232
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00022006486736670975,
+      "loss": 1.4913,
+      "step": 3233
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021997141993661848,
+      "loss": 1.3519,
+      "step": 3234
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021987796810243456,
+      "loss": 1.3736,
+      "step": 3235
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021978451188476604,
+      "loss": 1.5752,
+      "step": 3236
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021969105130422177,
+      "loss": 1.4871,
+      "step": 3237
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002195975863814116,
+      "loss": 1.6028,
+      "step": 3238
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002195041171369466,
+      "loss": 1.4483,
+      "step": 3239
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021941064359143843,
+      "loss": 1.3213,
+      "step": 3240
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002193171657654999,
+      "loss": 1.2323,
+      "step": 3241
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021922368367974465,
+      "loss": 1.6199,
+      "step": 3242
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021913019735478746,
+      "loss": 1.5917,
+      "step": 3243
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021903670681124374,
+      "loss": 1.5713,
+      "step": 3244
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002189432120697301,
+      "loss": 1.3313,
+      "step": 3245
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021884971315086404,
+      "loss": 1.4835,
+      "step": 3246
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021875621007526382,
+      "loss": 1.3442,
+      "step": 3247
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021866270286354875,
+      "loss": 1.3806,
+      "step": 3248
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021856919153633895,
+      "loss": 1.4992,
+      "step": 3249
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021847567611425569,
+      "loss": 1.1651,
+      "step": 3250
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021838215661792076,
+      "loss": 1.5147,
+      "step": 3251
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021828863306795713,
+      "loss": 1.495,
+      "step": 3252
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021819510548498866,
+      "loss": 1.6071,
+      "step": 3253
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021810157388963997,
+      "loss": 1.3831,
+      "step": 3254
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021800803830253666,
+      "loss": 1.3828,
+      "step": 3255
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002179144987443051,
+      "loss": 1.2186,
+      "step": 3256
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021782095523557274,
+      "loss": 1.4267,
+      "step": 3257
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021772740779696763,
+      "loss": 1.5035,
+      "step": 3258
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021763385644911894,
+      "loss": 1.5715,
+      "step": 3259
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002175403012126565,
+      "loss": 1.2847,
+      "step": 3260
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002174467421082111,
+      "loss": 1.351,
+      "step": 3261
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021735317915641447,
+      "loss": 1.566,
+      "step": 3262
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000217259612377899,
+      "loss": 1.4017,
+      "step": 3263
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000217166041793298,
+      "loss": 1.4468,
+      "step": 3264
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021707246742324568,
+      "loss": 1.5444,
+      "step": 3265
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021697888928837706,
+      "loss": 1.511,
+      "step": 3266
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021688530740932786,
+      "loss": 1.2569,
+      "step": 3267
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021679172180673483,
+      "loss": 1.6451,
+      "step": 3268
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021669813250123538,
+      "loss": 1.6356,
+      "step": 3269
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021660453951346788,
+      "loss": 1.4577,
+      "step": 3270
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021651094286407143,
+      "loss": 1.3282,
+      "step": 3271
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021641734257368583,
+      "loss": 1.5421,
+      "step": 3272
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002163237386629519,
+      "loss": 1.3798,
+      "step": 3273
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002162301311525111,
+      "loss": 1.225,
+      "step": 3274
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002161365200630058,
+      "loss": 1.66,
+      "step": 3275
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000216042905415079,
+      "loss": 1.3666,
+      "step": 3276
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002159492872293746,
+      "loss": 1.3543,
+      "step": 3277
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021585566552653732,
+      "loss": 1.5218,
+      "step": 3278
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021576204032721253,
+      "loss": 1.6022,
+      "step": 3279
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002156684116520465,
+      "loss": 1.3646,
+      "step": 3280
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021557477952168617,
+      "loss": 1.484,
+      "step": 3281
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021548114395677926,
+      "loss": 1.5282,
+      "step": 3282
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021538750497797425,
+      "loss": 1.437,
+      "step": 3283
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002152938626059204,
+      "loss": 1.4393,
+      "step": 3284
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002152002168612677,
+      "loss": 1.5964,
+      "step": 3285
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021510656776466692,
+      "loss": 1.3314,
+      "step": 3286
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021501291533676954,
+      "loss": 1.4291,
+      "step": 3287
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021491925959822773,
+      "loss": 1.5728,
+      "step": 3288
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021482560056969443,
+      "loss": 1.4993,
+      "step": 3289
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021473193827182332,
+      "loss": 1.3302,
+      "step": 3290
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021463827272526884,
+      "loss": 1.7345,
+      "step": 3291
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021454460395068598,
+      "loss": 1.3079,
+      "step": 3292
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00021445093196873058,
+      "loss": 1.5179,
+      "step": 3293
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002143572568000593,
+      "loss": 1.443,
+      "step": 3294
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002142635784653292,
+      "loss": 1.3913,
+      "step": 3295
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002141698969851983,
+      "loss": 1.3597,
+      "step": 3296
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002140762123803252,
+      "loss": 1.6395,
+      "step": 3297
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021398252467136913,
+      "loss": 1.6796,
+      "step": 3298
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021388883387899016,
+      "loss": 1.5334,
+      "step": 3299
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021379514002384893,
+      "loss": 1.4029,
+      "step": 3300
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021370144312660684,
+      "loss": 1.5003,
+      "step": 3301
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021360774320792586,
+      "loss": 1.4076,
+      "step": 3302
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021351404028846865,
+      "loss": 1.4523,
+      "step": 3303
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021342033438889857,
+      "loss": 1.5187,
+      "step": 3304
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021332662552987967,
+      "loss": 1.0828,
+      "step": 3305
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021323291373207654,
+      "loss": 1.3643,
+      "step": 3306
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021313919901615457,
+      "loss": 1.2249,
+      "step": 3307
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021304548140277957,
+      "loss": 1.6569,
+      "step": 3308
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021295176091261827,
+      "loss": 1.6204,
+      "step": 3309
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021285803756633778,
+      "loss": 1.4242,
+      "step": 3310
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021276431138460598,
+      "loss": 1.4654,
+      "step": 3311
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002126705823880914,
+      "loss": 1.5498,
+      "step": 3312
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021257685059746306,
+      "loss": 1.3662,
+      "step": 3313
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021248311603339076,
+      "loss": 1.4937,
+      "step": 3314
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002123893787165447,
+      "loss": 1.4215,
+      "step": 3315
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021229563866759593,
+      "loss": 1.4272,
+      "step": 3316
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002122018959072159,
+      "loss": 1.243,
+      "step": 3317
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021210815045607677,
+      "loss": 1.3493,
+      "step": 3318
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002120144023348513,
+      "loss": 1.352,
+      "step": 3319
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021192065156421265,
+      "loss": 1.371,
+      "step": 3320
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021182689816483493,
+      "loss": 1.4649,
+      "step": 3321
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021173314215739252,
+      "loss": 1.4334,
+      "step": 3322
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021163938356256047,
+      "loss": 1.4118,
+      "step": 3323
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021154562240101438,
+      "loss": 1.263,
+      "step": 3324
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021145185869343052,
+      "loss": 1.7381,
+      "step": 3325
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021135809246048553,
+      "loss": 1.5526,
+      "step": 3326
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021126432372285678,
+      "loss": 1.5065,
+      "step": 3327
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021117055250122215,
+      "loss": 1.592,
+      "step": 3328
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021107677881625997,
+      "loss": 1.5077,
+      "step": 3329
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021098300268864928,
+      "loss": 1.4896,
+      "step": 3330
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021088922413906952,
+      "loss": 1.4342,
+      "step": 3331
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021079544318820071,
+      "loss": 1.3754,
+      "step": 3332
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002107016598567234,
+      "loss": 1.3636,
+      "step": 3333
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002106078741653186,
+      "loss": 1.6193,
+      "step": 3334
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002105140861346681,
+      "loss": 1.3429,
+      "step": 3335
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002104202957854539,
+      "loss": 1.398,
+      "step": 3336
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021032650313835852,
+      "loss": 1.4952,
+      "step": 3337
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021023270821406525,
+      "loss": 1.3688,
+      "step": 3338
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021013891103325772,
+      "loss": 1.2289,
+      "step": 3339
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00021004511161661996,
+      "loss": 1.4125,
+      "step": 3340
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00020995130998483665,
+      "loss": 1.5116,
+      "step": 3341
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00020985750615859288,
+      "loss": 1.3925,
+      "step": 3342
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00020976370015857426,
+      "loss": 1.3106,
+      "step": 3343
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00020966989200546694,
+      "loss": 1.3369,
+      "step": 3344
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00020957608171995733,
+      "loss": 1.6139,
+      "step": 3345
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00020948226932273256,
+      "loss": 1.6698,
+      "step": 3346
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00020938845483448005,
+      "loss": 1.4465,
+      "step": 3347
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00020929463827588785,
+      "loss": 1.4162,
+      "step": 3348
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00020920081966764419,
+      "loss": 1.4922,
+      "step": 3349
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00020910699903043805,
+      "loss": 1.2607,
+      "step": 3350
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00020901317638495878,
+      "loss": 1.445,
+      "step": 3351
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000208919351751896,
+      "loss": 1.2716,
+      "step": 3352
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00020882552515194,
+      "loss": 1.3808,
+      "step": 3353
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00020873169660578126,
+      "loss": 1.6711,
+      "step": 3354
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00020863786613411098,
+      "loss": 1.7704,
+      "step": 3355
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002085440337576206,
+      "loss": 1.516,
+      "step": 3356
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00020845019949700198,
+      "loss": 1.3072,
+      "step": 3357
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002083563633729474,
+      "loss": 1.5349,
+      "step": 3358
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00020826252540614962,
+      "loss": 1.3915,
+      "step": 3359
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00020816868561730179,
+      "loss": 1.3614,
+      "step": 3360
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00020807484402709739,
+      "loss": 1.4322,
+      "step": 3361
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020798100065623037,
+      "loss": 1.495,
+      "step": 3362
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000207887155525395,
+      "loss": 1.337,
+      "step": 3363
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020779330865528607,
+      "loss": 1.1892,
+      "step": 3364
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002076994600665986,
+      "loss": 1.4234,
+      "step": 3365
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020760560978002807,
+      "loss": 1.4365,
+      "step": 3366
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020751175781627036,
+      "loss": 1.3232,
+      "step": 3367
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020741790419602163,
+      "loss": 1.6024,
+      "step": 3368
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020732404893997853,
+      "loss": 1.4187,
+      "step": 3369
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002072301920688379,
+      "loss": 1.1901,
+      "step": 3370
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020713633360329708,
+      "loss": 1.4108,
+      "step": 3371
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002070424735640537,
+      "loss": 1.8073,
+      "step": 3372
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020694861197180587,
+      "loss": 1.4523,
+      "step": 3373
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020685474884725173,
+      "loss": 1.2239,
+      "step": 3374
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002067608842110901,
+      "loss": 1.1935,
+      "step": 3375
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020666701808401996,
+      "loss": 1.5544,
+      "step": 3376
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020657315048674057,
+      "loss": 1.5539,
+      "step": 3377
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002064792814399517,
+      "loss": 1.5637,
+      "step": 3378
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020638541096435322,
+      "loss": 1.4767,
+      "step": 3379
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020629153908064551,
+      "loss": 1.5302,
+      "step": 3380
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020619766580952912,
+      "loss": 1.637,
+      "step": 3381
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020610379117170496,
+      "loss": 1.7188,
+      "step": 3382
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020600991518787432,
+      "loss": 1.2621,
+      "step": 3383
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020591603787873864,
+      "loss": 1.5259,
+      "step": 3384
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020582215926499972,
+      "loss": 1.5251,
+      "step": 3385
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020572827936735964,
+      "loss": 1.5052,
+      "step": 3386
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020563439820652085,
+      "loss": 1.5836,
+      "step": 3387
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020554051580318588,
+      "loss": 1.1373,
+      "step": 3388
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020544663217805776,
+      "loss": 1.3328,
+      "step": 3389
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020535274735183966,
+      "loss": 1.4298,
+      "step": 3390
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000205258861345235,
+      "loss": 1.6353,
+      "step": 3391
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002051649741789475,
+      "loss": 1.5716,
+      "step": 3392
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002050710858736812,
+      "loss": 1.6959,
+      "step": 3393
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020497719645014032,
+      "loss": 1.4229,
+      "step": 3394
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002048833059290292,
+      "loss": 1.4719,
+      "step": 3395
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020478941433105274,
+      "loss": 1.5632,
+      "step": 3396
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020469552167691574,
+      "loss": 1.55,
+      "step": 3397
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020460162798732347,
+      "loss": 1.4819,
+      "step": 3398
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020450773328298136,
+      "loss": 1.6565,
+      "step": 3399
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020441383758459494,
+      "loss": 1.4314,
+      "step": 3400
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020431994091287022,
+      "loss": 1.6098,
+      "step": 3401
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002042260432885131,
+      "loss": 1.451,
+      "step": 3402
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020413214473223002,
+      "loss": 1.4233,
+      "step": 3403
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002040382452647273,
+      "loss": 1.587,
+      "step": 3404
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020394434490671182,
+      "loss": 1.4471,
+      "step": 3405
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002038504436788903,
+      "loss": 1.5384,
+      "step": 3406
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020375654160196996,
+      "loss": 1.3955,
+      "step": 3407
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002036626386966579,
+      "loss": 1.5427,
+      "step": 3408
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020356873498366166,
+      "loss": 1.4926,
+      "step": 3409
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002034748304836889,
+      "loss": 1.4493,
+      "step": 3410
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020338092521744736,
+      "loss": 1.4659,
+      "step": 3411
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020328701920564504,
+      "loss": 1.395,
+      "step": 3412
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020319311246899,
+      "loss": 1.5147,
+      "step": 3413
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020309920502819062,
+      "loss": 1.4268,
+      "step": 3414
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020300529690395528,
+      "loss": 1.6258,
+      "step": 3415
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020291138811699257,
+      "loss": 1.5956,
+      "step": 3416
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020281747868801134,
+      "loss": 1.441,
+      "step": 3417
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020272356863772038,
+      "loss": 1.4121,
+      "step": 3418
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020262965798682872,
+      "loss": 1.4509,
+      "step": 3419
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002025357467560455,
+      "loss": 1.4144,
+      "step": 3420
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020244183496608004,
+      "loss": 1.6757,
+      "step": 3421
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020234792263764174,
+      "loss": 1.6586,
+      "step": 3422
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020225400979144008,
+      "loss": 1.6226,
+      "step": 3423
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020216009644818475,
+      "loss": 1.4983,
+      "step": 3424
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002020661826285855,
+      "loss": 1.2512,
+      "step": 3425
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020197226835335218,
+      "loss": 1.5106,
+      "step": 3426
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002018783536431947,
+      "loss": 1.4492,
+      "step": 3427
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020178443851882313,
+      "loss": 1.4819,
+      "step": 3428
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0002016905230009476,
+      "loss": 1.5536,
+      "step": 3429
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0002015966071102784,
+      "loss": 1.3814,
+      "step": 3430
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00020150269086752575,
+      "loss": 1.368,
+      "step": 3431
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00020140877429340005,
+      "loss": 1.3233,
+      "step": 3432
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0002013148574086119,
+      "loss": 1.4376,
+      "step": 3433
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0002012209402338716,
+      "loss": 1.5402,
+      "step": 3434
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00020112702278988996,
+      "loss": 1.5709,
+      "step": 3435
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0002010331050973774,
+      "loss": 1.2618,
+      "step": 3436
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00020093918717704488,
+      "loss": 1.2464,
+      "step": 3437
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00020084526904960294,
+      "loss": 1.3257,
+      "step": 3438
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00020075135073576253,
+      "loss": 1.3132,
+      "step": 3439
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00020065743225623441,
+      "loss": 1.5267,
+      "step": 3440
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00020056351363172947,
+      "loss": 1.7704,
+      "step": 3441
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00020046959488295872,
+      "loss": 1.262,
+      "step": 3442
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00020037567603063295,
+      "loss": 1.3893,
+      "step": 3443
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00020028175709546324,
+      "loss": 1.3381,
+      "step": 3444
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0002001878380981605,
+      "loss": 1.6744,
+      "step": 3445
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0002000939190594358,
+      "loss": 1.1899,
+      "step": 3446
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0002,
+      "loss": 1.4267,
+      "step": 3447
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019990608094056425,
+      "loss": 1.4414,
+      "step": 3448
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019981216190183954,
+      "loss": 1.5146,
+      "step": 3449
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019971824290453672,
+      "loss": 1.5898,
+      "step": 3450
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019962432396936706,
+      "loss": 1.4582,
+      "step": 3451
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001995304051170413,
+      "loss": 1.6195,
+      "step": 3452
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019943648636827057,
+      "loss": 1.4174,
+      "step": 3453
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001993425677437656,
+      "loss": 1.3316,
+      "step": 3454
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019924864926423752,
+      "loss": 1.2891,
+      "step": 3455
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019915473095039708,
+      "loss": 1.4178,
+      "step": 3456
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019906081282295522,
+      "loss": 1.6494,
+      "step": 3457
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019896689490262262,
+      "loss": 1.4759,
+      "step": 3458
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019887297721011014,
+      "loss": 1.5744,
+      "step": 3459
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019877905976612846,
+      "loss": 1.36,
+      "step": 3460
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019868514259138823,
+      "loss": 1.3996,
+      "step": 3461
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019859122570659997,
+      "loss": 1.41,
+      "step": 3462
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019849730913247432,
+      "loss": 1.7267,
+      "step": 3463
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019840339288972162,
+      "loss": 1.4497,
+      "step": 3464
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019830947699905237,
+      "loss": 1.4615,
+      "step": 3465
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001982155614811769,
+      "loss": 1.5179,
+      "step": 3466
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019812164635680531,
+      "loss": 1.6862,
+      "step": 3467
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019802773164664787,
+      "loss": 1.524,
+      "step": 3468
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001979338173714145,
+      "loss": 1.5758,
+      "step": 3469
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019783990355181526,
+      "loss": 1.2212,
+      "step": 3470
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019774599020855993,
+      "loss": 1.6342,
+      "step": 3471
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019765207736235834,
+      "loss": 1.3827,
+      "step": 3472
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019755816503392,
+      "loss": 1.42,
+      "step": 3473
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001974642532439546,
+      "loss": 1.5513,
+      "step": 3474
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019737034201317133,
+      "loss": 1.4571,
+      "step": 3475
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019727643136227974,
+      "loss": 1.38,
+      "step": 3476
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019718252131198868,
+      "loss": 1.4189,
+      "step": 3477
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019708861188300745,
+      "loss": 1.5015,
+      "step": 3478
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019699470309604477,
+      "loss": 1.6329,
+      "step": 3479
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019690079497180945,
+      "loss": 1.7699,
+      "step": 3480
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019680688753101003,
+      "loss": 1.3971,
+      "step": 3481
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000196712980794355,
+      "loss": 1.3949,
+      "step": 3482
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001966190747825527,
+      "loss": 1.5877,
+      "step": 3483
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019652516951631108,
+      "loss": 1.5971,
+      "step": 3484
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019643126501633836,
+      "loss": 1.5184,
+      "step": 3485
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001963373613033421,
+      "loss": 1.5698,
+      "step": 3486
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019624345839803012,
+      "loss": 1.3053,
+      "step": 3487
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019614955632110968,
+      "loss": 1.733,
+      "step": 3488
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019605565509328826,
+      "loss": 1.6144,
+      "step": 3489
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001959617547352727,
+      "loss": 1.4383,
+      "step": 3490
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019586785526777009,
+      "loss": 1.5438,
+      "step": 3491
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019577395671148694,
+      "loss": 1.5491,
+      "step": 3492
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001956800590871299,
+      "loss": 1.4859,
+      "step": 3493
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001955861624154051,
+      "loss": 1.3013,
+      "step": 3494
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019549226671701874,
+      "loss": 1.6414,
+      "step": 3495
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00019539837201267658,
+      "loss": 1.5158,
+      "step": 3496
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019530447832308433,
+      "loss": 1.6411,
+      "step": 3497
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001952105856689473,
+      "loss": 1.5443,
+      "step": 3498
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001951166940709708,
+      "loss": 1.1671,
+      "step": 3499
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019502280354985975,
+      "loss": 1.5172,
+      "step": 3500
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001949289141263188,
+      "loss": 1.6549,
+      "step": 3501
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019483502582105252,
+      "loss": 1.5727,
+      "step": 3502
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000194741138654765,
+      "loss": 1.299,
+      "step": 3503
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019464725264816042,
+      "loss": 1.4621,
+      "step": 3504
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019455336782194226,
+      "loss": 1.538,
+      "step": 3505
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019445948419681417,
+      "loss": 1.6543,
+      "step": 3506
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019436560179347917,
+      "loss": 1.4214,
+      "step": 3507
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019427172063264043,
+      "loss": 1.3656,
+      "step": 3508
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001941778407350003,
+      "loss": 1.1357,
+      "step": 3509
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019408396212126144,
+      "loss": 1.5178,
+      "step": 3510
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019399008481212572,
+      "loss": 1.5097,
+      "step": 3511
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019389620882829508,
+      "loss": 1.3267,
+      "step": 3512
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019380233419047092,
+      "loss": 1.3226,
+      "step": 3513
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019370846091935459,
+      "loss": 1.4646,
+      "step": 3514
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019361458903564683,
+      "loss": 1.3792,
+      "step": 3515
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019352071856004833,
+      "loss": 1.4395,
+      "step": 3516
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001934268495132595,
+      "loss": 1.479,
+      "step": 3517
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019333298191598006,
+      "loss": 1.6198,
+      "step": 3518
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019323911578890995,
+      "loss": 1.3731,
+      "step": 3519
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019314525115274824,
+      "loss": 1.2347,
+      "step": 3520
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019305138802819418,
+      "loss": 1.5135,
+      "step": 3521
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019295752643594627,
+      "loss": 1.319,
+      "step": 3522
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019286366639670297,
+      "loss": 1.604,
+      "step": 3523
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019276980793116215,
+      "loss": 1.3078,
+      "step": 3524
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019267595106002154,
+      "loss": 1.4658,
+      "step": 3525
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001925820958039784,
+      "loss": 1.2569,
+      "step": 3526
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019248824218372969,
+      "loss": 1.4584,
+      "step": 3527
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019239439021997198,
+      "loss": 1.401,
+      "step": 3528
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019230053993340147,
+      "loss": 1.5826,
+      "step": 3529
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019220669134471395,
+      "loss": 1.3942,
+      "step": 3530
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019211284447460508,
+      "loss": 1.6092,
+      "step": 3531
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019201899934376967,
+      "loss": 1.5233,
+      "step": 3532
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001919251559729026,
+      "loss": 1.6646,
+      "step": 3533
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019183131438269826,
+      "loss": 1.5225,
+      "step": 3534
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001917374745938504,
+      "loss": 1.2881,
+      "step": 3535
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019164363662705264,
+      "loss": 1.477,
+      "step": 3536
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019154980050299807,
+      "loss": 1.4684,
+      "step": 3537
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019145596624237947,
+      "loss": 1.3738,
+      "step": 3538
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019136213386588898,
+      "loss": 1.4262,
+      "step": 3539
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001912683033942188,
+      "loss": 1.284,
+      "step": 3540
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019117447484806005,
+      "loss": 1.3823,
+      "step": 3541
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001910806482481041,
+      "loss": 1.2839,
+      "step": 3542
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019098682361504127,
+      "loss": 1.253,
+      "step": 3543
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019089300096956197,
+      "loss": 1.3364,
+      "step": 3544
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019079918033235583,
+      "loss": 1.469,
+      "step": 3545
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019070536172411228,
+      "loss": 1.1669,
+      "step": 3546
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019061154516551997,
+      "loss": 1.5722,
+      "step": 3547
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019051773067726752,
+      "loss": 1.3722,
+      "step": 3548
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019042391828004272,
+      "loss": 1.347,
+      "step": 3549
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019033010799453308,
+      "loss": 1.4844,
+      "step": 3550
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00019023629984142576,
+      "loss": 1.6101,
+      "step": 3551
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001901424938414071,
+      "loss": 1.5685,
+      "step": 3552
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001900486900151634,
+      "loss": 1.4698,
+      "step": 3553
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00018995488838338006,
+      "loss": 1.4333,
+      "step": 3554
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00018986108896674236,
+      "loss": 1.2694,
+      "step": 3555
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00018976729178593474,
+      "loss": 1.3964,
+      "step": 3556
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001896734968616415,
+      "loss": 1.3768,
+      "step": 3557
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00018957970421454618,
+      "loss": 1.3675,
+      "step": 3558
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00018948591386533197,
+      "loss": 1.5105,
+      "step": 3559
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001893921258346814,
+      "loss": 1.494,
+      "step": 3560
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001892983401432767,
+      "loss": 1.4588,
+      "step": 3561
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00018920455681179933,
+      "loss": 1.4066,
+      "step": 3562
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001891107758609306,
+      "loss": 1.4732,
+      "step": 3563
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00018901699731135074,
+      "loss": 1.4711,
+      "step": 3564
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001889232211837401,
+      "loss": 1.5395,
+      "step": 3565
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001888294474987779,
+      "loss": 1.326,
+      "step": 3566
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018873567627714321,
+      "loss": 1.3515,
+      "step": 3567
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018864190753951448,
+      "loss": 1.3692,
+      "step": 3568
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001885481413065695,
+      "loss": 1.6288,
+      "step": 3569
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018845437759898564,
+      "loss": 1.3668,
+      "step": 3570
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018836061643743955,
+      "loss": 1.4791,
+      "step": 3571
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018826685784260753,
+      "loss": 1.4855,
+      "step": 3572
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018817310183516503,
+      "loss": 1.702,
+      "step": 3573
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018807934843578737,
+      "loss": 1.5062,
+      "step": 3574
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018798559766514876,
+      "loss": 1.425,
+      "step": 3575
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018789184954392328,
+      "loss": 1.2828,
+      "step": 3576
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018779810409278412,
+      "loss": 1.4183,
+      "step": 3577
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018770436133240415,
+      "loss": 1.4378,
+      "step": 3578
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018761062128345532,
+      "loss": 1.5622,
+      "step": 3579
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018751688396660931,
+      "loss": 1.4181,
+      "step": 3580
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018742314940253696,
+      "loss": 1.2908,
+      "step": 3581
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018732941761190867,
+      "loss": 1.6148,
+      "step": 3582
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018723568861539404,
+      "loss": 1.3825,
+      "step": 3583
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018714196243366221,
+      "loss": 1.7409,
+      "step": 3584
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018704823908738177,
+      "loss": 1.4178,
+      "step": 3585
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018695451859722042,
+      "loss": 1.1976,
+      "step": 3586
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018686080098384548,
+      "loss": 1.451,
+      "step": 3587
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018676708626792345,
+      "loss": 1.3662,
+      "step": 3588
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018667337447012038,
+      "loss": 1.7593,
+      "step": 3589
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018657966561110145,
+      "loss": 1.6199,
+      "step": 3590
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018648595971153142,
+      "loss": 1.4157,
+      "step": 3591
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001863922567920742,
+      "loss": 1.5963,
+      "step": 3592
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018629855687339323,
+      "loss": 1.4608,
+      "step": 3593
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001862048599761511,
+      "loss": 1.2655,
+      "step": 3594
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001861111661210099,
+      "loss": 1.598,
+      "step": 3595
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018601747532863088,
+      "loss": 1.401,
+      "step": 3596
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018592378761967494,
+      "loss": 1.499,
+      "step": 3597
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018583010301480171,
+      "loss": 1.5004,
+      "step": 3598
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018573642153467084,
+      "loss": 1.6075,
+      "step": 3599
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018564274319994074,
+      "loss": 1.5927,
+      "step": 3600
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001855490680312694,
+      "loss": 1.5468,
+      "step": 3601
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001854553960493141,
+      "loss": 1.2445,
+      "step": 3602
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018536172727473123,
+      "loss": 1.4718,
+      "step": 3603
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018526806172817673,
+      "loss": 1.4699,
+      "step": 3604
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018517439943030554,
+      "loss": 1.456,
+      "step": 3605
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018508074040177235,
+      "loss": 1.4742,
+      "step": 3606
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018498708466323046,
+      "loss": 1.1504,
+      "step": 3607
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018489343223533313,
+      "loss": 1.665,
+      "step": 3608
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001847997831387323,
+      "loss": 1.3263,
+      "step": 3609
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018470613739407965,
+      "loss": 1.4121,
+      "step": 3610
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001846124950220258,
+      "loss": 1.67,
+      "step": 3611
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018451885604322084,
+      "loss": 1.3614,
+      "step": 3612
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001844252204783139,
+      "loss": 1.4562,
+      "step": 3613
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018433158834795356,
+      "loss": 1.333,
+      "step": 3614
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001842379596727875,
+      "loss": 1.5278,
+      "step": 3615
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018414433447346275,
+      "loss": 1.3722,
+      "step": 3616
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018405071277062545,
+      "loss": 1.4852,
+      "step": 3617
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000183957094584921,
+      "loss": 1.3215,
+      "step": 3618
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018386347993699424,
+      "loss": 1.6243,
+      "step": 3619
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018376986884748888,
+      "loss": 1.3411,
+      "step": 3620
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001836762613370481,
+      "loss": 1.5707,
+      "step": 3621
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018358265742631416,
+      "loss": 1.4804,
+      "step": 3622
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018348905713592862,
+      "loss": 1.2851,
+      "step": 3623
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001833954604865321,
+      "loss": 1.1895,
+      "step": 3624
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018330186749876463,
+      "loss": 1.7503,
+      "step": 3625
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018320827819326522,
+      "loss": 1.4416,
+      "step": 3626
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018311469259067221,
+      "loss": 1.5993,
+      "step": 3627
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000183021110711623,
+      "loss": 1.4267,
+      "step": 3628
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018292753257675442,
+      "loss": 1.6687,
+      "step": 3629
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018283395820670203,
+      "loss": 1.4204,
+      "step": 3630
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018274038762210114,
+      "loss": 1.3839,
+      "step": 3631
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018264682084358558,
+      "loss": 1.4549,
+      "step": 3632
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00018255325789178894,
+      "loss": 1.5895,
+      "step": 3633
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00018245969878734358,
+      "loss": 1.5206,
+      "step": 3634
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001823661435508811,
+      "loss": 1.3865,
+      "step": 3635
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00018227259220303241,
+      "loss": 1.5038,
+      "step": 3636
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001821790447644273,
+      "loss": 1.4459,
+      "step": 3637
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00018208550125569493,
+      "loss": 1.3324,
+      "step": 3638
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00018199196169746336,
+      "loss": 1.3851,
+      "step": 3639
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001818984261103601,
+      "loss": 1.1415,
+      "step": 3640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00018180489451501136,
+      "loss": 1.0758,
+      "step": 3641
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001817113669320429,
+      "loss": 1.6153,
+      "step": 3642
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001816178433820793,
+      "loss": 1.5651,
+      "step": 3643
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00018152432388574444,
+      "loss": 1.3644,
+      "step": 3644
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00018143080846366106,
+      "loss": 1.4522,
+      "step": 3645
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00018133729713645135,
+      "loss": 1.521,
+      "step": 3646
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00018124378992473626,
+      "loss": 1.2852,
+      "step": 3647
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00018115028684913604,
+      "loss": 1.6442,
+      "step": 3648
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00018105678793026994,
+      "loss": 1.4695,
+      "step": 3649
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00018096329318875634,
+      "loss": 1.5783,
+      "step": 3650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00018086980264521262,
+      "loss": 1.596,
+      "step": 3651
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00018077631632025534,
+      "loss": 1.6445,
+      "step": 3652
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00018068283423450015,
+      "loss": 1.531,
+      "step": 3653
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001805893564085616,
+      "loss": 1.3696,
+      "step": 3654
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00018049588286305342,
+      "loss": 1.4743,
+      "step": 3655
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00018040241361858835,
+      "loss": 1.3333,
+      "step": 3656
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001803089486957783,
+      "loss": 1.5639,
+      "step": 3657
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000180215488115234,
+      "loss": 1.4279,
+      "step": 3658
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001801220318975655,
+      "loss": 1.4206,
+      "step": 3659
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00018002858006338157,
+      "loss": 1.533,
+      "step": 3660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017993513263329035,
+      "loss": 1.6151,
+      "step": 3661
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017984168962789868,
+      "loss": 1.5427,
+      "step": 3662
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017974825106781278,
+      "loss": 1.1644,
+      "step": 3663
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017965481697363752,
+      "loss": 1.534,
+      "step": 3664
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001795613873659771,
+      "loss": 1.2437,
+      "step": 3665
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017946796226543448,
+      "loss": 1.2831,
+      "step": 3666
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017937454169261188,
+      "loss": 1.4001,
+      "step": 3667
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017928112566811022,
+      "loss": 1.6179,
+      "step": 3668
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001791877142125297,
+      "loss": 1.3082,
+      "step": 3669
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017909430734646935,
+      "loss": 1.4029,
+      "step": 3670
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017900090509052714,
+      "loss": 1.2826,
+      "step": 3671
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017890750746530034,
+      "loss": 1.4571,
+      "step": 3672
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017881411449138473,
+      "loss": 1.2009,
+      "step": 3673
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017872072618937545,
+      "loss": 1.4809,
+      "step": 3674
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001786273425798664,
+      "loss": 1.4999,
+      "step": 3675
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001785339636834506,
+      "loss": 1.5286,
+      "step": 3676
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001784405895207198,
+      "loss": 1.4906,
+      "step": 3677
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017834722011226498,
+      "loss": 1.6479,
+      "step": 3678
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017825385547867584,
+      "loss": 1.614,
+      "step": 3679
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001781604956405412,
+      "loss": 1.2393,
+      "step": 3680
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001780671406184487,
+      "loss": 1.5663,
+      "step": 3681
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017797379043298501,
+      "loss": 1.3122,
+      "step": 3682
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001778804451047356,
+      "loss": 1.485,
+      "step": 3683
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001777871046542851,
+      "loss": 1.3881,
+      "step": 3684
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017769376910221674,
+      "loss": 1.5274,
+      "step": 3685
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000177600438469113,
+      "loss": 1.3143,
+      "step": 3686
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017750711277555515,
+      "loss": 1.7434,
+      "step": 3687
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017741379204212323,
+      "loss": 1.3662,
+      "step": 3688
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017732047628939642,
+      "loss": 1.4299,
+      "step": 3689
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017722716553795256,
+      "loss": 1.3854,
+      "step": 3690
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017713385980836863,
+      "loss": 1.4622,
+      "step": 3691
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001770405591212203,
+      "loss": 1.6199,
+      "step": 3692
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001769472634970823,
+      "loss": 1.4102,
+      "step": 3693
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017685397295652804,
+      "loss": 1.2267,
+      "step": 3694
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001767606875201301,
+      "loss": 1.3416,
+      "step": 3695
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017666740720845956,
+      "loss": 1.5479,
+      "step": 3696
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017657413204208674,
+      "loss": 1.3505,
+      "step": 3697
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017648086204158055,
+      "loss": 1.4348,
+      "step": 3698
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017638759722750895,
+      "loss": 1.4262,
+      "step": 3699
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017629433762043862,
+      "loss": 1.4625,
+      "step": 3700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017620108324093516,
+      "loss": 1.4374,
+      "step": 3701
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017610783410956294,
+      "loss": 1.4246,
+      "step": 3702
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001760145902468853,
+      "loss": 1.1806,
+      "step": 3703
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017592135167346446,
+      "loss": 1.5397,
+      "step": 3704
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017582811840986112,
+      "loss": 1.4094,
+      "step": 3705
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017573489047663534,
+      "loss": 1.4526,
+      "step": 3706
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017564166789434544,
+      "loss": 1.2876,
+      "step": 3707
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017554845068354905,
+      "loss": 1.6941,
+      "step": 3708
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001754552388648023,
+      "loss": 1.4479,
+      "step": 3709
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001753620324586603,
+      "loss": 1.4684,
+      "step": 3710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017526883148567682,
+      "loss": 1.3978,
+      "step": 3711
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017517563596640466,
+      "loss": 1.6162,
+      "step": 3712
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001750824459213951,
+      "loss": 1.3785,
+      "step": 3713
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017498926137119852,
+      "loss": 1.3874,
+      "step": 3714
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017489608233636384,
+      "loss": 1.5013,
+      "step": 3715
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000174802908837439,
+      "loss": 1.5097,
+      "step": 3716
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017470974089497045,
+      "loss": 1.6127,
+      "step": 3717
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017461657852950379,
+      "loss": 1.4182,
+      "step": 3718
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017452342176158288,
+      "loss": 1.5645,
+      "step": 3719
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001744302706117508,
+      "loss": 1.3734,
+      "step": 3720
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017433712510054927,
+      "loss": 1.5109,
+      "step": 3721
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017424398524851854,
+      "loss": 1.2866,
+      "step": 3722
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017415085107619796,
+      "loss": 1.5304,
+      "step": 3723
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017405772260412533,
+      "loss": 1.3401,
+      "step": 3724
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017396459985283738,
+      "loss": 1.5118,
+      "step": 3725
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017387148284286942,
+      "loss": 1.439,
+      "step": 3726
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001737783715947558,
+      "loss": 1.6858,
+      "step": 3727
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017368526612902916,
+      "loss": 1.5324,
+      "step": 3728
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001735921664662213,
+      "loss": 1.4981,
+      "step": 3729
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001734990726268623,
+      "loss": 1.4727,
+      "step": 3730
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017340598463148142,
+      "loss": 1.3022,
+      "step": 3731
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017331290250060621,
+      "loss": 1.4258,
+      "step": 3732
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001732198262547633,
+      "loss": 1.5375,
+      "step": 3733
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001731267559144777,
+      "loss": 1.5744,
+      "step": 3734
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017303369150027334,
+      "loss": 1.4614,
+      "step": 3735
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001729406330326727,
+      "loss": 1.2381,
+      "step": 3736
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017284758053219697,
+      "loss": 1.3975,
+      "step": 3737
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001727545340193663,
+      "loss": 1.4678,
+      "step": 3738
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017266149351469898,
+      "loss": 1.6308,
+      "step": 3739
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001725684590387125,
+      "loss": 1.3744,
+      "step": 3740
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017247543061192267,
+      "loss": 1.4771,
+      "step": 3741
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017238240825484424,
+      "loss": 1.2911,
+      "step": 3742
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001722893919879903,
+      "loss": 1.4236,
+      "step": 3743
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001721963818318729,
+      "loss": 1.5266,
+      "step": 3744
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001721033778070025,
+      "loss": 1.5195,
+      "step": 3745
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017201037993388852,
+      "loss": 1.2341,
+      "step": 3746
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017191738823303863,
+      "loss": 1.341,
+      "step": 3747
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017182440272495942,
+      "loss": 1.4769,
+      "step": 3748
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017173142343015598,
+      "loss": 1.6219,
+      "step": 3749
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017163845036913226,
+      "loss": 1.3345,
+      "step": 3750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017154548356239038,
+      "loss": 1.4534,
+      "step": 3751
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017145252303043163,
+      "loss": 1.5673,
+      "step": 3752
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001713595687937554,
+      "loss": 1.546,
+      "step": 3753
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017126662087286007,
+      "loss": 1.5918,
+      "step": 3754
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017117367928824254,
+      "loss": 1.4186,
+      "step": 3755
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017108074406039814,
+      "loss": 1.3886,
+      "step": 3756
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017098781520982103,
+      "loss": 1.5131,
+      "step": 3757
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017089489275700376,
+      "loss": 1.4016,
+      "step": 3758
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017080197672243768,
+      "loss": 1.4917,
+      "step": 3759
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017070906712661246,
+      "loss": 1.3416,
+      "step": 3760
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001706161639900167,
+      "loss": 1.5693,
+      "step": 3761
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017052326733313716,
+      "loss": 1.3407,
+      "step": 3762
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017043037717645956,
+      "loss": 1.4562,
+      "step": 3763
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001703374935404679,
+      "loss": 1.3788,
+      "step": 3764
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017024461644564497,
+      "loss": 1.4814,
+      "step": 3765
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001701517459124719,
+      "loss": 1.4798,
+      "step": 3766
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00017005888196142854,
+      "loss": 1.531,
+      "step": 3767
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00016996602461299317,
+      "loss": 1.262,
+      "step": 3768
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016987317388764277,
+      "loss": 1.5031,
+      "step": 3769
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016978032980585262,
+      "loss": 1.3687,
+      "step": 3770
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001696874923880967,
+      "loss": 1.3055,
+      "step": 3771
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016959466165484768,
+      "loss": 1.3966,
+      "step": 3772
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001695018376265763,
+      "loss": 1.4191,
+      "step": 3773
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001694090203237523,
+      "loss": 1.3824,
+      "step": 3774
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001693162097668436,
+      "loss": 1.3579,
+      "step": 3775
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016922340597631688,
+      "loss": 1.5564,
+      "step": 3776
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001691306089726371,
+      "loss": 1.255,
+      "step": 3777
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016903781877626792,
+      "loss": 1.4952,
+      "step": 3778
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016894503540767133,
+      "loss": 1.5926,
+      "step": 3779
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016885225888730796,
+      "loss": 1.5149,
+      "step": 3780
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001687594892356368,
+      "loss": 1.5468,
+      "step": 3781
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016866672647311548,
+      "loss": 1.4746,
+      "step": 3782
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016857397062019987,
+      "loss": 1.4095,
+      "step": 3783
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016848122169734467,
+      "loss": 1.5063,
+      "step": 3784
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016838847972500267,
+      "loss": 1.4838,
+      "step": 3785
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016829574472362546,
+      "loss": 1.3756,
+      "step": 3786
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001682030167136628,
+      "loss": 1.4238,
+      "step": 3787
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016811029571556314,
+      "loss": 1.4745,
+      "step": 3788
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001680175817497733,
+      "loss": 1.468,
+      "step": 3789
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016792487483673844,
+      "loss": 1.2754,
+      "step": 3790
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016783217499690241,
+      "loss": 1.3574,
+      "step": 3791
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016773948225070717,
+      "loss": 1.6624,
+      "step": 3792
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016764679661859355,
+      "loss": 1.6925,
+      "step": 3793
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016755411812100027,
+      "loss": 1.5826,
+      "step": 3794
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016746144677836508,
+      "loss": 1.524,
+      "step": 3795
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016736878261112354,
+      "loss": 1.4148,
+      "step": 3796
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016727612563971017,
+      "loss": 1.3754,
+      "step": 3797
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001671834758845575,
+      "loss": 1.3707,
+      "step": 3798
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016709083336609675,
+      "loss": 1.4784,
+      "step": 3799
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016699819810475736,
+      "loss": 1.4993,
+      "step": 3800
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016690557012096727,
+      "loss": 1.6315,
+      "step": 3801
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016681294943515272,
+      "loss": 1.5977,
+      "step": 3802
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016672033606773847,
+      "loss": 1.6527,
+      "step": 3803
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016662773003914754,
+      "loss": 1.4217,
+      "step": 3804
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016653513136980134,
+      "loss": 1.28,
+      "step": 3805
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001664425400801199,
+      "loss": 1.5495,
+      "step": 3806
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016634995619052126,
+      "loss": 1.599,
+      "step": 3807
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016625737972142206,
+      "loss": 1.6405,
+      "step": 3808
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001661648106932372,
+      "loss": 1.5159,
+      "step": 3809
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016607224912638005,
+      "loss": 1.2115,
+      "step": 3810
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016597969504126222,
+      "loss": 1.5828,
+      "step": 3811
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001658871484582937,
+      "loss": 1.5234,
+      "step": 3812
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016579460939788286,
+      "loss": 1.5423,
+      "step": 3813
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001657020778804364,
+      "loss": 1.5283,
+      "step": 3814
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016560955392635923,
+      "loss": 1.3509,
+      "step": 3815
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016551703755605495,
+      "loss": 1.3183,
+      "step": 3816
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000165424528789925,
+      "loss": 1.4361,
+      "step": 3817
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016533202764836962,
+      "loss": 1.4687,
+      "step": 3818
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001652395341517869,
+      "loss": 1.544,
+      "step": 3819
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016514704832057365,
+      "loss": 1.3955,
+      "step": 3820
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001650545701751248,
+      "loss": 1.4223,
+      "step": 3821
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016496209973583352,
+      "loss": 1.4833,
+      "step": 3822
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001648696370230915,
+      "loss": 1.5131,
+      "step": 3823
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001647771820572885,
+      "loss": 1.5541,
+      "step": 3824
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001646847348588127,
+      "loss": 1.4337,
+      "step": 3825
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016459229544805046,
+      "loss": 1.5211,
+      "step": 3826
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001644998638453867,
+      "loss": 1.333,
+      "step": 3827
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016440744007120415,
+      "loss": 1.3747,
+      "step": 3828
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016431502414588427,
+      "loss": 1.1768,
+      "step": 3829
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001642226160898065,
+      "loss": 1.4446,
+      "step": 3830
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016413021592334872,
+      "loss": 1.3206,
+      "step": 3831
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001640378236668869,
+      "loss": 1.4766,
+      "step": 3832
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016394543934079546,
+      "loss": 1.4543,
+      "step": 3833
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00016385306296544687,
+      "loss": 1.6452,
+      "step": 3834
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000163760694561212,
+      "loss": 1.5049,
+      "step": 3835
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001636683341484599,
+      "loss": 1.3327,
+      "step": 3836
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016357598174755787,
+      "loss": 1.4649,
+      "step": 3837
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016348363737887138,
+      "loss": 1.2861,
+      "step": 3838
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016339130106276425,
+      "loss": 1.5064,
+      "step": 3839
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001632989728195985,
+      "loss": 1.3072,
+      "step": 3840
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016320665266973428,
+      "loss": 1.3647,
+      "step": 3841
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016311434063353,
+      "loss": 1.481,
+      "step": 3842
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016302203673134228,
+      "loss": 1.42,
+      "step": 3843
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.000162929740983526,
+      "loss": 1.4914,
+      "step": 3844
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016283745341043413,
+      "loss": 1.5088,
+      "step": 3845
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016274517403241798,
+      "loss": 1.5087,
+      "step": 3846
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016265290286982692,
+      "loss": 1.5419,
+      "step": 3847
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001625606399430086,
+      "loss": 1.4995,
+      "step": 3848
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001624683852723087,
+      "loss": 1.6052,
+      "step": 3849
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016237613887807144,
+      "loss": 1.2752,
+      "step": 3850
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016228390078063868,
+      "loss": 1.5137,
+      "step": 3851
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.000162191671000351,
+      "loss": 1.3435,
+      "step": 3852
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016209944955754673,
+      "loss": 1.4484,
+      "step": 3853
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016200723647256262,
+      "loss": 1.7031,
+      "step": 3854
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016191503176573338,
+      "loss": 1.3559,
+      "step": 3855
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016182283545739199,
+      "loss": 1.4148,
+      "step": 3856
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016173064756786965,
+      "loss": 1.5083,
+      "step": 3857
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016163846811749542,
+      "loss": 1.1208,
+      "step": 3858
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.000161546297126597,
+      "loss": 1.2748,
+      "step": 3859
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016145413461549958,
+      "loss": 1.4945,
+      "step": 3860
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.000161361980604527,
+      "loss": 1.3588,
+      "step": 3861
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016126983511400103,
+      "loss": 1.3677,
+      "step": 3862
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016117769816424154,
+      "loss": 1.2158,
+      "step": 3863
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016108556977556654,
+      "loss": 1.5144,
+      "step": 3864
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016099344996829218,
+      "loss": 1.3052,
+      "step": 3865
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016090133876273265,
+      "loss": 1.496,
+      "step": 3866
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001608092361792004,
+      "loss": 1.3607,
+      "step": 3867
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016071714223800574,
+      "loss": 1.7108,
+      "step": 3868
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001606250569594573,
+      "loss": 1.3849,
+      "step": 3869
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016053298036386158,
+      "loss": 1.3623,
+      "step": 3870
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016044091247152347,
+      "loss": 1.4677,
+      "step": 3871
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016034885330274556,
+      "loss": 1.2394,
+      "step": 3872
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016025680287782887,
+      "loss": 1.5879,
+      "step": 3873
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001601647612170723,
+      "loss": 1.3684,
+      "step": 3874
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016007272834077282,
+      "loss": 1.5064,
+      "step": 3875
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015998070426922556,
+      "loss": 1.2837,
+      "step": 3876
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015988868902272355,
+      "loss": 1.3801,
+      "step": 3877
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015979668262155807,
+      "loss": 1.4114,
+      "step": 3878
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015970468508601823,
+      "loss": 1.5055,
+      "step": 3879
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015961269643639144,
+      "loss": 1.6037,
+      "step": 3880
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015952071669296286,
+      "loss": 1.4703,
+      "step": 3881
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015942874587601607,
+      "loss": 1.4309,
+      "step": 3882
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015933678400583217,
+      "loss": 1.4165,
+      "step": 3883
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001592448311026908,
+      "loss": 1.5553,
+      "step": 3884
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015915288718686924,
+      "loss": 1.3561,
+      "step": 3885
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015906095227864308,
+      "loss": 1.5221,
+      "step": 3886
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015896902639828565,
+      "loss": 1.5243,
+      "step": 3887
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001588771095660685,
+      "loss": 1.4631,
+      "step": 3888
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015878520180226106,
+      "loss": 1.6563,
+      "step": 3889
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015869330312713077,
+      "loss": 1.4543,
+      "step": 3890
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015860141356094333,
+      "loss": 1.3859,
+      "step": 3891
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001585095331239619,
+      "loss": 1.4383,
+      "step": 3892
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015841766183644817,
+      "loss": 1.2907,
+      "step": 3893
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001583257997186614,
+      "loss": 1.3133,
+      "step": 3894
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015823394679085912,
+      "loss": 1.4119,
+      "step": 3895
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015814210307329665,
+      "loss": 1.5526,
+      "step": 3896
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001580502685862274,
+      "loss": 1.4463,
+      "step": 3897
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015795844334990262,
+      "loss": 1.4398,
+      "step": 3898
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015786662738457167,
+      "loss": 1.6007,
+      "step": 3899
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015777482071048168,
+      "loss": 1.6081,
+      "step": 3900
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015768302334787798,
+      "loss": 1.7558,
+      "step": 3901
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015759123531700352,
+      "loss": 1.3625,
+      "step": 3902
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00015749945663809953,
+      "loss": 1.3287,
+      "step": 3903
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001574076873314049,
+      "loss": 1.4921,
+      "step": 3904
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015731592741715674,
+      "loss": 1.3357,
+      "step": 3905
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015722417691558968,
+      "loss": 1.602,
+      "step": 3906
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015713243584693673,
+      "loss": 1.0457,
+      "step": 3907
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015704070423142857,
+      "loss": 1.1865,
+      "step": 3908
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015694898208929377,
+      "loss": 1.6022,
+      "step": 3909
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015685726944075895,
+      "loss": 1.5503,
+      "step": 3910
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015676556630604845,
+      "loss": 1.4013,
+      "step": 3911
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001566738727053847,
+      "loss": 1.5691,
+      "step": 3912
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001565821886589879,
+      "loss": 1.4147,
+      "step": 3913
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015649051418707637,
+      "loss": 1.3905,
+      "step": 3914
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015639884930986585,
+      "loss": 1.5954,
+      "step": 3915
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015630719404757055,
+      "loss": 1.5868,
+      "step": 3916
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000156215548420402,
+      "loss": 1.3968,
+      "step": 3917
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015612391244857008,
+      "loss": 1.3334,
+      "step": 3918
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001560322861522822,
+      "loss": 1.5883,
+      "step": 3919
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015594066955174386,
+      "loss": 1.4765,
+      "step": 3920
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015584906266715825,
+      "loss": 1.3468,
+      "step": 3921
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001557574655187266,
+      "loss": 1.5827,
+      "step": 3922
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015566587812664777,
+      "loss": 1.5872,
+      "step": 3923
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015557430051111858,
+      "loss": 1.3055,
+      "step": 3924
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015548273269233394,
+      "loss": 1.222,
+      "step": 3925
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015539117469048604,
+      "loss": 1.547,
+      "step": 3926
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015529962652576545,
+      "loss": 1.44,
+      "step": 3927
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001552080882183603,
+      "loss": 1.3355,
+      "step": 3928
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015511655978845654,
+      "loss": 1.443,
+      "step": 3929
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015502504125623802,
+      "loss": 1.5233,
+      "step": 3930
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015493353264188646,
+      "loss": 1.4485,
+      "step": 3931
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001548420339655812,
+      "loss": 1.4701,
+      "step": 3932
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001547505452474996,
+      "loss": 1.2966,
+      "step": 3933
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001546590665078166,
+      "loss": 1.5997,
+      "step": 3934
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015456759776670526,
+      "loss": 1.6566,
+      "step": 3935
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001544761390443361,
+      "loss": 1.3635,
+      "step": 3936
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015438469036087763,
+      "loss": 1.4599,
+      "step": 3937
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000154293251736496,
+      "loss": 1.4238,
+      "step": 3938
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015420182319135546,
+      "loss": 1.3969,
+      "step": 3939
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001541104047456175,
+      "loss": 1.3743,
+      "step": 3940
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015401899641944193,
+      "loss": 1.3016,
+      "step": 3941
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015392759823298606,
+      "loss": 1.346,
+      "step": 3942
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001538362102064049,
+      "loss": 1.4975,
+      "step": 3943
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001537448323598514,
+      "loss": 1.5042,
+      "step": 3944
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015365346471347613,
+      "loss": 1.1579,
+      "step": 3945
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001535621072874275,
+      "loss": 1.3804,
+      "step": 3946
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015347076010185156,
+      "loss": 1.7045,
+      "step": 3947
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001533794231768923,
+      "loss": 1.4637,
+      "step": 3948
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015328809653269112,
+      "loss": 1.5597,
+      "step": 3949
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015319678018938754,
+      "loss": 1.4709,
+      "step": 3950
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015310547416711853,
+      "loss": 1.4817,
+      "step": 3951
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001530141784860189,
+      "loss": 1.5771,
+      "step": 3952
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015292289316622107,
+      "loss": 1.686,
+      "step": 3953
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015283161822785538,
+      "loss": 1.4763,
+      "step": 3954
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015274035369104965,
+      "loss": 1.5335,
+      "step": 3955
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015264909957592962,
+      "loss": 1.4896,
+      "step": 3956
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001525578559026185,
+      "loss": 1.1671,
+      "step": 3957
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015246662269123736,
+      "loss": 1.2872,
+      "step": 3958
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015237539996190507,
+      "loss": 1.619,
+      "step": 3959
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001522841877347378,
+      "loss": 1.2967,
+      "step": 3960
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015219298602984983,
+      "loss": 1.4011,
+      "step": 3961
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015210179486735287,
+      "loss": 1.3928,
+      "step": 3962
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015201061426735644,
+      "loss": 1.4269,
+      "step": 3963
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015191944424996755,
+      "loss": 1.4544,
+      "step": 3964
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001518282848352911,
+      "loss": 1.1863,
+      "step": 3965
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015173713604342944,
+      "loss": 1.5975,
+      "step": 3966
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015164599789448276,
+      "loss": 1.4002,
+      "step": 3967
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015155487040854876,
+      "loss": 1.592,
+      "step": 3968
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015146375360572295,
+      "loss": 1.3964,
+      "step": 3969
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015137264750609818,
+      "loss": 1.4385,
+      "step": 3970
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00015128155212976545,
+      "loss": 1.5781,
+      "step": 3971
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001511904674968128,
+      "loss": 1.3721,
+      "step": 3972
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00015109939362732634,
+      "loss": 1.4035,
+      "step": 3973
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00015100833054138964,
+      "loss": 1.4517,
+      "step": 3974
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001509172782590839,
+      "loss": 1.3413,
+      "step": 3975
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00015082623680048797,
+      "loss": 1.4764,
+      "step": 3976
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00015073520618567824,
+      "loss": 1.4709,
+      "step": 3977
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001506441864347288,
+      "loss": 1.2146,
+      "step": 3978
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00015055317756771126,
+      "loss": 1.4756,
+      "step": 3979
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.000150462179604695,
+      "loss": 1.1801,
+      "step": 3980
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00015037119256574663,
+      "loss": 1.3945,
+      "step": 3981
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00015028021647093087,
+      "loss": 1.5291,
+      "step": 3982
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00015018925134030947,
+      "loss": 1.3246,
+      "step": 3983
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001500982971939423,
+      "loss": 1.5777,
+      "step": 3984
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00015000735405188632,
+      "loss": 1.4263,
+      "step": 3985
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014991642193419645,
+      "loss": 1.6574,
+      "step": 3986
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014982550086092492,
+      "loss": 1.456,
+      "step": 3987
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014973459085212173,
+      "loss": 1.5215,
+      "step": 3988
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001496436919278342,
+      "loss": 1.4223,
+      "step": 3989
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014955280410810745,
+      "loss": 1.4757,
+      "step": 3990
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014946192741298396,
+      "loss": 1.4177,
+      "step": 3991
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001493710618625038,
+      "loss": 1.3957,
+      "step": 3992
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014928020747670482,
+      "loss": 1.5148,
+      "step": 3993
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.000149189364275622,
+      "loss": 1.3999,
+      "step": 3994
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014909853227928818,
+      "loss": 1.5666,
+      "step": 3995
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014900771150773353,
+      "loss": 1.5765,
+      "step": 3996
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014891690198098588,
+      "loss": 1.5927,
+      "step": 3997
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014882610371907048,
+      "loss": 1.4599,
+      "step": 3998
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014873531674201022,
+      "loss": 1.4095,
+      "step": 3999
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001486445410698253,
+      "loss": 1.546,
+      "step": 4000
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001485537767225337,
+      "loss": 1.4721,
+      "step": 4001
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014846302372015054,
+      "loss": 1.5146,
+      "step": 4002
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014837228208268893,
+      "loss": 1.3026,
+      "step": 4003
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014828155183015888,
+      "loss": 1.3127,
+      "step": 4004
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014819083298256853,
+      "loss": 1.3784,
+      "step": 4005
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001481001255599229,
+      "loss": 1.4629,
+      "step": 4006
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014800942958222493,
+      "loss": 1.4777,
+      "step": 4007
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001479187450694748,
+      "loss": 1.525,
+      "step": 4008
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014782807204167023,
+      "loss": 1.4217,
+      "step": 4009
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014773741051880653,
+      "loss": 1.3679,
+      "step": 4010
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014764676052087618,
+      "loss": 1.4404,
+      "step": 4011
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014755612206786945,
+      "loss": 1.4544,
+      "step": 4012
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014746549517977378,
+      "loss": 1.4064,
+      "step": 4013
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014737487987657438,
+      "loss": 1.3097,
+      "step": 4014
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001472842761782535,
+      "loss": 1.3996,
+      "step": 4015
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014719368410479118,
+      "loss": 1.2374,
+      "step": 4016
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014710310367616468,
+      "loss": 1.2583,
+      "step": 4017
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014701253491234884,
+      "loss": 1.689,
+      "step": 4018
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014692197783331584,
+      "loss": 1.3779,
+      "step": 4019
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001468314324590353,
+      "loss": 1.501,
+      "step": 4020
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014674089880947424,
+      "loss": 1.4598,
+      "step": 4021
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014665037690459717,
+      "loss": 1.475,
+      "step": 4022
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001465598667643659,
+      "loss": 1.5337,
+      "step": 4023
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014646936840873975,
+      "loss": 1.2012,
+      "step": 4024
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014637888185767533,
+      "loss": 1.5735,
+      "step": 4025
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001462884071311268,
+      "loss": 1.5754,
+      "step": 4026
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014619794424904558,
+      "loss": 1.5419,
+      "step": 4027
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014610749323138055,
+      "loss": 1.363,
+      "step": 4028
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014601705409807793,
+      "loss": 1.355,
+      "step": 4029
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001459266268690813,
+      "loss": 1.4101,
+      "step": 4030
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014583621156433172,
+      "loss": 1.3874,
+      "step": 4031
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014574580820376747,
+      "loss": 1.5682,
+      "step": 4032
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014565541680732435,
+      "loss": 1.519,
+      "step": 4033
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014556503739493538,
+      "loss": 1.3361,
+      "step": 4034
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001454746699865311,
+      "loss": 1.4448,
+      "step": 4035
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014538431460203914,
+      "loss": 1.5179,
+      "step": 4036
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001452939712613849,
+      "loss": 1.1576,
+      "step": 4037
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001452036399844906,
+      "loss": 1.2798,
+      "step": 4038
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014511332079127632,
+      "loss": 1.3621,
+      "step": 4039
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00014502301370165904,
+      "loss": 1.5725,
+      "step": 4040
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014493271873555337,
+      "loss": 1.3383,
+      "step": 4041
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014484243591287105,
+      "loss": 1.3511,
+      "step": 4042
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014475216525352123,
+      "loss": 1.5054,
+      "step": 4043
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014466190677741046,
+      "loss": 1.3032,
+      "step": 4044
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014457166050444242,
+      "loss": 1.6976,
+      "step": 4045
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014448142645451836,
+      "loss": 1.5473,
+      "step": 4046
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014439120464753642,
+      "loss": 1.391,
+      "step": 4047
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014430099510339247,
+      "loss": 1.4206,
+      "step": 4048
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014421079784197947,
+      "loss": 1.2335,
+      "step": 4049
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014412061288318766,
+      "loss": 1.5874,
+      "step": 4050
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001440304402469046,
+      "loss": 1.5648,
+      "step": 4051
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014394027995301523,
+      "loss": 1.4569,
+      "step": 4052
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014385013202140154,
+      "loss": 1.626,
+      "step": 4053
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014375999647194302,
+      "loss": 1.3524,
+      "step": 4054
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014366987332451632,
+      "loss": 1.3486,
+      "step": 4055
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014357976259899536,
+      "loss": 1.417,
+      "step": 4056
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014348966431525136,
+      "loss": 1.4001,
+      "step": 4057
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014339957849315277,
+      "loss": 1.4426,
+      "step": 4058
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001433095051525652,
+      "loss": 1.3689,
+      "step": 4059
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001432194443133517,
+      "loss": 1.5094,
+      "step": 4060
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001431293959953725,
+      "loss": 1.6864,
+      "step": 4061
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014303936021848497,
+      "loss": 1.6229,
+      "step": 4062
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014294933700254378,
+      "loss": 1.3572,
+      "step": 4063
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014285932636740083,
+      "loss": 1.4597,
+      "step": 4064
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014276932833290527,
+      "loss": 1.3748,
+      "step": 4065
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014267934291890336,
+      "loss": 1.3587,
+      "step": 4066
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001425893701452388,
+      "loss": 1.5782,
+      "step": 4067
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014249941003175218,
+      "loss": 1.6506,
+      "step": 4068
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014240946259828173,
+      "loss": 1.4558,
+      "step": 4069
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014231952786466234,
+      "loss": 1.4906,
+      "step": 4070
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014222960585072672,
+      "loss": 1.5168,
+      "step": 4071
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014213969657630413,
+      "loss": 1.4394,
+      "step": 4072
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014204980006122156,
+      "loss": 1.2999,
+      "step": 4073
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014195991632530287,
+      "loss": 1.3073,
+      "step": 4074
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014187004538836927,
+      "loss": 1.4078,
+      "step": 4075
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014178018727023902,
+      "loss": 1.5508,
+      "step": 4076
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014169034199072755,
+      "loss": 1.4167,
+      "step": 4077
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014160050956964773,
+      "loss": 1.6654,
+      "step": 4078
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014151069002680915,
+      "loss": 1.1898,
+      "step": 4079
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014142088338201897,
+      "loss": 1.5366,
+      "step": 4080
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014133108965508115,
+      "loss": 1.3162,
+      "step": 4081
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014124130886579718,
+      "loss": 1.6783,
+      "step": 4082
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014115154103396535,
+      "loss": 1.5308,
+      "step": 4083
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014106178617938133,
+      "loss": 1.4494,
+      "step": 4084
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014097204432183774,
+      "loss": 1.4024,
+      "step": 4085
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014088231548112457,
+      "loss": 1.5181,
+      "step": 4086
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014079259967702865,
+      "loss": 1.192,
+      "step": 4087
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001407028969293342,
+      "loss": 1.3928,
+      "step": 4088
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014061320725782239,
+      "loss": 1.2224,
+      "step": 4089
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001405235306822716,
+      "loss": 1.2262,
+      "step": 4090
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001404338672224572,
+      "loss": 1.2806,
+      "step": 4091
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014034421689815196,
+      "loss": 1.3579,
+      "step": 4092
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014025457972912528,
+      "loss": 1.5083,
+      "step": 4093
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001401649557351441,
+      "loss": 1.0533,
+      "step": 4094
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00014007534493597228,
+      "loss": 1.4872,
+      "step": 4095
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00013998574735137067,
+      "loss": 1.2764,
+      "step": 4096
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00013989616300109743,
+      "loss": 1.5377,
+      "step": 4097
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00013980659190490754,
+      "loss": 1.5388,
+      "step": 4098
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00013971703408255338,
+      "loss": 1.3951,
+      "step": 4099
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00013962748955378398,
+      "loss": 1.3062,
+      "step": 4100
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00013953795833834595,
+      "loss": 1.3837,
+      "step": 4101
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00013944844045598242,
+      "loss": 1.3684,
+      "step": 4102
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001393589359264341,
+      "loss": 1.4094,
+      "step": 4103
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001392694447694383,
+      "loss": 1.5443,
+      "step": 4104
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00013917996700472973,
+      "loss": 1.459,
+      "step": 4105
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00013909050265203992,
+      "loss": 1.283,
+      "step": 4106
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001390010517310976,
+      "loss": 1.3704,
+      "step": 4107
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013891161426162838,
+      "loss": 1.3909,
+      "step": 4108
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013882219026335507,
+      "loss": 1.3977,
+      "step": 4109
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013873277975599737,
+      "loss": 1.4394,
+      "step": 4110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013864338275927204,
+      "loss": 1.443,
+      "step": 4111
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013855399929289304,
+      "loss": 1.5868,
+      "step": 4112
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013846462937657097,
+      "loss": 1.4295,
+      "step": 4113
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013837527303001384,
+      "loss": 1.4844,
+      "step": 4114
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013828593027292638,
+      "loss": 1.5096,
+      "step": 4115
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013819660112501054,
+      "loss": 1.3751,
+      "step": 4116
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013810728560596504,
+      "loss": 1.3445,
+      "step": 4117
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001380179837354858,
+      "loss": 1.3413,
+      "step": 4118
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013792869553326558,
+      "loss": 1.5578,
+      "step": 4119
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001378394210189943,
+      "loss": 1.6083,
+      "step": 4120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001377501602123586,
+      "loss": 1.4284,
+      "step": 4121
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013766091313304244,
+      "loss": 1.4128,
+      "step": 4122
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013757167980072636,
+      "loss": 1.4399,
+      "step": 4123
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013748246023508824,
+      "loss": 1.4415,
+      "step": 4124
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013739325445580262,
+      "loss": 1.4123,
+      "step": 4125
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013730406248254132,
+      "loss": 1.6901,
+      "step": 4126
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001372148843349727,
+      "loss": 1.3888,
+      "step": 4127
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013712572003276246,
+      "loss": 1.2742,
+      "step": 4128
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013703656959557307,
+      "loss": 1.4939,
+      "step": 4129
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013694743304306392,
+      "loss": 1.386,
+      "step": 4130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013685831039489144,
+      "loss": 1.5251,
+      "step": 4131
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001367692016707089,
+      "loss": 1.444,
+      "step": 4132
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013668010689016656,
+      "loss": 1.4973,
+      "step": 4133
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013659102607291144,
+      "loss": 1.3775,
+      "step": 4134
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013650195923858787,
+      "loss": 1.5361,
+      "step": 4135
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013641290640683659,
+      "loss": 1.4622,
+      "step": 4136
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013632386759729573,
+      "loss": 1.2315,
+      "step": 4137
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013623484282959992,
+      "loss": 1.4655,
+      "step": 4138
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013614583212338103,
+      "loss": 1.2794,
+      "step": 4139
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001360568354982676,
+      "loss": 1.416,
+      "step": 4140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013596785297388515,
+      "loss": 1.5032,
+      "step": 4141
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013587888456985606,
+      "loss": 1.3916,
+      "step": 4142
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013578993030579972,
+      "loss": 1.4817,
+      "step": 4143
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001357009902013322,
+      "loss": 1.4351,
+      "step": 4144
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013561206427606653,
+      "loss": 1.2832,
+      "step": 4145
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001355231525496128,
+      "loss": 1.2948,
+      "step": 4146
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001354342550415776,
+      "loss": 1.3117,
+      "step": 4147
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013534537177156475,
+      "loss": 1.3539,
+      "step": 4148
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013525650275917462,
+      "loss": 1.5672,
+      "step": 4149
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001351676480240047,
+      "loss": 1.4126,
+      "step": 4150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001350788075856492,
+      "loss": 1.5884,
+      "step": 4151
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013498998146369915,
+      "loss": 1.3408,
+      "step": 4152
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013490116967774247,
+      "loss": 1.5279,
+      "step": 4153
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013481237224736395,
+      "loss": 1.4442,
+      "step": 4154
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013472358919214508,
+      "loss": 1.5106,
+      "step": 4155
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001346348205316644,
+      "loss": 1.5932,
+      "step": 4156
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013454606628549704,
+      "loss": 1.5247,
+      "step": 4157
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013445732647321523,
+      "loss": 1.526,
+      "step": 4158
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013436860111438765,
+      "loss": 1.4817,
+      "step": 4159
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013427989022858015,
+      "loss": 1.2434,
+      "step": 4160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013419119383535513,
+      "loss": 1.4353,
+      "step": 4161
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013410251195427196,
+      "loss": 1.3305,
+      "step": 4162
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013401384460488678,
+      "loss": 1.4647,
+      "step": 4163
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013392519180675235,
+      "loss": 1.2236,
+      "step": 4164
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013383655357941852,
+      "loss": 1.4763,
+      "step": 4165
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013374792994243163,
+      "loss": 1.2927,
+      "step": 4166
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013365932091533516,
+      "loss": 1.4313,
+      "step": 4167
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013357072651766884,
+      "loss": 1.4033,
+      "step": 4168
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001334821467689698,
+      "loss": 1.2799,
+      "step": 4169
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013339358168877133,
+      "loss": 1.3602,
+      "step": 4170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013330503129660402,
+      "loss": 1.5937,
+      "step": 4171
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013321649561199488,
+      "loss": 1.3462,
+      "step": 4172
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001331279746544678,
+      "loss": 1.3863,
+      "step": 4173
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013303946844354334,
+      "loss": 1.4365,
+      "step": 4174
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00013295097699873897,
+      "loss": 1.3175,
+      "step": 4175
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001328625003395687,
+      "loss": 1.3862,
+      "step": 4176
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013277403848554347,
+      "loss": 1.2749,
+      "step": 4177
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013268559145617075,
+      "loss": 1.3735,
+      "step": 4178
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001325971592709549,
+      "loss": 1.4545,
+      "step": 4179
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001325087419493971,
+      "loss": 1.453,
+      "step": 4180
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013242033951099496,
+      "loss": 1.3253,
+      "step": 4181
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.000132331951975243,
+      "loss": 1.4904,
+      "step": 4182
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013224357936163247,
+      "loss": 1.5006,
+      "step": 4183
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013215522168965124,
+      "loss": 1.4799,
+      "step": 4184
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001320668789787839,
+      "loss": 1.5781,
+      "step": 4185
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001319785512485118,
+      "loss": 1.4123,
+      "step": 4186
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013189023851831292,
+      "loss": 1.3535,
+      "step": 4187
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013180194080766197,
+      "loss": 1.496,
+      "step": 4188
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001317136581360303,
+      "loss": 1.3635,
+      "step": 4189
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013162539052288615,
+      "loss": 1.7743,
+      "step": 4190
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.000131537137987694,
+      "loss": 1.4197,
+      "step": 4191
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013144890054991558,
+      "loss": 1.2918,
+      "step": 4192
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001313606782290087,
+      "loss": 1.2159,
+      "step": 4193
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013127247104442836,
+      "loss": 1.4458,
+      "step": 4194
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013118427901562583,
+      "loss": 1.4626,
+      "step": 4195
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013109610216204923,
+      "loss": 1.5021,
+      "step": 4196
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013100794050314337,
+      "loss": 1.1234,
+      "step": 4197
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013091979405834958,
+      "loss": 1.5599,
+      "step": 4198
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013083166284710593,
+      "loss": 1.3212,
+      "step": 4199
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013074354688884696,
+      "loss": 1.3623,
+      "step": 4200
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001306554462030042,
+      "loss": 1.3424,
+      "step": 4201
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013056736080900537,
+      "loss": 1.5376,
+      "step": 4202
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013047929072627522,
+      "loss": 1.3847,
+      "step": 4203
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013039123597423483,
+      "loss": 1.4111,
+      "step": 4204
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001303031965723021,
+      "loss": 1.2838,
+      "step": 4205
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013021517253989135,
+      "loss": 1.1769,
+      "step": 4206
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001301271638964137,
+      "loss": 1.3277,
+      "step": 4207
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00013003917066127678,
+      "loss": 1.3136,
+      "step": 4208
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012995119285388483,
+      "loss": 1.5382,
+      "step": 4209
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012986323049363863,
+      "loss": 1.4289,
+      "step": 4210
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012977528359993574,
+      "loss": 1.2661,
+      "step": 4211
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012968735219217006,
+      "loss": 1.544,
+      "step": 4212
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012959943628973228,
+      "loss": 1.3432,
+      "step": 4213
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012951153591200963,
+      "loss": 1.6017,
+      "step": 4214
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012942365107838582,
+      "loss": 1.454,
+      "step": 4215
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001293357818082412,
+      "loss": 1.3108,
+      "step": 4216
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012924792812095268,
+      "loss": 1.4731,
+      "step": 4217
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012916009003589374,
+      "loss": 1.4617,
+      "step": 4218
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012907226757243444,
+      "loss": 1.4111,
+      "step": 4219
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001289844607499413,
+      "loss": 1.3247,
+      "step": 4220
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012889666958777748,
+      "loss": 1.6429,
+      "step": 4221
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001288088941053027,
+      "loss": 1.5138,
+      "step": 4222
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012872113432187317,
+      "loss": 1.2302,
+      "step": 4223
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012863339025684172,
+      "loss": 1.3034,
+      "step": 4224
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012854566192955748,
+      "loss": 1.3346,
+      "step": 4225
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012845794935936646,
+      "loss": 1.3913,
+      "step": 4226
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001283702525656109,
+      "loss": 1.4029,
+      "step": 4227
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012828257156762972,
+      "loss": 1.3983,
+      "step": 4228
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012819490638475833,
+      "loss": 1.5253,
+      "step": 4229
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012810725703632858,
+      "loss": 1.446,
+      "step": 4230
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012801962354166898,
+      "loss": 1.3592,
+      "step": 4231
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012793200592010426,
+      "loss": 1.4538,
+      "step": 4232
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0001278444041909561,
+      "loss": 1.5118,
+      "step": 4233
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012775681837354217,
+      "loss": 1.4665,
+      "step": 4234
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012766924848717706,
+      "loss": 1.4096,
+      "step": 4235
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012758169455117147,
+      "loss": 1.6136,
+      "step": 4236
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012749415658483296,
+      "loss": 1.269,
+      "step": 4237
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012740663460746526,
+      "loss": 1.5011,
+      "step": 4238
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012731912863836876,
+      "loss": 1.4092,
+      "step": 4239
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012723163869684018,
+      "loss": 1.3091,
+      "step": 4240
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012714416480217287,
+      "loss": 1.4337,
+      "step": 4241
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012705670697365647,
+      "loss": 1.5291,
+      "step": 4242
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00012696926523057727,
+      "loss": 1.385,
+      "step": 4243
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012688183959221776,
+      "loss": 1.5216,
+      "step": 4244
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012679443007785715,
+      "loss": 1.431,
+      "step": 4245
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012670703670677083,
+      "loss": 1.619,
+      "step": 4246
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012661965949823084,
+      "loss": 1.5466,
+      "step": 4247
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012653229847150565,
+      "loss": 1.5093,
+      "step": 4248
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012644495364585998,
+      "loss": 1.336,
+      "step": 4249
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012635762504055516,
+      "loss": 1.3527,
+      "step": 4250
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001262703126748488,
+      "loss": 1.4285,
+      "step": 4251
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012618301656799509,
+      "loss": 1.5427,
+      "step": 4252
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012609573673924449,
+      "loss": 1.6709,
+      "step": 4253
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012600847320784397,
+      "loss": 1.5119,
+      "step": 4254
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012592122599303673,
+      "loss": 1.3768,
+      "step": 4255
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012583399511406277,
+      "loss": 1.3384,
+      "step": 4256
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012574678059015787,
+      "loss": 1.5113,
+      "step": 4257
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001256595824405549,
+      "loss": 1.236,
+      "step": 4258
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001255724006844825,
+      "loss": 1.5722,
+      "step": 4259
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012548523534116617,
+      "loss": 1.4242,
+      "step": 4260
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012539808642982747,
+      "loss": 1.6749,
+      "step": 4261
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012531095396968453,
+      "loss": 1.1931,
+      "step": 4262
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001252238379799517,
+      "loss": 1.6093,
+      "step": 4263
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012513673847983978,
+      "loss": 1.5424,
+      "step": 4264
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012504965548855608,
+      "loss": 1.5897,
+      "step": 4265
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001249625890253039,
+      "loss": 1.4477,
+      "step": 4266
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012487553910928338,
+      "loss": 1.2188,
+      "step": 4267
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012478850575969042,
+      "loss": 1.3908,
+      "step": 4268
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012470148899571787,
+      "loss": 1.3214,
+      "step": 4269
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012461448883655448,
+      "loss": 1.3832,
+      "step": 4270
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012452750530138565,
+      "loss": 1.3532,
+      "step": 4271
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001244405384093928,
+      "loss": 1.5133,
+      "step": 4272
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.000124353588179754,
+      "loss": 1.238,
+      "step": 4273
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012426665463164335,
+      "loss": 1.4714,
+      "step": 4274
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012417973778423153,
+      "loss": 1.2851,
+      "step": 4275
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012409283765668535,
+      "loss": 1.2859,
+      "step": 4276
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001240059542681681,
+      "loss": 1.4669,
+      "step": 4277
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001239190876378391,
+      "loss": 1.3611,
+      "step": 4278
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012383223778485443,
+      "loss": 1.557,
+      "step": 4279
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012374540472836592,
+      "loss": 1.4775,
+      "step": 4280
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001236585884875221,
+      "loss": 1.4763,
+      "step": 4281
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012357178908146773,
+      "loss": 1.1961,
+      "step": 4282
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012348500652934372,
+      "loss": 1.5508,
+      "step": 4283
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012339824085028735,
+      "loss": 1.4637,
+      "step": 4284
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012331149206343213,
+      "loss": 1.3625,
+      "step": 4285
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012322476018790794,
+      "loss": 1.5032,
+      "step": 4286
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012313804524284082,
+      "loss": 1.5781,
+      "step": 4287
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001230513472473532,
+      "loss": 1.32,
+      "step": 4288
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012296466622056356,
+      "loss": 1.232,
+      "step": 4289
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.000122878002181587,
+      "loss": 1.2693,
+      "step": 4290
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012279135514953443,
+      "loss": 1.6021,
+      "step": 4291
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012270472514351336,
+      "loss": 1.4783,
+      "step": 4292
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001226181121826274,
+      "loss": 1.4317,
+      "step": 4293
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012253151628597642,
+      "loss": 1.323,
+      "step": 4294
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012244493747265645,
+      "loss": 1.4237,
+      "step": 4295
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001223583757617599,
+      "loss": 1.4708,
+      "step": 4296
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001222718311723754,
+      "loss": 1.5456,
+      "step": 4297
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012218530372358756,
+      "loss": 1.3263,
+      "step": 4298
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012209879343447763,
+      "loss": 1.3483,
+      "step": 4299
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012201230032412257,
+      "loss": 1.3215,
+      "step": 4300
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012192582441159606,
+      "loss": 1.4685,
+      "step": 4301
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012183936571596758,
+      "loss": 1.5713,
+      "step": 4302
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012175292425630313,
+      "loss": 1.595,
+      "step": 4303
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001216665000516646,
+      "loss": 1.481,
+      "step": 4304
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012158009312111036,
+      "loss": 1.5155,
+      "step": 4305
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012149370348369476,
+      "loss": 1.4808,
+      "step": 4306
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012140733115846847,
+      "loss": 1.5088,
+      "step": 4307
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012132097616447828,
+      "loss": 1.2301,
+      "step": 4308
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012123463852076717,
+      "loss": 1.4324,
+      "step": 4309
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001211483182463743,
+      "loss": 1.4184,
+      "step": 4310
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00012106201536033501,
+      "loss": 1.5074,
+      "step": 4311
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00012097572988168072,
+      "loss": 1.5277,
+      "step": 4312
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00012088946182943914,
+      "loss": 1.4734,
+      "step": 4313
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00012080321122263414,
+      "loss": 1.5725,
+      "step": 4314
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00012071697808028557,
+      "loss": 1.2035,
+      "step": 4315
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00012063076242140967,
+      "loss": 1.4649,
+      "step": 4316
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00012054456426501852,
+      "loss": 1.3006,
+      "step": 4317
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001204583836301207,
+      "loss": 1.3756,
+      "step": 4318
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00012037222053572059,
+      "loss": 1.3542,
+      "step": 4319
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00012028607500081898,
+      "loss": 1.584,
+      "step": 4320
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00012019994704441253,
+      "loss": 1.6489,
+      "step": 4321
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00012011383668549432,
+      "loss": 1.4246,
+      "step": 4322
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001200277439430532,
+      "loss": 1.4131,
+      "step": 4323
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011994166883607449,
+      "loss": 1.3749,
+      "step": 4324
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011985561138353936,
+      "loss": 1.521,
+      "step": 4325
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001197695716044252,
+      "loss": 1.6786,
+      "step": 4326
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011968354951770544,
+      "loss": 1.2756,
+      "step": 4327
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011959754514234973,
+      "loss": 1.3084,
+      "step": 4328
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011951155849732365,
+      "loss": 1.4947,
+      "step": 4329
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011942558960158896,
+      "loss": 1.3608,
+      "step": 4330
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011933963847410366,
+      "loss": 1.3635,
+      "step": 4331
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011925370513382141,
+      "loss": 1.4182,
+      "step": 4332
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011916778959969248,
+      "loss": 1.4099,
+      "step": 4333
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011908189189066268,
+      "loss": 1.1892,
+      "step": 4334
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011899601202567434,
+      "loss": 1.4964,
+      "step": 4335
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001189101500236656,
+      "loss": 1.3493,
+      "step": 4336
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011882430590357074,
+      "loss": 1.197,
+      "step": 4337
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011873847968432006,
+      "loss": 1.4022,
+      "step": 4338
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011865267138483999,
+      "loss": 1.316,
+      "step": 4339
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011856688102405287,
+      "loss": 1.3347,
+      "step": 4340
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011848110862087729,
+      "loss": 1.455,
+      "step": 4341
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011839535419422763,
+      "loss": 1.3964,
+      "step": 4342
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011830961776301453,
+      "loss": 1.3377,
+      "step": 4343
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011822389934614441,
+      "loss": 1.4097,
+      "step": 4344
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011813819896252019,
+      "loss": 1.4702,
+      "step": 4345
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011805251663104013,
+      "loss": 1.3874,
+      "step": 4346
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001179668523705991,
+      "loss": 1.4534,
+      "step": 4347
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011788120620008776,
+      "loss": 1.6284,
+      "step": 4348
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011779557813839268,
+      "loss": 1.4947,
+      "step": 4349
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011770996820439664,
+      "loss": 1.4208,
+      "step": 4350
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011762437641697826,
+      "loss": 1.6158,
+      "step": 4351
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011753880279501224,
+      "loss": 1.3955,
+      "step": 4352
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011745324735736918,
+      "loss": 1.4783,
+      "step": 4353
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011736771012291594,
+      "loss": 1.3542,
+      "step": 4354
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011728219111051491,
+      "loss": 1.57,
+      "step": 4355
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011719669033902497,
+      "loss": 1.3957,
+      "step": 4356
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011711120782730049,
+      "loss": 1.4009,
+      "step": 4357
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001170257435941923,
+      "loss": 1.1789,
+      "step": 4358
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011694029765854675,
+      "loss": 1.4662,
+      "step": 4359
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011685487003920648,
+      "loss": 1.4846,
+      "step": 4360
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011676946075500984,
+      "loss": 1.2525,
+      "step": 4361
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011668406982479148,
+      "loss": 1.4866,
+      "step": 4362
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011659869726738155,
+      "loss": 1.5372,
+      "step": 4363
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011651334310160651,
+      "loss": 1.3851,
+      "step": 4364
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011642800734628862,
+      "loss": 1.6052,
+      "step": 4365
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011634269002024609,
+      "loss": 1.3262,
+      "step": 4366
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011625739114229316,
+      "loss": 1.3953,
+      "step": 4367
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011617211073123977,
+      "loss": 1.5429,
+      "step": 4368
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011608684880589203,
+      "loss": 1.4319,
+      "step": 4369
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001160016053850518,
+      "loss": 1.3278,
+      "step": 4370
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011591638048751714,
+      "loss": 1.581,
+      "step": 4371
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011583117413208149,
+      "loss": 1.3854,
+      "step": 4372
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011574598633753488,
+      "loss": 1.4162,
+      "step": 4373
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011566081712266265,
+      "loss": 1.3609,
+      "step": 4374
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011557566650624641,
+      "loss": 1.5203,
+      "step": 4375
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011549053450706355,
+      "loss": 1.4895,
+      "step": 4376
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011540542114388739,
+      "loss": 1.3025,
+      "step": 4377
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011532032643548699,
+      "loss": 1.3553,
+      "step": 4378
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00011523525040062746,
+      "loss": 1.3474,
+      "step": 4379
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001151501930580698,
+      "loss": 1.6392,
+      "step": 4380
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001150651544265708,
+      "loss": 1.5818,
+      "step": 4381
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011498013452488323,
+      "loss": 1.5031,
+      "step": 4382
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011489513337175544,
+      "loss": 1.2948,
+      "step": 4383
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011481015098593215,
+      "loss": 1.3366,
+      "step": 4384
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011472518738615345,
+      "loss": 1.4105,
+      "step": 4385
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011464024259115558,
+      "loss": 1.3779,
+      "step": 4386
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001145553166196705,
+      "loss": 1.449,
+      "step": 4387
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011447040949042618,
+      "loss": 1.3783,
+      "step": 4388
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011438552122214617,
+      "loss": 1.5115,
+      "step": 4389
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011430065183355003,
+      "loss": 1.3167,
+      "step": 4390
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001142158013433532,
+      "loss": 1.4183,
+      "step": 4391
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011413096977026693,
+      "loss": 1.3565,
+      "step": 4392
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011404615713299805,
+      "loss": 1.3204,
+      "step": 4393
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001139613634502497,
+      "loss": 1.6272,
+      "step": 4394
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011387658874072038,
+      "loss": 1.5089,
+      "step": 4395
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001137918330231047,
+      "loss": 1.2854,
+      "step": 4396
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011370709631609276,
+      "loss": 1.4682,
+      "step": 4397
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011362237863837086,
+      "loss": 1.3993,
+      "step": 4398
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011353768000862102,
+      "loss": 1.3975,
+      "step": 4399
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011345300044552072,
+      "loss": 1.3953,
+      "step": 4400
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011336833996774356,
+      "loss": 1.446,
+      "step": 4401
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001132836985939589,
+      "loss": 1.4509,
+      "step": 4402
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011319907634283189,
+      "loss": 1.5127,
+      "step": 4403
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011311447323302315,
+      "loss": 1.5164,
+      "step": 4404
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011302988928318967,
+      "loss": 1.2968,
+      "step": 4405
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011294532451198367,
+      "loss": 1.3739,
+      "step": 4406
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011286077893805335,
+      "loss": 1.3586,
+      "step": 4407
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011277625258004278,
+      "loss": 1.6629,
+      "step": 4408
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011269174545659171,
+      "loss": 1.5977,
+      "step": 4409
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001126072575863355,
+      "loss": 1.2847,
+      "step": 4410
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011252278898790546,
+      "loss": 1.2718,
+      "step": 4411
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011243833967992858,
+      "loss": 1.5737,
+      "step": 4412
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011235390968102768,
+      "loss": 1.6657,
+      "step": 4413
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011226949900982111,
+      "loss": 1.3434,
+      "step": 4414
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011218510768492306,
+      "loss": 1.4893,
+      "step": 4415
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011210073572494371,
+      "loss": 1.4993,
+      "step": 4416
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011201638314848854,
+      "loss": 1.1413,
+      "step": 4417
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011193204997415905,
+      "loss": 1.3289,
+      "step": 4418
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011184773622055234,
+      "loss": 1.5738,
+      "step": 4419
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011176344190626133,
+      "loss": 1.2357,
+      "step": 4420
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011167916704987442,
+      "loss": 1.3426,
+      "step": 4421
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011159491166997602,
+      "loss": 1.3703,
+      "step": 4422
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011151067578514607,
+      "loss": 1.4258,
+      "step": 4423
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011142645941396028,
+      "loss": 1.5245,
+      "step": 4424
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011134226257498994,
+      "loss": 1.6451,
+      "step": 4425
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011125808528680215,
+      "loss": 1.3878,
+      "step": 4426
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011117392756795966,
+      "loss": 1.2096,
+      "step": 4427
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011108978943702104,
+      "loss": 1.2473,
+      "step": 4428
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011100567091254012,
+      "loss": 1.164,
+      "step": 4429
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011092157201306702,
+      "loss": 1.4803,
+      "step": 4430
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011083749275714699,
+      "loss": 1.271,
+      "step": 4431
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011075343316332125,
+      "loss": 1.4944,
+      "step": 4432
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011066939325012657,
+      "loss": 1.3548,
+      "step": 4433
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011058537303609546,
+      "loss": 1.7251,
+      "step": 4434
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011050137253975605,
+      "loss": 1.4626,
+      "step": 4435
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011041739177963193,
+      "loss": 1.3995,
+      "step": 4436
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001103334307742428,
+      "loss": 1.3956,
+      "step": 4437
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001102494895421035,
+      "loss": 1.3831,
+      "step": 4438
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001101655681017248,
+      "loss": 1.4895,
+      "step": 4439
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00011008166647161306,
+      "loss": 1.4063,
+      "step": 4440
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00010999778467027027,
+      "loss": 1.3947,
+      "step": 4441
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00010991392271619391,
+      "loss": 1.2596,
+      "step": 4442
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001098300806278773,
+      "loss": 1.2921,
+      "step": 4443
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00010974625842380921,
+      "loss": 1.4103,
+      "step": 4444
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00010966245612247422,
+      "loss": 1.3427,
+      "step": 4445
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00010957867374235222,
+      "loss": 1.5719,
+      "step": 4446
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00010949491130191893,
+      "loss": 1.4693,
+      "step": 4447
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010941116881964571,
+      "loss": 1.5452,
+      "step": 4448
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010932744631399931,
+      "loss": 1.5387,
+      "step": 4449
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010924374380344231,
+      "loss": 1.3748,
+      "step": 4450
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001091600613064327,
+      "loss": 1.5915,
+      "step": 4451
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010907639884142423,
+      "loss": 1.5793,
+      "step": 4452
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010899275642686593,
+      "loss": 1.5219,
+      "step": 4453
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010890913408120273,
+      "loss": 1.4268,
+      "step": 4454
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010882553182287498,
+      "loss": 1.4562,
+      "step": 4455
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010874194967031874,
+      "loss": 1.4731,
+      "step": 4456
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010865838764196534,
+      "loss": 1.3013,
+      "step": 4457
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010857484575624194,
+      "loss": 1.3968,
+      "step": 4458
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001084913240315712,
+      "loss": 1.5833,
+      "step": 4459
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010840782248637136,
+      "loss": 1.3098,
+      "step": 4460
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010832434113905594,
+      "loss": 1.395,
+      "step": 4461
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010824088000803452,
+      "loss": 1.6093,
+      "step": 4462
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001081574391117117,
+      "loss": 1.4906,
+      "step": 4463
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010807401846848795,
+      "loss": 1.2107,
+      "step": 4464
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010799061809675915,
+      "loss": 1.5187,
+      "step": 4465
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010790723801491673,
+      "loss": 1.4462,
+      "step": 4466
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010782387824134772,
+      "loss": 1.2378,
+      "step": 4467
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010774053879443437,
+      "loss": 1.494,
+      "step": 4468
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010765721969255498,
+      "loss": 1.5209,
+      "step": 4469
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010757392095408283,
+      "loss": 1.4835,
+      "step": 4470
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010749064259738713,
+      "loss": 1.2914,
+      "step": 4471
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010740738464083207,
+      "loss": 1.5449,
+      "step": 4472
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010732414710277807,
+      "loss": 1.3469,
+      "step": 4473
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010724093000158042,
+      "loss": 1.5858,
+      "step": 4474
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010715773335559017,
+      "loss": 1.5603,
+      "step": 4475
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010707455718315382,
+      "loss": 1.5611,
+      "step": 4476
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010699140150261347,
+      "loss": 1.3506,
+      "step": 4477
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010690826633230644,
+      "loss": 1.2353,
+      "step": 4478
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001068251516905657,
+      "loss": 1.5772,
+      "step": 4479
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010674205759571974,
+      "loss": 1.4742,
+      "step": 4480
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001066589840660925,
+      "loss": 1.3866,
+      "step": 4481
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010657593112000306,
+      "loss": 1.6228,
+      "step": 4482
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010649289877576654,
+      "loss": 1.4728,
+      "step": 4483
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010640988705169317,
+      "loss": 1.3626,
+      "step": 4484
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010632689596608854,
+      "loss": 1.5099,
+      "step": 4485
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010624392553725386,
+      "loss": 1.321,
+      "step": 4486
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010616097578348578,
+      "loss": 1.2419,
+      "step": 4487
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010607804672307647,
+      "loss": 1.4169,
+      "step": 4488
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010599513837431319,
+      "loss": 1.6238,
+      "step": 4489
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010591225075547901,
+      "loss": 1.2601,
+      "step": 4490
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010582938388485226,
+      "loss": 1.3912,
+      "step": 4491
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010574653778070685,
+      "loss": 1.1897,
+      "step": 4492
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010566371246131166,
+      "loss": 1.3734,
+      "step": 4493
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001055809079449317,
+      "loss": 1.5745,
+      "step": 4494
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010549812424982671,
+      "loss": 1.3075,
+      "step": 4495
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010541536139425226,
+      "loss": 1.3719,
+      "step": 4496
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010533261939645916,
+      "loss": 1.3891,
+      "step": 4497
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010524989827469373,
+      "loss": 1.3517,
+      "step": 4498
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010516719804719747,
+      "loss": 1.5496,
+      "step": 4499
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010508451873220751,
+      "loss": 1.3229,
+      "step": 4500
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010500186034795625,
+      "loss": 1.5622,
+      "step": 4501
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010491922291267149,
+      "loss": 1.4306,
+      "step": 4502
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010483660644457649,
+      "loss": 1.0458,
+      "step": 4503
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010475401096188962,
+      "loss": 1.4169,
+      "step": 4504
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010467143648282509,
+      "loss": 1.2511,
+      "step": 4505
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010458888302559197,
+      "loss": 1.4455,
+      "step": 4506
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010450635060839502,
+      "loss": 1.4126,
+      "step": 4507
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010442383924943428,
+      "loss": 1.4329,
+      "step": 4508
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001043413489669052,
+      "loss": 1.3134,
+      "step": 4509
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010425887977899833,
+      "loss": 1.4351,
+      "step": 4510
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010417643170389988,
+      "loss": 1.7247,
+      "step": 4511
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010409400475979127,
+      "loss": 1.3825,
+      "step": 4512
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010401159896484935,
+      "loss": 1.4659,
+      "step": 4513
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010392921433724596,
+      "loss": 1.368,
+      "step": 4514
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00010384685089514888,
+      "loss": 1.366,
+      "step": 4515
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010376450865672063,
+      "loss": 1.3127,
+      "step": 4516
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010368218764011937,
+      "loss": 1.6328,
+      "step": 4517
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010359988786349852,
+      "loss": 1.3049,
+      "step": 4518
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010351760934500683,
+      "loss": 1.629,
+      "step": 4519
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010343535210278839,
+      "loss": 1.5206,
+      "step": 4520
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010335311615498237,
+      "loss": 1.4297,
+      "step": 4521
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010327090151972356,
+      "loss": 1.3456,
+      "step": 4522
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010318870821514186,
+      "loss": 1.4205,
+      "step": 4523
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010310653625936262,
+      "loss": 1.4095,
+      "step": 4524
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010302438567050612,
+      "loss": 1.4072,
+      "step": 4525
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010294225646668851,
+      "loss": 1.4491,
+      "step": 4526
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001028601486660207,
+      "loss": 1.2713,
+      "step": 4527
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010277806228660914,
+      "loss": 1.4811,
+      "step": 4528
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010269599734655546,
+      "loss": 1.4966,
+      "step": 4529
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010261395386395674,
+      "loss": 1.379,
+      "step": 4530
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010253193185690499,
+      "loss": 1.6439,
+      "step": 4531
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001024499313434878,
+      "loss": 1.298,
+      "step": 4532
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010236795234178785,
+      "loss": 1.4191,
+      "step": 4533
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010228599486988317,
+      "loss": 1.4204,
+      "step": 4534
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010220405894584706,
+      "loss": 1.4786,
+      "step": 4535
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001021221445877478,
+      "loss": 1.509,
+      "step": 4536
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001020402518136494,
+      "loss": 1.45,
+      "step": 4537
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010195838064161063,
+      "loss": 1.4684,
+      "step": 4538
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010187653108968578,
+      "loss": 1.5232,
+      "step": 4539
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010179470317592428,
+      "loss": 1.5321,
+      "step": 4540
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010171289691837085,
+      "loss": 1.4285,
+      "step": 4541
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010163111233506528,
+      "loss": 1.2188,
+      "step": 4542
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010154934944404275,
+      "loss": 1.398,
+      "step": 4543
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010146760826333359,
+      "loss": 1.3143,
+      "step": 4544
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010138588881096342,
+      "loss": 1.3093,
+      "step": 4545
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010130419110495286,
+      "loss": 1.5098,
+      "step": 4546
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010122251516331792,
+      "loss": 1.3947,
+      "step": 4547
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010114086100406974,
+      "loss": 1.5356,
+      "step": 4548
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010105922864521478,
+      "loss": 1.5004,
+      "step": 4549
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010097761810475437,
+      "loss": 1.3127,
+      "step": 4550
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010089602940068545,
+      "loss": 1.5857,
+      "step": 4551
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010081446255099996,
+      "loss": 1.6307,
+      "step": 4552
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010073291757368481,
+      "loss": 1.362,
+      "step": 4553
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001006513944867224,
+      "loss": 1.7285,
+      "step": 4554
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010056989330809018,
+      "loss": 1.5675,
+      "step": 4555
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010048841405576083,
+      "loss": 1.21,
+      "step": 4556
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010040695674770191,
+      "loss": 1.333,
+      "step": 4557
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010032552140187666,
+      "loss": 1.1945,
+      "step": 4558
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010024410803624298,
+      "loss": 1.4449,
+      "step": 4559
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010016271666875426,
+      "loss": 1.6169,
+      "step": 4560
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010008134731735865,
+      "loss": 1.3114,
+      "step": 4561
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00010000000000000005,
+      "loss": 1.3881,
+      "step": 4562
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.99186747346169e-05,
+      "loss": 1.4153,
+      "step": 4563
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.983737153914309e-05,
+      "loss": 1.5848,
+      "step": 4564
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.975609043150762e-05,
+      "loss": 1.4073,
+      "step": 4565
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.967483142963463e-05,
+      "loss": 1.4268,
+      "step": 4566
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.959359455144317e-05,
+      "loss": 1.5223,
+      "step": 4567
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.95123798148476e-05,
+      "loss": 1.2617,
+      "step": 4568
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.943118723775761e-05,
+      "loss": 1.317,
+      "step": 4569
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.935001683807752e-05,
+      "loss": 1.3448,
+      "step": 4570
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.926886863370708e-05,
+      "loss": 1.4521,
+      "step": 4571
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.918774264254107e-05,
+      "loss": 1.5791,
+      "step": 4572
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.910663888246947e-05,
+      "loss": 1.3762,
+      "step": 4573
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.902555737137709e-05,
+      "loss": 1.2235,
+      "step": 4574
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.89444981271441e-05,
+      "loss": 1.363,
+      "step": 4575
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.886346116764565e-05,
+      "loss": 1.0464,
+      "step": 4576
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.878244651075208e-05,
+      "loss": 1.5581,
+      "step": 4577
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.870145417432855e-05,
+      "loss": 1.3916,
+      "step": 4578
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.862048417623555e-05,
+      "loss": 1.3682,
+      "step": 4579
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.853953653432858e-05,
+      "loss": 1.2406,
+      "step": 4580
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.845861126645825e-05,
+      "loss": 1.2536,
+      "step": 4581
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.837770839046998e-05,
+      "loss": 1.2856,
+      "step": 4582
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.829682792420472e-05,
+      "loss": 1.5128,
+      "step": 4583
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.821596988549801e-05,
+      "loss": 1.3469,
+      "step": 4584
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.813513429218074e-05,
+      "loss": 1.5052,
+      "step": 4585
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.805432116207868e-05,
+      "loss": 1.3618,
+      "step": 4586
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.797353051301278e-05,
+      "loss": 1.4113,
+      "step": 4587
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.789276236279907e-05,
+      "loss": 1.5057,
+      "step": 4588
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.781201672924834e-05,
+      "loss": 1.5877,
+      "step": 4589
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.773129363016664e-05,
+      "loss": 1.4269,
+      "step": 4590
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.765059308335507e-05,
+      "loss": 1.3257,
+      "step": 4591
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.756991510660975e-05,
+      "loss": 1.3438,
+      "step": 4592
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.748925971772152e-05,
+      "loss": 1.3192,
+      "step": 4593
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.740862693447679e-05,
+      "loss": 1.5617,
+      "step": 4594
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.732801677465648e-05,
+      "loss": 1.5154,
+      "step": 4595
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.72474292560368e-05,
+      "loss": 1.4035,
+      "step": 4596
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.716686439638884e-05,
+      "loss": 1.5343,
+      "step": 4597
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.708632221347889e-05,
+      "loss": 1.4605,
+      "step": 4598
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.700580272506789e-05,
+      "loss": 1.4544,
+      "step": 4599
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.692530594891208e-05,
+      "loss": 1.3357,
+      "step": 4600
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.684483190276256e-05,
+      "loss": 1.5918,
+      "step": 4601
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.676438060436549e-05,
+      "loss": 1.5354,
+      "step": 4602
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.668395207146192e-05,
+      "loss": 1.4119,
+      "step": 4603
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.660354632178799e-05,
+      "loss": 1.2204,
+      "step": 4604
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.652316337307476e-05,
+      "loss": 1.6531,
+      "step": 4605
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.644280324304818e-05,
+      "loss": 1.6183,
+      "step": 4606
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.636246594942926e-05,
+      "loss": 1.4105,
+      "step": 4607
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.6282151509934e-05,
+      "loss": 1.3843,
+      "step": 4608
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.620185994227338e-05,
+      "loss": 1.5129,
+      "step": 4609
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.612159126415314e-05,
+      "loss": 1.5474,
+      "step": 4610
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.604134549327415e-05,
+      "loss": 1.6303,
+      "step": 4611
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.596112264733223e-05,
+      "loss": 1.3689,
+      "step": 4612
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.588092274401817e-05,
+      "loss": 1.298,
+      "step": 4613
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.580074580101739e-05,
+      "loss": 1.3543,
+      "step": 4614
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.572059183601078e-05,
+      "loss": 1.5351,
+      "step": 4615
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.564046086667368e-05,
+      "loss": 1.4277,
+      "step": 4616
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.55603529106766e-05,
+      "loss": 1.4559,
+      "step": 4617
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.548026798568494e-05,
+      "loss": 1.4988,
+      "step": 4618
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.540020610935902e-05,
+      "loss": 1.4448,
+      "step": 4619
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.532016729935414e-05,
+      "loss": 1.3787,
+      "step": 4620
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.524015157332025e-05,
+      "loss": 1.5542,
+      "step": 4621
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.51601589489025e-05,
+      "loss": 1.411,
+      "step": 4622
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.508018944374086e-05,
+      "loss": 1.5527,
+      "step": 4623
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.500024307547024e-05,
+      "loss": 1.345,
+      "step": 4624
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.492031986172017e-05,
+      "loss": 1.6383,
+      "step": 4625
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.484041982011558e-05,
+      "loss": 1.6149,
+      "step": 4626
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.476054296827583e-05,
+      "loss": 1.4425,
+      "step": 4627
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.468068932381538e-05,
+      "loss": 1.4708,
+      "step": 4628
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.460085890434353e-05,
+      "loss": 1.3772,
+      "step": 4629
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.45210517274646e-05,
+      "loss": 1.1166,
+      "step": 4630
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.444126781077742e-05,
+      "loss": 1.2327,
+      "step": 4631
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.436150717187604e-05,
+      "loss": 1.205,
+      "step": 4632
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.428176982834926e-05,
+      "loss": 1.5731,
+      "step": 4633
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.420205579778079e-05,
+      "loss": 1.3271,
+      "step": 4634
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.412236509774901e-05,
+      "loss": 1.3388,
+      "step": 4635
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.404269774582731e-05,
+      "loss": 1.2963,
+      "step": 4636
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.39630537595841e-05,
+      "loss": 1.4888,
+      "step": 4637
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.388343315658225e-05,
+      "loss": 1.4618,
+      "step": 4638
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.380383595437978e-05,
+      "loss": 1.6773,
+      "step": 4639
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.372426217052941e-05,
+      "loss": 1.447,
+      "step": 4640
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.364471182257879e-05,
+      "loss": 1.4646,
+      "step": 4641
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.356518492807024e-05,
+      "loss": 1.4545,
+      "step": 4642
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.348568150454106e-05,
+      "loss": 1.4589,
+      "step": 4643
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.340620156952333e-05,
+      "loss": 1.4339,
+      "step": 4644
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.332674514054403e-05,
+      "loss": 1.3747,
+      "step": 4645
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.324731223512462e-05,
+      "loss": 1.4208,
+      "step": 4646
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.316790287078195e-05,
+      "loss": 1.3849,
+      "step": 4647
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.308851706502712e-05,
+      "loss": 1.4925,
+      "step": 4648
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.300915483536635e-05,
+      "loss": 1.448,
+      "step": 4649
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.292981619930057e-05,
+      "loss": 1.7085,
+      "step": 4650
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.285050117432559e-05,
+      "loss": 1.3412,
+      "step": 4651
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.277120977793178e-05,
+      "loss": 1.7296,
+      "step": 4652
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.269194202760459e-05,
+      "loss": 1.4695,
+      "step": 4653
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.261269794082406e-05,
+      "loss": 1.4827,
+      "step": 4654
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.253347753506511e-05,
+      "loss": 1.418,
+      "step": 4655
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.245428082779745e-05,
+      "loss": 1.3794,
+      "step": 4656
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.237510783648536e-05,
+      "loss": 1.4865,
+      "step": 4657
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.229595857858826e-05,
+      "loss": 1.3423,
+      "step": 4658
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.221683307155995e-05,
+      "loss": 1.3367,
+      "step": 4659
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.213773133284924e-05,
+      "loss": 1.3684,
+      "step": 4660
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.205865337989963e-05,
+      "loss": 1.3492,
+      "step": 4661
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.197959923014942e-05,
+      "loss": 1.4585,
+      "step": 4662
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.190056890103149e-05,
+      "loss": 1.6022,
+      "step": 4663
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.182156240997365e-05,
+      "loss": 1.5136,
+      "step": 4664
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.174257977439842e-05,
+      "loss": 1.3992,
+      "step": 4665
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.166362101172305e-05,
+      "loss": 1.5672,
+      "step": 4666
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.15846861393594e-05,
+      "loss": 1.3759,
+      "step": 4667
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.150577517471424e-05,
+      "loss": 1.5001,
+      "step": 4668
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.142688813518896e-05,
+      "loss": 1.3267,
+      "step": 4669
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.134802503817976e-05,
+      "loss": 1.4039,
+      "step": 4670
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.126918590107749e-05,
+      "loss": 1.3663,
+      "step": 4671
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.119037074126775e-05,
+      "loss": 1.4265,
+      "step": 4672
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.111157957613088e-05,
+      "loss": 1.3884,
+      "step": 4673
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.103281242304176e-05,
+      "loss": 1.4896,
+      "step": 4674
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.095406929937021e-05,
+      "loss": 1.466,
+      "step": 4675
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.087535022248057e-05,
+      "loss": 1.5143,
+      "step": 4676
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.079665520973211e-05,
+      "loss": 1.1977,
+      "step": 4677
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.071798427847835e-05,
+      "loss": 1.5323,
+      "step": 4678
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.06393374460681e-05,
+      "loss": 1.6486,
+      "step": 4679
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.056071472984433e-05,
+      "loss": 1.5469,
+      "step": 4680
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.048211614714503e-05,
+      "loss": 1.5127,
+      "step": 4681
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.040354171530255e-05,
+      "loss": 1.3972,
+      "step": 4682
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.032499145164435e-05,
+      "loss": 1.2695,
+      "step": 4683
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.024646537349213e-05,
+      "loss": 1.4949,
+      "step": 4684
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.016796349816253e-05,
+      "loss": 1.2027,
+      "step": 4685
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.008948584296673e-05,
+      "loss": 1.4575,
+      "step": 4686
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.00110324252106e-05,
+      "loss": 1.4175,
+      "step": 4687
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.993260326219477e-05,
+      "loss": 1.4553,
+      "step": 4688
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.98541983712142e-05,
+      "loss": 1.3483,
+      "step": 4689
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.977581776955899e-05,
+      "loss": 1.4818,
+      "step": 4690
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.969746147451339e-05,
+      "loss": 1.4045,
+      "step": 4691
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.96191295033566e-05,
+      "loss": 1.2975,
+      "step": 4692
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.954082187336236e-05,
+      "loss": 1.483,
+      "step": 4693
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.946253860179911e-05,
+      "loss": 1.4083,
+      "step": 4694
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.938427970592975e-05,
+      "loss": 1.4331,
+      "step": 4695
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.930604520301196e-05,
+      "loss": 1.496,
+      "step": 4696
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.922783511029798e-05,
+      "loss": 1.4267,
+      "step": 4697
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.914964944503476e-05,
+      "loss": 1.397,
+      "step": 4698
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.907148822446365e-05,
+      "loss": 1.3054,
+      "step": 4699
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.899335146582082e-05,
+      "loss": 1.3976,
+      "step": 4700
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.891523918633695e-05,
+      "loss": 1.2665,
+      "step": 4701
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.883715140323743e-05,
+      "loss": 1.2853,
+      "step": 4702
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.875908813374194e-05,
+      "loss": 1.407,
+      "step": 4703
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.868104939506519e-05,
+      "loss": 1.3584,
+      "step": 4704
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.860303520441624e-05,
+      "loss": 1.5266,
+      "step": 4705
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.852504557899866e-05,
+      "loss": 1.3297,
+      "step": 4706
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.844708053601072e-05,
+      "loss": 1.4414,
+      "step": 4707
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.83691400926453e-05,
+      "loss": 1.4162,
+      "step": 4708
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.829122426608984e-05,
+      "loss": 1.4923,
+      "step": 4709
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.821333307352622e-05,
+      "loss": 1.2909,
+      "step": 4710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.813546653213102e-05,
+      "loss": 1.5229,
+      "step": 4711
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.805762465907539e-05,
+      "loss": 1.6438,
+      "step": 4712
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.797980747152503e-05,
+      "loss": 1.3667,
+      "step": 4713
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.790201498663997e-05,
+      "loss": 1.2919,
+      "step": 4714
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.78242472215753e-05,
+      "loss": 1.3884,
+      "step": 4715
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.774650419348007e-05,
+      "loss": 1.3874,
+      "step": 4716
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.766878591949829e-05,
+      "loss": 1.5619,
+      "step": 4717
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.759109241676835e-05,
+      "loss": 1.4253,
+      "step": 4718
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.751342370242329e-05,
+      "loss": 1.3493,
+      "step": 4719
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.743577979359041e-05,
+      "loss": 1.454,
+      "step": 4720
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.735816070739176e-05,
+      "loss": 1.264,
+      "step": 4721
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.728056646094409e-05,
+      "loss": 1.3253,
+      "step": 4722
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.720299707135824e-05,
+      "loss": 1.5584,
+      "step": 4723
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.712545255573996e-05,
+      "loss": 1.3596,
+      "step": 4724
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.704793293118911e-05,
+      "loss": 1.3688,
+      "step": 4725
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.69704382148006e-05,
+      "loss": 1.4926,
+      "step": 4726
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.689296842366332e-05,
+      "loss": 1.4962,
+      "step": 4727
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.681552357486096e-05,
+      "loss": 1.3822,
+      "step": 4728
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.673810368547163e-05,
+      "loss": 1.1469,
+      "step": 4729
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.666070877256806e-05,
+      "loss": 1.3749,
+      "step": 4730
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.658333885321717e-05,
+      "loss": 1.3266,
+      "step": 4731
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.650599394448065e-05,
+      "loss": 1.4069,
+      "step": 4732
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.642867406341458e-05,
+      "loss": 1.3197,
+      "step": 4733
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.635137922706955e-05,
+      "loss": 1.49,
+      "step": 4734
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.627410945249044e-05,
+      "loss": 1.3015,
+      "step": 4735
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.619686475671703e-05,
+      "loss": 1.4307,
+      "step": 4736
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.611964515678308e-05,
+      "loss": 1.3564,
+      "step": 4737
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.604245066971712e-05,
+      "loss": 1.3471,
+      "step": 4738
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.596528131254205e-05,
+      "loss": 1.2145,
+      "step": 4739
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.588813710227521e-05,
+      "loss": 1.2201,
+      "step": 4740
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.581101805592855e-05,
+      "loss": 1.4895,
+      "step": 4741
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.573392419050818e-05,
+      "loss": 1.5092,
+      "step": 4742
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.565685552301488e-05,
+      "loss": 1.5781,
+      "step": 4743
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.557981207044381e-05,
+      "loss": 1.5071,
+      "step": 4744
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.550279384978465e-05,
+      "loss": 1.5976,
+      "step": 4745
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.542580087802125e-05,
+      "loss": 1.4705,
+      "step": 4746
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.534883317213233e-05,
+      "loss": 1.3828,
+      "step": 4747
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.52718907490906e-05,
+      "loss": 1.3728,
+      "step": 4748
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.519497362586344e-05,
+      "loss": 1.4741,
+      "step": 4749
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.511808181941261e-05,
+      "loss": 1.5449,
+      "step": 4750
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.504121534669431e-05,
+      "loss": 1.325,
+      "step": 4751
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.496437422465904e-05,
+      "loss": 1.5082,
+      "step": 4752
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.488755847025178e-05,
+      "loss": 1.3657,
+      "step": 4753
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.481076810041198e-05,
+      "loss": 1.3965,
+      "step": 4754
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.473400313207338e-05,
+      "loss": 1.1661,
+      "step": 4755
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.46572635821643e-05,
+      "loss": 1.4627,
+      "step": 4756
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.45805494676071e-05,
+      "loss": 1.4958,
+      "step": 4757
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.450386080531901e-05,
+      "loss": 1.4834,
+      "step": 4758
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.442719761221118e-05,
+      "loss": 1.3746,
+      "step": 4759
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.435055990518948e-05,
+      "loss": 1.3961,
+      "step": 4760
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.4273947701154e-05,
+      "loss": 1.2447,
+      "step": 4761
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.419736101699933e-05,
+      "loss": 1.3042,
+      "step": 4762
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.41207998696142e-05,
+      "loss": 1.5107,
+      "step": 4763
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.404426427588192e-05,
+      "loss": 1.4441,
+      "step": 4764
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.39677542526801e-05,
+      "loss": 1.7535,
+      "step": 4765
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.38912698168808e-05,
+      "loss": 1.3072,
+      "step": 4766
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.381481098535013e-05,
+      "loss": 1.3282,
+      "step": 4767
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.373837777494904e-05,
+      "loss": 0.9761,
+      "step": 4768
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.366197020253237e-05,
+      "loss": 1.3591,
+      "step": 4769
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.358558828494962e-05,
+      "loss": 1.2575,
+      "step": 4770
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.350923203904433e-05,
+      "loss": 1.4512,
+      "step": 4771
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.343290148165474e-05,
+      "loss": 1.372,
+      "step": 4772
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.335659662961326e-05,
+      "loss": 1.3759,
+      "step": 4773
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.32803174997465e-05,
+      "loss": 1.4949,
+      "step": 4774
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.320406410887553e-05,
+      "loss": 1.6036,
+      "step": 4775
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.31278364738158e-05,
+      "loss": 1.4034,
+      "step": 4776
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.3051634611377e-05,
+      "loss": 1.3056,
+      "step": 4777
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.2975458538363e-05,
+      "loss": 1.339,
+      "step": 4778
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.289930827157238e-05,
+      "loss": 1.35,
+      "step": 4779
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.282318382779756e-05,
+      "loss": 1.5169,
+      "step": 4780
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.27470852238256e-05,
+      "loss": 1.2921,
+      "step": 4781
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.26710124764377e-05,
+      "loss": 1.368,
+      "step": 4782
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.259496560240948e-05,
+      "loss": 1.6407,
+      "step": 4783
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.251894461851063e-05,
+      "loss": 1.3528,
+      "step": 4784
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.24429495415054e-05,
+      "loss": 1.5099,
+      "step": 4785
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.236698038815213e-05,
+      "loss": 1.4574,
+      "step": 4786
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.229103717520366e-05,
+      "loss": 1.4044,
+      "step": 4787
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.22151199194068e-05,
+      "loss": 1.2496,
+      "step": 4788
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.213922863750279e-05,
+      "loss": 1.5378,
+      "step": 4789
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.206336334622738e-05,
+      "loss": 1.4656,
+      "step": 4790
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.198752406231014e-05,
+      "loss": 1.5516,
+      "step": 4791
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.191171080247524e-05,
+      "loss": 1.5614,
+      "step": 4792
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.183592358344096e-05,
+      "loss": 1.5182,
+      "step": 4793
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.176016242191998e-05,
+      "loss": 1.5513,
+      "step": 4794
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.168442733461896e-05,
+      "loss": 1.3106,
+      "step": 4795
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.160871833823909e-05,
+      "loss": 1.4495,
+      "step": 4796
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.153303544947566e-05,
+      "loss": 1.4153,
+      "step": 4797
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.145737868501835e-05,
+      "loss": 1.4624,
+      "step": 4798
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.138174806155081e-05,
+      "loss": 1.4336,
+      "step": 4799
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.130614359575115e-05,
+      "loss": 1.3756,
+      "step": 4800
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.123056530429168e-05,
+      "loss": 1.3711,
+      "step": 4801
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.115501320383893e-05,
+      "loss": 1.5583,
+      "step": 4802
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.107948731105346e-05,
+      "loss": 1.0864,
+      "step": 4803
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.100398764259047e-05,
+      "loss": 1.3567,
+      "step": 4804
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.092851421509895e-05,
+      "loss": 1.3901,
+      "step": 4805
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.085306704522235e-05,
+      "loss": 1.5322,
+      "step": 4806
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.077764614959824e-05,
+      "loss": 1.535,
+      "step": 4807
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.070225154485842e-05,
+      "loss": 1.5552,
+      "step": 4808
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.062688324762898e-05,
+      "loss": 1.5405,
+      "step": 4809
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.055154127452986e-05,
+      "loss": 1.3429,
+      "step": 4810
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.047622564217578e-05,
+      "loss": 1.4801,
+      "step": 4811
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.040093636717511e-05,
+      "loss": 1.4495,
+      "step": 4812
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.03256734661307e-05,
+      "loss": 1.5616,
+      "step": 4813
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.025043695563948e-05,
+      "loss": 1.5101,
+      "step": 4814
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.017522685229266e-05,
+      "loss": 1.3003,
+      "step": 4815
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.010004317267543e-05,
+      "loss": 1.5907,
+      "step": 4816
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.002488593336735e-05,
+      "loss": 1.4855,
+      "step": 4817
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.994975515094205e-05,
+      "loss": 1.2606,
+      "step": 4818
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.987465084196744e-05,
+      "loss": 1.5447,
+      "step": 4819
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.979957302300536e-05,
+      "loss": 1.3762,
+      "step": 4820
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.972452171061202e-05,
+      "loss": 1.482,
+      "step": 4821
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.964949692133772e-05,
+      "loss": 1.3202,
+      "step": 4822
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.957449867172693e-05,
+      "loss": 1.3503,
+      "step": 4823
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.94995269783182e-05,
+      "loss": 1.4354,
+      "step": 4824
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.942458185764429e-05,
+      "loss": 1.1771,
+      "step": 4825
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.934966332623217e-05,
+      "loss": 1.218,
+      "step": 4826
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.927477140060268e-05,
+      "loss": 1.467,
+      "step": 4827
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.919990609727106e-05,
+      "loss": 1.4878,
+      "step": 4828
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.912506743274659e-05,
+      "loss": 1.5256,
+      "step": 4829
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.905025542353274e-05,
+      "loss": 1.2958,
+      "step": 4830
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.897547008612691e-05,
+      "loss": 1.1762,
+      "step": 4831
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.89007114370208e-05,
+      "loss": 1.5237,
+      "step": 4832
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.882597949270017e-05,
+      "loss": 1.2508,
+      "step": 4833
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.875127426964495e-05,
+      "loss": 1.6476,
+      "step": 4834
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.867659578432891e-05,
+      "loss": 1.2838,
+      "step": 4835
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.860194405322043e-05,
+      "loss": 1.3945,
+      "step": 4836
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.852731909278146e-05,
+      "loss": 1.3428,
+      "step": 4837
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.845272091946836e-05,
+      "loss": 1.7628,
+      "step": 4838
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.83781495497315e-05,
+      "loss": 1.3397,
+      "step": 4839
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.830360500001535e-05,
+      "loss": 1.2342,
+      "step": 4840
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.822908728675851e-05,
+      "loss": 1.4522,
+      "step": 4841
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.81545964263934e-05,
+      "loss": 1.5593,
+      "step": 4842
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.8080132435347e-05,
+      "loss": 1.5156,
+      "step": 4843
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.80056953300399e-05,
+      "loss": 1.2345,
+      "step": 4844
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.793128512688708e-05,
+      "loss": 1.5705,
+      "step": 4845
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.785690184229726e-05,
+      "loss": 1.5304,
+      "step": 4846
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.778254549267369e-05,
+      "loss": 1.4002,
+      "step": 4847
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.77082160944132e-05,
+      "loss": 1.5076,
+      "step": 4848
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.7633913663907e-05,
+      "loss": 1.4164,
+      "step": 4849
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.75596382175402e-05,
+      "loss": 1.3455,
+      "step": 4850
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.748538977169209e-05,
+      "loss": 1.284,
+      "step": 4851
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.741116834273577e-05,
+      "loss": 1.344,
+      "step": 4852
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.73369739470386e-05,
+      "loss": 1.7534,
+      "step": 4853
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 7.726280660096192e-05,
+      "loss": 1.2735,
+      "step": 4854
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.718866632086116e-05,
+      "loss": 1.328,
+      "step": 4855
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.711455312308554e-05,
+      "loss": 1.3594,
+      "step": 4856
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.704046702397865e-05,
+      "loss": 1.7196,
+      "step": 4857
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.696640803987794e-05,
+      "loss": 1.3558,
+      "step": 4858
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.689237618711475e-05,
+      "loss": 1.3221,
+      "step": 4859
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.681837148201465e-05,
+      "loss": 1.2908,
+      "step": 4860
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.674439394089712e-05,
+      "loss": 1.3251,
+      "step": 4861
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.667044358007576e-05,
+      "loss": 1.3685,
+      "step": 4862
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.659652041585792e-05,
+      "loss": 1.3143,
+      "step": 4863
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.652262446454519e-05,
+      "loss": 1.2003,
+      "step": 4864
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.644875574243311e-05,
+      "loss": 1.2862,
+      "step": 4865
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.637491426581125e-05,
+      "loss": 1.5387,
+      "step": 4866
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.630110005096289e-05,
+      "loss": 1.5809,
+      "step": 4867
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.622731311416585e-05,
+      "loss": 1.2867,
+      "step": 4868
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.615355347169133e-05,
+      "loss": 1.5561,
+      "step": 4869
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.60798211398049e-05,
+      "loss": 1.4985,
+      "step": 4870
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.600611613476598e-05,
+      "loss": 1.5261,
+      "step": 4871
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.593243847282808e-05,
+      "loss": 1.3887,
+      "step": 4872
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.585878817023839e-05,
+      "loss": 1.399,
+      "step": 4873
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.578516524323837e-05,
+      "loss": 1.4893,
+      "step": 4874
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.571156970806333e-05,
+      "loss": 1.4717,
+      "step": 4875
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.56380015809425e-05,
+      "loss": 1.4365,
+      "step": 4876
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.55644608780992e-05,
+      "loss": 1.2204,
+      "step": 4877
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.549094761575042e-05,
+      "loss": 1.4682,
+      "step": 4878
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.541746181010752e-05,
+      "loss": 1.3436,
+      "step": 4879
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.534400347737539e-05,
+      "loss": 1.4406,
+      "step": 4880
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.52705726337531e-05,
+      "loss": 1.5899,
+      "step": 4881
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.519716929543361e-05,
+      "loss": 1.4307,
+      "step": 4882
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.51237934786039e-05,
+      "loss": 1.3262,
+      "step": 4883
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.505044519944459e-05,
+      "loss": 1.5212,
+      "step": 4884
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.497712447413057e-05,
+      "loss": 1.4373,
+      "step": 4885
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.490383131883044e-05,
+      "loss": 1.5448,
+      "step": 4886
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.483056574970689e-05,
+      "loss": 1.2963,
+      "step": 4887
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.475732778291626e-05,
+      "loss": 1.3218,
+      "step": 4888
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.468411743460908e-05,
+      "loss": 1.3312,
+      "step": 4889
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.461093472092966e-05,
+      "loss": 1.6142,
+      "step": 4890
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.453777965801623e-05,
+      "loss": 1.4425,
+      "step": 4891
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.446465226200092e-05,
+      "loss": 1.2737,
+      "step": 4892
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.439155254900977e-05,
+      "loss": 1.3883,
+      "step": 4893
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.43184805351628e-05,
+      "loss": 1.6356,
+      "step": 4894
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.424543623657367e-05,
+      "loss": 1.204,
+      "step": 4895
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.417241966935018e-05,
+      "loss": 1.4583,
+      "step": 4896
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.40994308495939e-05,
+      "loss": 1.6071,
+      "step": 4897
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.402646979340037e-05,
+      "loss": 1.5721,
+      "step": 4898
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.395353651685879e-05,
+      "loss": 1.6619,
+      "step": 4899
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.38806310360526e-05,
+      "loss": 1.378,
+      "step": 4900
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.380775336705874e-05,
+      "loss": 1.4804,
+      "step": 4901
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.373490352594823e-05,
+      "loss": 1.3764,
+      "step": 4902
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.366208152878589e-05,
+      "loss": 1.187,
+      "step": 4903
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.358928739163049e-05,
+      "loss": 1.4742,
+      "step": 4904
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.351652113053442e-05,
+      "loss": 1.8533,
+      "step": 4905
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.344378276154418e-05,
+      "loss": 1.1955,
+      "step": 4906
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.337107230069997e-05,
+      "loss": 1.2599,
+      "step": 4907
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.329838976403591e-05,
+      "loss": 1.4044,
+      "step": 4908
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.322573516758003e-05,
+      "loss": 1.4643,
+      "step": 4909
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.315310852735385e-05,
+      "loss": 1.6774,
+      "step": 4910
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.308050985937327e-05,
+      "loss": 1.4649,
+      "step": 4911
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.300793917964755e-05,
+      "loss": 1.4776,
+      "step": 4912
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.293539650417997e-05,
+      "loss": 1.3826,
+      "step": 4913
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.286288184896768e-05,
+      "loss": 1.6168,
+      "step": 4914
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.279039523000164e-05,
+      "loss": 1.4385,
+      "step": 4915
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.271793666326645e-05,
+      "loss": 1.4153,
+      "step": 4916
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.264550616474073e-05,
+      "loss": 1.651,
+      "step": 4917
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.257310375039681e-05,
+      "loss": 1.1879,
+      "step": 4918
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.250072943620095e-05,
+      "loss": 1.4644,
+      "step": 4919
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.242838323811293e-05,
+      "loss": 1.1835,
+      "step": 4920
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.235606517208666e-05,
+      "loss": 1.4936,
+      "step": 4921
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 7.228377525406964e-05,
+      "loss": 1.6768,
+      "step": 4922
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.221151350000332e-05,
+      "loss": 1.4912,
+      "step": 4923
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.213927992582266e-05,
+      "loss": 1.4008,
+      "step": 4924
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.206707454745674e-05,
+      "loss": 1.354,
+      "step": 4925
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.19948973808283e-05,
+      "loss": 1.4804,
+      "step": 4926
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.192274844185372e-05,
+      "loss": 1.3528,
+      "step": 4927
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.185062774644329e-05,
+      "loss": 1.4573,
+      "step": 4928
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.177853531050109e-05,
+      "loss": 1.5004,
+      "step": 4929
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.170647114992496e-05,
+      "loss": 1.2537,
+      "step": 4930
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.163443528060629e-05,
+      "loss": 1.4979,
+      "step": 4931
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.156242771843066e-05,
+      "loss": 1.3858,
+      "step": 4932
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.1490448479277e-05,
+      "loss": 1.3739,
+      "step": 4933
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.141849757901824e-05,
+      "loss": 1.5,
+      "step": 4934
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.13465750335208e-05,
+      "loss": 1.3461,
+      "step": 4935
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.127468085864528e-05,
+      "loss": 1.7089,
+      "step": 4936
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.120281507024558e-05,
+      "loss": 1.3835,
+      "step": 4937
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.113097768416956e-05,
+      "loss": 1.5495,
+      "step": 4938
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.105916871625884e-05,
+      "loss": 1.4551,
+      "step": 4939
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.098738818234871e-05,
+      "loss": 1.3892,
+      "step": 4940
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.091563609826813e-05,
+      "loss": 1.4094,
+      "step": 4941
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.084391247983981e-05,
+      "loss": 1.358,
+      "step": 4942
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.077221734288045e-05,
+      "loss": 1.3633,
+      "step": 4943
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.070055070320001e-05,
+      "loss": 1.4541,
+      "step": 4944
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.062891257660252e-05,
+      "loss": 1.4483,
+      "step": 4945
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.055730297888555e-05,
+      "loss": 1.6267,
+      "step": 4946
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.04857219258405e-05,
+      "loss": 1.3908,
+      "step": 4947
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.041416943325232e-05,
+      "loss": 1.4205,
+      "step": 4948
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.034264551689978e-05,
+      "loss": 1.4743,
+      "step": 4949
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.02711501925553e-05,
+      "loss": 1.6647,
+      "step": 4950
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.019968347598513e-05,
+      "loss": 1.4804,
+      "step": 4951
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.012824538294891e-05,
+      "loss": 1.3246,
+      "step": 4952
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 7.005683592920024e-05,
+      "loss": 1.2661,
+      "step": 4953
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.99854551304863e-05,
+      "loss": 1.2556,
+      "step": 4954
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.991410300254808e-05,
+      "loss": 1.2603,
+      "step": 4955
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.984277956111988e-05,
+      "loss": 1.5709,
+      "step": 4956
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.977148482193021e-05,
+      "loss": 1.2987,
+      "step": 4957
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.970021880070078e-05,
+      "loss": 1.4515,
+      "step": 4958
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.962898151314721e-05,
+      "loss": 1.4883,
+      "step": 4959
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.955777297497874e-05,
+      "loss": 1.3107,
+      "step": 4960
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.948659320189825e-05,
+      "loss": 1.4521,
+      "step": 4961
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.941544220960235e-05,
+      "loss": 1.4475,
+      "step": 4962
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.93443200137811e-05,
+      "loss": 1.441,
+      "step": 4963
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.927322663011847e-05,
+      "loss": 1.3915,
+      "step": 4964
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.920216207429186e-05,
+      "loss": 1.458,
+      "step": 4965
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.913112636197256e-05,
+      "loss": 1.4132,
+      "step": 4966
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.906011950882511e-05,
+      "loss": 1.5305,
+      "step": 4967
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.898914153050819e-05,
+      "loss": 1.4423,
+      "step": 4968
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.891819244267367e-05,
+      "loss": 1.1961,
+      "step": 4969
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.884727226096727e-05,
+      "loss": 1.3611,
+      "step": 4970
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.87763810010283e-05,
+      "loss": 1.3377,
+      "step": 4971
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.870551867848974e-05,
+      "loss": 1.5908,
+      "step": 4972
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.8634685308978e-05,
+      "loss": 1.3912,
+      "step": 4973
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.856388090811331e-05,
+      "loss": 1.2894,
+      "step": 4974
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.849310549150944e-05,
+      "loss": 1.465,
+      "step": 4975
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.842235907477377e-05,
+      "loss": 1.24,
+      "step": 4976
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.83516416735073e-05,
+      "loss": 1.3449,
+      "step": 4977
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.828095330330455e-05,
+      "loss": 1.584,
+      "step": 4978
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.821029397975383e-05,
+      "loss": 1.3499,
+      "step": 4979
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.813966371843676e-05,
+      "loss": 1.4349,
+      "step": 4980
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.80690625349288e-05,
+      "loss": 1.4491,
+      "step": 4981
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.799849044479886e-05,
+      "loss": 1.4666,
+      "step": 4982
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.792794746360958e-05,
+      "loss": 1.4351,
+      "step": 4983
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.785743360691694e-05,
+      "loss": 1.4535,
+      "step": 4984
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.778694889027071e-05,
+      "loss": 1.34,
+      "step": 4985
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.771649332921415e-05,
+      "loss": 1.5698,
+      "step": 4986
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.764606693928418e-05,
+      "loss": 1.336,
+      "step": 4987
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.7575669736011e-05,
+      "loss": 1.2909,
+      "step": 4988
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.750530173491887e-05,
+      "loss": 1.1949,
+      "step": 4989
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.74349629515251e-05,
+      "loss": 1.4885,
+      "step": 4990
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.736465340134086e-05,
+      "loss": 1.2903,
+      "step": 4991
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.72943730998708e-05,
+      "loss": 1.447,
+      "step": 4992
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.72241220626131e-05,
+      "loss": 1.2741,
+      "step": 4993
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.715390030505959e-05,
+      "loss": 1.4104,
+      "step": 4994
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.708370784269542e-05,
+      "loss": 1.3718,
+      "step": 4995
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.701354469099945e-05,
+      "loss": 1.3087,
+      "step": 4996
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.694341086544407e-05,
+      "loss": 1.4037,
+      "step": 4997
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.687330638149523e-05,
+      "loss": 1.503,
+      "step": 4998
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.680323125461214e-05,
+      "loss": 1.5119,
+      "step": 4999
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.673318550024806e-05,
+      "loss": 1.0325,
+      "step": 5000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.666316913384923e-05,
+      "loss": 1.3127,
+      "step": 5001
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.65931821708557e-05,
+      "loss": 1.0905,
+      "step": 5002
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.6523224626701e-05,
+      "loss": 1.0697,
+      "step": 5003
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.64532965168122e-05,
+      "loss": 1.4711,
+      "step": 5004
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.638339785660972e-05,
+      "loss": 1.4947,
+      "step": 5005
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.631352866150764e-05,
+      "loss": 1.2798,
+      "step": 5006
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.624368894691351e-05,
+      "loss": 1.4244,
+      "step": 5007
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.617387872822842e-05,
+      "loss": 1.2675,
+      "step": 5008
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.610409802084679e-05,
+      "loss": 1.4157,
+      "step": 5009
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.603434684015661e-05,
+      "loss": 1.3126,
+      "step": 5010
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.596462520153964e-05,
+      "loss": 1.3973,
+      "step": 5011
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.589493312037063e-05,
+      "loss": 1.4431,
+      "step": 5012
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.582527061201817e-05,
+      "loss": 1.4702,
+      "step": 5013
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.575563769184419e-05,
+      "loss": 1.3967,
+      "step": 5014
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.568603437520422e-05,
+      "loss": 1.4169,
+      "step": 5015
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.5616460677447e-05,
+      "loss": 1.5053,
+      "step": 5016
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.554691661391499e-05,
+      "loss": 1.2544,
+      "step": 5017
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.5477402199944e-05,
+      "loss": 1.5576,
+      "step": 5018
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.540791745086345e-05,
+      "loss": 1.1581,
+      "step": 5019
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.533846238199588e-05,
+      "loss": 1.4789,
+      "step": 5020
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.526903700865776e-05,
+      "loss": 1.496,
+      "step": 5021
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.519964134615856e-05,
+      "loss": 1.3958,
+      "step": 5022
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.513027540980154e-05,
+      "loss": 1.398,
+      "step": 5023
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.506093921488304e-05,
+      "loss": 1.2809,
+      "step": 5024
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.499163277669334e-05,
+      "loss": 1.3464,
+      "step": 5025
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.492235611051567e-05,
+      "loss": 1.6676,
+      "step": 5026
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.485310923162697e-05,
+      "loss": 1.4681,
+      "step": 5027
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.478389215529756e-05,
+      "loss": 1.5825,
+      "step": 5028
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.471470489679117e-05,
+      "loss": 1.5408,
+      "step": 5029
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.4645547471365e-05,
+      "loss": 1.537,
+      "step": 5030
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.457641989426946e-05,
+      "loss": 1.4603,
+      "step": 5031
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.450732218074879e-05,
+      "loss": 1.4904,
+      "step": 5032
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.443825434604021e-05,
+      "loss": 1.2668,
+      "step": 5033
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.436921640537461e-05,
+      "loss": 1.3426,
+      "step": 5034
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.430020837397621e-05,
+      "loss": 1.501,
+      "step": 5035
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.423123026706273e-05,
+      "loss": 1.3895,
+      "step": 5036
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.416228209984507e-05,
+      "loss": 1.6046,
+      "step": 5037
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.409336388752772e-05,
+      "loss": 1.4894,
+      "step": 5038
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.40244756453085e-05,
+      "loss": 1.501,
+      "step": 5039
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.395561738837867e-05,
+      "loss": 1.418,
+      "step": 5040
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.388678913192276e-05,
+      "loss": 1.3173,
+      "step": 5041
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.381799089111882e-05,
+      "loss": 1.1386,
+      "step": 5042
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.37492226811382e-05,
+      "loss": 1.4629,
+      "step": 5043
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.368048451714563e-05,
+      "loss": 1.2279,
+      "step": 5044
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.361177641429928e-05,
+      "loss": 1.3965,
+      "step": 5045
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.354309838775061e-05,
+      "loss": 1.4288,
+      "step": 5046
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.347445045264455e-05,
+      "loss": 1.3723,
+      "step": 5047
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.340583262411921e-05,
+      "loss": 1.4848,
+      "step": 5048
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.333724491730621e-05,
+      "loss": 1.2222,
+      "step": 5049
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.326868734733056e-05,
+      "loss": 1.7153,
+      "step": 5050
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.320015992931055e-05,
+      "loss": 1.3073,
+      "step": 5051
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.313166267835768e-05,
+      "loss": 1.3779,
+      "step": 5052
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.306319560957719e-05,
+      "loss": 1.4207,
+      "step": 5053
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.299475873806722e-05,
+      "loss": 1.3065,
+      "step": 5054
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.29263520789196e-05,
+      "loss": 1.285,
+      "step": 5055
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.285797564721915e-05,
+      "loss": 1.267,
+      "step": 5056
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.278962945804447e-05,
+      "loss": 1.5704,
+      "step": 5057
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.272131352646706e-05,
+      "loss": 1.4607,
+      "step": 5058
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.2653027867552e-05,
+      "loss": 1.4169,
+      "step": 5059
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.258477249635764e-05,
+      "loss": 1.6787,
+      "step": 5060
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.251654742793562e-05,
+      "loss": 1.5715,
+      "step": 5061
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.244835267733095e-05,
+      "loss": 1.6077,
+      "step": 5062
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.23801882595818e-05,
+      "loss": 1.6022,
+      "step": 5063
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.231205418971997e-05,
+      "loss": 1.3677,
+      "step": 5064
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.224395048277019e-05,
+      "loss": 1.3963,
+      "step": 5065
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.217587715375072e-05,
+      "loss": 1.2977,
+      "step": 5066
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.210783421767312e-05,
+      "loss": 1.3674,
+      "step": 5067
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.203982168954223e-05,
+      "loss": 1.3697,
+      "step": 5068
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.197183958435604e-05,
+      "loss": 1.741,
+      "step": 5069
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.190388791710602e-05,
+      "loss": 1.469,
+      "step": 5070
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.183596670277682e-05,
+      "loss": 1.4727,
+      "step": 5071
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.17680759563465e-05,
+      "loss": 1.3404,
+      "step": 5072
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.170021569278619e-05,
+      "loss": 1.4368,
+      "step": 5073
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.163238592706046e-05,
+      "loss": 1.3877,
+      "step": 5074
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.156458667412715e-05,
+      "loss": 1.5912,
+      "step": 5075
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.149681794893734e-05,
+      "loss": 1.3066,
+      "step": 5076
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.142907976643524e-05,
+      "loss": 1.4528,
+      "step": 5077
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.136137214155861e-05,
+      "loss": 1.4854,
+      "step": 5078
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.129369508923835e-05,
+      "loss": 1.2665,
+      "step": 5079
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.122604862439843e-05,
+      "loss": 1.4005,
+      "step": 5080
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.11584327619563e-05,
+      "loss": 1.4135,
+      "step": 5081
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.109084751682259e-05,
+      "loss": 1.1091,
+      "step": 5082
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.1023292903901296e-05,
+      "loss": 1.5585,
+      "step": 5083
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.0955768938089344e-05,
+      "loss": 1.5017,
+      "step": 5084
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.088827563427721e-05,
+      "loss": 1.51,
+      "step": 5085
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.08208130073485e-05,
+      "loss": 1.2932,
+      "step": 5086
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.075338107218012e-05,
+      "loss": 1.4158,
+      "step": 5087
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.0685979843641926e-05,
+      "loss": 1.2396,
+      "step": 5088
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.061860933659749e-05,
+      "loss": 1.4128,
+      "step": 5089
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.055126956590316e-05,
+      "loss": 1.5426,
+      "step": 5090
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.0483960546408746e-05,
+      "loss": 1.4249,
+      "step": 5091
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.0416682292957205e-05,
+      "loss": 1.2967,
+      "step": 5092
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.034943482038478e-05,
+      "loss": 1.491,
+      "step": 5093
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.0282218143520754e-05,
+      "loss": 1.4177,
+      "step": 5094
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.0215032277187744e-05,
+      "loss": 1.5244,
+      "step": 5095
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.014787723620172e-05,
+      "loss": 1.2266,
+      "step": 5096
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.008075303537153e-05,
+      "loss": 1.062,
+      "step": 5097
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.00136596894995e-05,
+      "loss": 1.5372,
+      "step": 5098
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.994659721338087e-05,
+      "loss": 1.3232,
+      "step": 5099
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.987956562180445e-05,
+      "loss": 1.3923,
+      "step": 5100
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.98125649295519e-05,
+      "loss": 1.4182,
+      "step": 5101
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.9745595151398194e-05,
+      "loss": 1.5496,
+      "step": 5102
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.9678656302111534e-05,
+      "loss": 1.4849,
+      "step": 5103
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.961174839645331e-05,
+      "loss": 1.2247,
+      "step": 5104
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.9544871449177906e-05,
+      "loss": 1.1836,
+      "step": 5105
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.9478025475033096e-05,
+      "loss": 1.4477,
+      "step": 5106
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.941121048875971e-05,
+      "loss": 1.2778,
+      "step": 5107
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.934442650509184e-05,
+      "loss": 1.4727,
+      "step": 5108
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.927767353875651e-05,
+      "loss": 1.4074,
+      "step": 5109
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.9210951604474294e-05,
+      "loss": 1.3928,
+      "step": 5110
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.914426071695851e-05,
+      "loss": 1.518,
+      "step": 5111
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.907760089091589e-05,
+      "loss": 1.3832,
+      "step": 5112
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.901097214104623e-05,
+      "loss": 1.4399,
+      "step": 5113
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.8944374482042486e-05,
+      "loss": 1.4885,
+      "step": 5114
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.8877807928590836e-05,
+      "loss": 1.2353,
+      "step": 5115
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.8811272495370394e-05,
+      "loss": 1.3999,
+      "step": 5116
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.874476819705357e-05,
+      "loss": 1.4157,
+      "step": 5117
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.8678295048305906e-05,
+      "loss": 1.4303,
+      "step": 5118
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.861185306378611e-05,
+      "loss": 1.5316,
+      "step": 5119
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.854544225814573e-05,
+      "loss": 1.4214,
+      "step": 5120
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.847906264602996e-05,
+      "loss": 1.3981,
+      "step": 5121
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.841271424207657e-05,
+      "loss": 1.5083,
+      "step": 5122
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.83463970609168e-05,
+      "loss": 1.3749,
+      "step": 5123
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.828011111717491e-05,
+      "loss": 1.2686,
+      "step": 5124
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.821385642546826e-05,
+      "loss": 1.4212,
+      "step": 5125
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.814763300040726e-05,
+      "loss": 1.1717,
+      "step": 5126
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.80814408565955e-05,
+      "loss": 1.0821,
+      "step": 5127
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.801528000862968e-05,
+      "loss": 1.5381,
+      "step": 5128
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.7949150471099567e-05,
+      "loss": 1.4526,
+      "step": 5129
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.78830522585881e-05,
+      "loss": 1.314,
+      "step": 5130
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.781698538567104e-05,
+      "loss": 1.4032,
+      "step": 5131
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.775094986691769e-05,
+      "loss": 1.4772,
+      "step": 5132
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.7684945716890005e-05,
+      "loss": 1.4516,
+      "step": 5133
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.7618972950143266e-05,
+      "loss": 1.3124,
+      "step": 5134
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.755303158122578e-05,
+      "loss": 1.043,
+      "step": 5135
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.748712162467895e-05,
+      "loss": 1.392,
+      "step": 5136
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.742124309503716e-05,
+      "loss": 1.3918,
+      "step": 5137
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.735539600682791e-05,
+      "loss": 1.3218,
+      "step": 5138
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.728958037457186e-05,
+      "loss": 1.5036,
+      "step": 5139
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.722379621278267e-05,
+      "loss": 1.1828,
+      "step": 5140
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.7158043535966875e-05,
+      "loss": 1.2537,
+      "step": 5141
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.7092322358624496e-05,
+      "loss": 1.4033,
+      "step": 5142
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.702663269524815e-05,
+      "loss": 1.4688,
+      "step": 5143
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.696097456032383e-05,
+      "loss": 1.2979,
+      "step": 5144
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.6895347968330294e-05,
+      "loss": 1.5385,
+      "step": 5145
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.682975293373964e-05,
+      "loss": 1.3649,
+      "step": 5146
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.6764189471016913e-05,
+      "loss": 1.5109,
+      "step": 5147
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.669865759462001e-05,
+      "loss": 1.2095,
+      "step": 5148
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.6633157319000076e-05,
+      "loss": 1.5682,
+      "step": 5149
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.656768865860118e-05,
+      "loss": 1.5981,
+      "step": 5150
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.650225162786056e-05,
+      "loss": 1.5968,
+      "step": 5151
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.643684624120818e-05,
+      "loss": 1.3672,
+      "step": 5152
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.637147251306745e-05,
+      "loss": 1.3979,
+      "step": 5153
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.63061304578544e-05,
+      "loss": 1.563,
+      "step": 5154
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.6240820089978284e-05,
+      "loss": 1.4884,
+      "step": 5155
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.617554142384134e-05,
+      "loss": 1.4243,
+      "step": 5156
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.6110294473838885e-05,
+      "loss": 1.4475,
+      "step": 5157
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.6045079254359e-05,
+      "loss": 1.584,
+      "step": 5158
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.597989577978302e-05,
+      "loss": 1.3581,
+      "step": 5159
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.591474406448518e-05,
+      "loss": 1.3454,
+      "step": 5160
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.584962412283279e-05,
+      "loss": 1.4256,
+      "step": 5161
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.578453596918596e-05,
+      "loss": 1.3238,
+      "step": 5162
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.5719479617897874e-05,
+      "loss": 1.4224,
+      "step": 5163
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.565445508331497e-05,
+      "loss": 1.2938,
+      "step": 5164
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.558946237977627e-05,
+      "loss": 1.4556,
+      "step": 5165
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.552450152161397e-05,
+      "loss": 1.4519,
+      "step": 5166
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.5459572523153234e-05,
+      "loss": 1.4293,
+      "step": 5167
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.539467539871226e-05,
+      "loss": 1.5144,
+      "step": 5168
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.5329810162602015e-05,
+      "loss": 1.4444,
+      "step": 5169
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.526497682912664e-05,
+      "loss": 1.2167,
+      "step": 5170
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.5200175412583154e-05,
+      "loss": 1.4196,
+      "step": 5171
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.5135405927261605e-05,
+      "loss": 1.2008,
+      "step": 5172
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.507066838744484e-05,
+      "loss": 1.7165,
+      "step": 5173
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.500596280740882e-05,
+      "loss": 1.2497,
+      "step": 5174
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.494128920142239e-05,
+      "loss": 1.422,
+      "step": 5175
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.487664758374742e-05,
+      "loss": 1.4528,
+      "step": 5176
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.48120379686385e-05,
+      "loss": 1.4672,
+      "step": 5177
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.4747460370343575e-05,
+      "loss": 1.5102,
+      "step": 5178
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.468291480310306e-05,
+      "loss": 1.5058,
+      "step": 5179
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.461840128115061e-05,
+      "loss": 1.4761,
+      "step": 5180
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.4553919818712765e-05,
+      "loss": 1.3422,
+      "step": 5181
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.448947043000889e-05,
+      "loss": 1.4114,
+      "step": 5182
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.442505312925148e-05,
+      "loss": 1.5931,
+      "step": 5183
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.43606679306456e-05,
+      "loss": 1.4786,
+      "step": 5184
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.4296314848389707e-05,
+      "loss": 1.5127,
+      "step": 5185
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.423199389667475e-05,
+      "loss": 1.442,
+      "step": 5186
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.416770508968487e-05,
+      "loss": 1.4093,
+      "step": 5187
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.410344844159687e-05,
+      "loss": 1.429,
+      "step": 5188
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.403922396658085e-05,
+      "loss": 1.308,
+      "step": 5189
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.397503167879938e-05,
+      "loss": 1.4917,
+      "step": 5190
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.3910871592408175e-05,
+      "loss": 1.4579,
+      "step": 5191
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.384674372155582e-05,
+      "loss": 1.4531,
+      "step": 5192
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.378264808038385e-05,
+      "loss": 1.5151,
+      "step": 5193
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 5.3718584683026464e-05,
+      "loss": 1.486,
+      "step": 5194
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.3654553543611e-05,
+      "loss": 1.5091,
+      "step": 5195
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.3590554676257575e-05,
+      "loss": 1.5584,
+      "step": 5196
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.35265880950792e-05,
+      "loss": 1.2393,
+      "step": 5197
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.346265381418179e-05,
+      "loss": 1.2966,
+      "step": 5198
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.3398751847664075e-05,
+      "loss": 1.5701,
+      "step": 5199
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.33348822096178e-05,
+      "loss": 1.5817,
+      "step": 5200
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.3271044914127354e-05,
+      "loss": 1.4643,
+      "step": 5201
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.320723997527017e-05,
+      "loss": 1.2481,
+      "step": 5202
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.3143467407116506e-05,
+      "loss": 1.6713,
+      "step": 5203
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.3079727223729536e-05,
+      "loss": 1.447,
+      "step": 5204
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.301601943916512e-05,
+      "loss": 1.3193,
+      "step": 5205
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2952344067472115e-05,
+      "loss": 1.5178,
+      "step": 5206
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.28887011226922e-05,
+      "loss": 1.5444,
+      "step": 5207
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.282509061885999e-05,
+      "loss": 1.3043,
+      "step": 5208
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2761512570002656e-05,
+      "loss": 1.2788,
+      "step": 5209
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.269796699014067e-05,
+      "loss": 1.4896,
+      "step": 5210
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.263445389328687e-05,
+      "loss": 1.3868,
+      "step": 5211
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.257097329344727e-05,
+      "loss": 1.2431,
+      "step": 5212
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.250752520462055e-05,
+      "loss": 1.4965,
+      "step": 5213
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2444109640798266e-05,
+      "loss": 1.444,
+      "step": 5214
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.238072661596487e-05,
+      "loss": 1.474,
+      "step": 5215
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.231737614409742e-05,
+      "loss": 1.3439,
+      "step": 5216
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.225405823916616e-05,
+      "loss": 1.7172,
+      "step": 5217
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2190772915133744e-05,
+      "loss": 1.5796,
+      "step": 5218
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2127520185955994e-05,
+      "loss": 1.6014,
+      "step": 5219
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.206430006558116e-05,
+      "loss": 1.3602,
+      "step": 5220
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.2001112567950816e-05,
+      "loss": 1.4751,
+      "step": 5221
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.193795770699885e-05,
+      "loss": 1.2675,
+      "step": 5222
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.187483549665222e-05,
+      "loss": 1.4587,
+      "step": 5223
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.181174595083063e-05,
+      "loss": 1.2608,
+      "step": 5224
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.174868908344659e-05,
+      "loss": 1.4702,
+      "step": 5225
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.1685664908405316e-05,
+      "loss": 1.3673,
+      "step": 5226
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.162267343960492e-05,
+      "loss": 1.5586,
+      "step": 5227
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.155971469093625e-05,
+      "loss": 1.4984,
+      "step": 5228
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.149678867628305e-05,
+      "loss": 1.2751,
+      "step": 5229
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.1433895409521547e-05,
+      "loss": 1.4079,
+      "step": 5230
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.1371034904521134e-05,
+      "loss": 1.548,
+      "step": 5231
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.130820717514378e-05,
+      "loss": 1.4239,
+      "step": 5232
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.124541223524415e-05,
+      "loss": 1.3678,
+      "step": 5233
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.11826500986698e-05,
+      "loss": 1.3013,
+      "step": 5234
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.111992077926104e-05,
+      "loss": 1.3259,
+      "step": 5235
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.105722429085098e-05,
+      "loss": 1.3637,
+      "step": 5236
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.099456064726531e-05,
+      "loss": 1.4211,
+      "step": 5237
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.093192986232265e-05,
+      "loss": 1.4019,
+      "step": 5238
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.0869331949834343e-05,
+      "loss": 1.4286,
+      "step": 5239
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.080676692360453e-05,
+      "loss": 1.4123,
+      "step": 5240
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.0744234797429845e-05,
+      "loss": 1.5135,
+      "step": 5241
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.06817355851001e-05,
+      "loss": 1.3375,
+      "step": 5242
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.061926930039742e-05,
+      "loss": 1.4716,
+      "step": 5243
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.05568359570969e-05,
+      "loss": 1.0529,
+      "step": 5244
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.049443556896638e-05,
+      "loss": 1.3717,
+      "step": 5245
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.0432068149766376e-05,
+      "loss": 1.4276,
+      "step": 5246
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.0369733713250046e-05,
+      "loss": 1.3372,
+      "step": 5247
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.0307432273163436e-05,
+      "loss": 1.2934,
+      "step": 5248
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.024516384324522e-05,
+      "loss": 1.2798,
+      "step": 5249
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.018292843722683e-05,
+      "loss": 1.2741,
+      "step": 5250
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.0120726068832446e-05,
+      "loss": 1.5044,
+      "step": 5251
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.0058556751778775e-05,
+      "loss": 1.5647,
+      "step": 5252
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.999642049977558e-05,
+      "loss": 1.4648,
+      "step": 5253
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.993431732652498e-05,
+      "loss": 1.3594,
+      "step": 5254
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.987224724572199e-05,
+      "loss": 1.632,
+      "step": 5255
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.9810210271054304e-05,
+      "loss": 1.2482,
+      "step": 5256
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.974820641620237e-05,
+      "loss": 1.3066,
+      "step": 5257
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.96862356948391e-05,
+      "loss": 1.2467,
+      "step": 5258
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.9624298120630366e-05,
+      "loss": 1.3865,
+      "step": 5259
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.956239370723461e-05,
+      "loss": 1.4421,
+      "step": 5260
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.9500522468303054e-05,
+      "loss": 1.5849,
+      "step": 5261
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.9438684417479386e-05,
+      "loss": 1.405,
+      "step": 5262
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.937687956840018e-05,
+      "loss": 1.4171,
+      "step": 5263
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.931510793469463e-05,
+      "loss": 1.2795,
+      "step": 5264
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.9253369529984605e-05,
+      "loss": 1.1118,
+      "step": 5265
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.919166436788465e-05,
+      "loss": 1.5906,
+      "step": 5266
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.912999246200196e-05,
+      "loss": 1.4629,
+      "step": 5267
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.9068353825936485e-05,
+      "loss": 1.4013,
+      "step": 5268
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.900674847328062e-05,
+      "loss": 1.3301,
+      "step": 5269
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.894517641761962e-05,
+      "loss": 1.2827,
+      "step": 5270
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.888363767253137e-05,
+      "loss": 1.5374,
+      "step": 5271
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.8822132251586406e-05,
+      "loss": 1.5383,
+      "step": 5272
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.876066016834773e-05,
+      "loss": 1.5564,
+      "step": 5273
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.869922143637138e-05,
+      "loss": 1.4251,
+      "step": 5274
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.8637816069205656e-05,
+      "loss": 1.1734,
+      "step": 5275
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.857644408039166e-05,
+      "loss": 1.4651,
+      "step": 5276
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.8515105483463183e-05,
+      "loss": 1.3311,
+      "step": 5277
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.8453800291946616e-05,
+      "loss": 1.3255,
+      "step": 5278
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.839252851936089e-05,
+      "loss": 1.3311,
+      "step": 5279
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.8331290179217645e-05,
+      "loss": 1.3861,
+      "step": 5280
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.8270085285021196e-05,
+      "loss": 1.6508,
+      "step": 5281
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.82089138502684e-05,
+      "loss": 1.2333,
+      "step": 5282
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.814777588844885e-05,
+      "loss": 1.4091,
+      "step": 5283
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.808667141304446e-05,
+      "loss": 1.407,
+      "step": 5284
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.802560043753026e-05,
+      "loss": 1.1673,
+      "step": 5285
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.7964562975373375e-05,
+      "loss": 1.4376,
+      "step": 5286
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.7903559040033875e-05,
+      "loss": 1.0462,
+      "step": 5287
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.7842588644964316e-05,
+      "loss": 1.3786,
+      "step": 5288
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.778165180360994e-05,
+      "loss": 1.4362,
+      "step": 5289
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.7720748529408375e-05,
+      "loss": 1.3287,
+      "step": 5290
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.7659878835790103e-05,
+      "loss": 1.2799,
+      "step": 5291
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.759904273617806e-05,
+      "loss": 1.3648,
+      "step": 5292
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.7538240243987875e-05,
+      "loss": 1.3623,
+      "step": 5293
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.747747137262759e-05,
+      "loss": 1.4981,
+      "step": 5294
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.7416736135498e-05,
+      "loss": 1.3311,
+      "step": 5295
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.735603454599242e-05,
+      "loss": 1.4946,
+      "step": 5296
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.7295366617496785e-05,
+      "loss": 1.1491,
+      "step": 5297
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.7234732363389445e-05,
+      "loss": 1.3677,
+      "step": 5298
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.7174131797041575e-05,
+      "loss": 1.3657,
+      "step": 5299
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.711356493181684e-05,
+      "loss": 1.2068,
+      "step": 5300
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.7053031781071276e-05,
+      "loss": 1.453,
+      "step": 5301
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.699253235815373e-05,
+      "loss": 1.5207,
+      "step": 5302
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.69320666764055e-05,
+      "loss": 1.3068,
+      "step": 5303
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.687163474916054e-05,
+      "loss": 1.4984,
+      "step": 5304
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.681123658974509e-05,
+      "loss": 1.4585,
+      "step": 5305
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.675087221147838e-05,
+      "loss": 1.4134,
+      "step": 5306
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.669054162767177e-05,
+      "loss": 1.459,
+      "step": 5307
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.6630244851629456e-05,
+      "loss": 1.1079,
+      "step": 5308
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.656998189664789e-05,
+      "loss": 1.3176,
+      "step": 5309
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.65097527760165e-05,
+      "loss": 1.3285,
+      "step": 5310
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.64495575030168e-05,
+      "loss": 1.6574,
+      "step": 5311
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.638939609092312e-05,
+      "loss": 1.5516,
+      "step": 5312
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.6329268553002213e-05,
+      "loss": 1.4291,
+      "step": 5313
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.626917490251346e-05,
+      "loss": 1.228,
+      "step": 5314
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.620911515270858e-05,
+      "loss": 1.3658,
+      "step": 5315
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.614908931683193e-05,
+      "loss": 1.6437,
+      "step": 5316
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.6089097408120576e-05,
+      "loss": 1.4756,
+      "step": 5317
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.6029139439803734e-05,
+      "loss": 1.3887,
+      "step": 5318
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.5969215425103395e-05,
+      "loss": 1.4509,
+      "step": 5319
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.5909325377233956e-05,
+      "loss": 1.5504,
+      "step": 5320
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.584946930940246e-05,
+      "loss": 1.4033,
+      "step": 5321
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.578964723480821e-05,
+      "loss": 1.4849,
+      "step": 5322
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.572985916664321e-05,
+      "loss": 1.4039,
+      "step": 5323
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.567010511809191e-05,
+      "loss": 1.3723,
+      "step": 5324
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.561038510233131e-05,
+      "loss": 1.359,
+      "step": 5325
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.555069913253076e-05,
+      "loss": 1.3657,
+      "step": 5326
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.5491047221852244e-05,
+      "loss": 1.5494,
+      "step": 5327
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.5431429383450195e-05,
+      "loss": 1.441,
+      "step": 5328
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.537184563047157e-05,
+      "loss": 1.5706,
+      "step": 5329
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.531229597605558e-05,
+      "loss": 1.3239,
+      "step": 5330
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.525278043333434e-05,
+      "loss": 1.1059,
+      "step": 5331
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.5193299015432036e-05,
+      "loss": 1.3917,
+      "step": 5332
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.513385173546556e-05,
+      "loss": 1.3176,
+      "step": 5333
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.507443860654419e-05,
+      "loss": 1.4296,
+      "step": 5334
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.501505964176971e-05,
+      "loss": 1.4248,
+      "step": 5335
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.495571485423642e-05,
+      "loss": 1.4099,
+      "step": 5336
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.489640425703088e-05,
+      "loss": 1.4315,
+      "step": 5337
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.483712786323231e-05,
+      "loss": 1.4053,
+      "step": 5338
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.477788568591235e-05,
+      "loss": 1.5405,
+      "step": 5339
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.47186777381351e-05,
+      "loss": 1.4313,
+      "step": 5340
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.465950403295693e-05,
+      "loss": 1.6608,
+      "step": 5341
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.4600364583427026e-05,
+      "loss": 1.4481,
+      "step": 5342
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.454125940258664e-05,
+      "loss": 1.32,
+      "step": 5343
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.44821885034697e-05,
+      "loss": 1.587,
+      "step": 5344
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.4423151899102486e-05,
+      "loss": 1.2123,
+      "step": 5345
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.436414960250379e-05,
+      "loss": 1.359,
+      "step": 5346
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.430518162668469e-05,
+      "loss": 1.4412,
+      "step": 5347
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.424624798464885e-05,
+      "loss": 1.2453,
+      "step": 5348
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.418734868939227e-05,
+      "loss": 1.3339,
+      "step": 5349
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.412848375390344e-05,
+      "loss": 1.4867,
+      "step": 5350
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.406965319116323e-05,
+      "loss": 1.5014,
+      "step": 5351
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.401085701414493e-05,
+      "loss": 1.3315,
+      "step": 5352
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.395209523581434e-05,
+      "loss": 1.3099,
+      "step": 5353
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.3893367869129454e-05,
+      "loss": 1.4843,
+      "step": 5354
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.383467492704087e-05,
+      "loss": 1.3047,
+      "step": 5355
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.3776016422491583e-05,
+      "loss": 1.4388,
+      "step": 5356
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.371739236841697e-05,
+      "loss": 1.5858,
+      "step": 5357
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.36588027777447e-05,
+      "loss": 1.3207,
+      "step": 5358
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.360024766339497e-05,
+      "loss": 1.5107,
+      "step": 5359
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.3541727038280364e-05,
+      "loss": 1.6765,
+      "step": 5360
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.348324091530587e-05,
+      "loss": 1.3783,
+      "step": 5361
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.342478930736869e-05,
+      "loss": 1.3268,
+      "step": 5362
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.3366372227358756e-05,
+      "loss": 1.4138,
+      "step": 5363
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.3307989688158036e-05,
+      "loss": 1.595,
+      "step": 5364
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.3249641702641076e-05,
+      "loss": 1.4219,
+      "step": 5365
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.319132828367478e-05,
+      "loss": 1.6525,
+      "step": 5366
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.31330494441184e-05,
+      "loss": 1.525,
+      "step": 5367
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.307480519682363e-05,
+      "loss": 1.3423,
+      "step": 5368
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.301659555463435e-05,
+      "loss": 1.244,
+      "step": 5369
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.2958420530387014e-05,
+      "loss": 1.3093,
+      "step": 5370
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.2900280136910345e-05,
+      "loss": 1.5122,
+      "step": 5371
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.284217438702551e-05,
+      "loss": 1.2069,
+      "step": 5372
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.2784103293545804e-05,
+      "loss": 1.4458,
+      "step": 5373
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.272606686927729e-05,
+      "loss": 1.418,
+      "step": 5374
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.2668065127017974e-05,
+      "loss": 1.2929,
+      "step": 5375
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.2610098079558425e-05,
+      "loss": 1.2742,
+      "step": 5376
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.255216573968153e-05,
+      "loss": 1.5063,
+      "step": 5377
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.249426812016257e-05,
+      "loss": 1.3464,
+      "step": 5378
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.243640523376902e-05,
+      "loss": 1.6538,
+      "step": 5379
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.2378577093260826e-05,
+      "loss": 1.412,
+      "step": 5380
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.232078371139023e-05,
+      "loss": 1.2987,
+      "step": 5381
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.22630251009019e-05,
+      "loss": 1.2524,
+      "step": 5382
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.220530127453259e-05,
+      "loss": 1.3716,
+      "step": 5383
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.21476122450116e-05,
+      "loss": 1.4637,
+      "step": 5384
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.208995802506062e-05,
+      "loss": 1.3666,
+      "step": 5385
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.203233862739342e-05,
+      "loss": 1.3502,
+      "step": 5386
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.197475406471623e-05,
+      "loss": 1.3926,
+      "step": 5387
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.1917204349727615e-05,
+      "loss": 1.2254,
+      "step": 5388
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.185968949511847e-05,
+      "loss": 1.6447,
+      "step": 5389
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.180220951357183e-05,
+      "loss": 1.3903,
+      "step": 5390
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.174476441776325e-05,
+      "loss": 1.2818,
+      "step": 5391
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.168735422036048e-05,
+      "loss": 1.2817,
+      "step": 5392
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.162997893402367e-05,
+      "loss": 1.1082,
+      "step": 5393
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.157263857140505e-05,
+      "loss": 1.3688,
+      "step": 5394
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.1515333145149506e-05,
+      "loss": 1.3193,
+      "step": 5395
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.145806266789387e-05,
+      "loss": 1.5548,
+      "step": 5396
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.140082715226752e-05,
+      "loss": 1.4573,
+      "step": 5397
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.1343626610891816e-05,
+      "loss": 1.7014,
+      "step": 5398
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.1286461056380876e-05,
+      "loss": 1.3426,
+      "step": 5399
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.122933050134066e-05,
+      "loss": 1.4993,
+      "step": 5400
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.117223495836961e-05,
+      "loss": 1.3187,
+      "step": 5401
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.1115174440058456e-05,
+      "loss": 1.3281,
+      "step": 5402
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.105814895899014e-05,
+      "loss": 1.6185,
+      "step": 5403
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.100115852773998e-05,
+      "loss": 1.2658,
+      "step": 5404
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.094420315887533e-05,
+      "loss": 1.2801,
+      "step": 5405
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.088728286495618e-05,
+      "loss": 1.2733,
+      "step": 5406
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.0830397658534425e-05,
+      "loss": 1.4699,
+      "step": 5407
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.077354755215441e-05,
+      "loss": 1.4133,
+      "step": 5408
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.071673255835275e-05,
+      "loss": 1.2106,
+      "step": 5409
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.06599526896583e-05,
+      "loss": 1.4726,
+      "step": 5410
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.0603207958592e-05,
+      "loss": 1.3932,
+      "step": 5411
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.054649837766731e-05,
+      "loss": 1.1435,
+      "step": 5412
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.0489823959389783e-05,
+      "loss": 1.2825,
+      "step": 5413
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.0433184716257275e-05,
+      "loss": 1.4939,
+      "step": 5414
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.037658066075978e-05,
+      "loss": 1.4354,
+      "step": 5415
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.032001180537965e-05,
+      "loss": 1.3777,
+      "step": 5416
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.026347816259146e-05,
+      "loss": 1.4166,
+      "step": 5417
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.020697974486196e-05,
+      "loss": 1.403,
+      "step": 5418
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.0150516564650186e-05,
+      "loss": 1.5645,
+      "step": 5419
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.00940886344074e-05,
+      "loss": 1.4992,
+      "step": 5420
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.00376959665771e-05,
+      "loss": 1.3622,
+      "step": 5421
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.9981338573594876e-05,
+      "loss": 1.2789,
+      "step": 5422
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.9925016467888706e-05,
+      "loss": 1.4904,
+      "step": 5423
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.9868729661878754e-05,
+      "loss": 1.5074,
+      "step": 5424
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.98124781679774e-05,
+      "loss": 1.4103,
+      "step": 5425
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.975626199858911e-05,
+      "loss": 1.4181,
+      "step": 5426
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.97000811661107e-05,
+      "loss": 1.3087,
+      "step": 5427
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.964393568293117e-05,
+      "loss": 1.3996,
+      "step": 5428
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.958782556143175e-05,
+      "loss": 1.14,
+      "step": 5429
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.953175081398568e-05,
+      "loss": 1.3396,
+      "step": 5430
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.9475711452958765e-05,
+      "loss": 1.5284,
+      "step": 5431
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.9419707490708626e-05,
+      "loss": 1.4733,
+      "step": 5432
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.936373893958527e-05,
+      "loss": 1.2852,
+      "step": 5433
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.930780581193092e-05,
+      "loss": 1.4066,
+      "step": 5434
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.925190812007991e-05,
+      "loss": 1.5386,
+      "step": 5435
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.919604587635883e-05,
+      "loss": 1.5034,
+      "step": 5436
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.914021909308625e-05,
+      "loss": 1.4842,
+      "step": 5437
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.908442778257331e-05,
+      "loss": 1.3435,
+      "step": 5438
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.902867195712292e-05,
+      "loss": 1.5665,
+      "step": 5439
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.8972951629030384e-05,
+      "loss": 1.4112,
+      "step": 5440
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.891726681058316e-05,
+      "loss": 1.4449,
+      "step": 5441
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.88616175140609e-05,
+      "loss": 1.2618,
+      "step": 5442
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.8806003751735265e-05,
+      "loss": 1.2105,
+      "step": 5443
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.8750425535870226e-05,
+      "loss": 1.5608,
+      "step": 5444
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.869488287872189e-05,
+      "loss": 1.3411,
+      "step": 5445
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.863937579253858e-05,
+      "loss": 1.1692,
+      "step": 5446
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.858390428956056e-05,
+      "loss": 1.3905,
+      "step": 5447
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.8528468382020466e-05,
+      "loss": 1.4086,
+      "step": 5448
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.8473068082143014e-05,
+      "loss": 1.2676,
+      "step": 5449
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.841770340214512e-05,
+      "loss": 1.6083,
+      "step": 5450
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.836237435423562e-05,
+      "loss": 1.4299,
+      "step": 5451
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.830708095061582e-05,
+      "loss": 1.6556,
+      "step": 5452
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.825182320347902e-05,
+      "loss": 1.3534,
+      "step": 5453
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.819660112501053e-05,
+      "loss": 1.3361,
+      "step": 5454
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.814141472738797e-05,
+      "loss": 1.6271,
+      "step": 5455
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.8086264022781015e-05,
+      "loss": 1.3883,
+      "step": 5456
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.803114902335156e-05,
+      "loss": 1.4374,
+      "step": 5457
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.7976069741253425e-05,
+      "loss": 1.3873,
+      "step": 5458
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.7921026188632756e-05,
+      "loss": 1.4819,
+      "step": 5459
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.786601837762771e-05,
+      "loss": 1.6438,
+      "step": 5460
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.781104632036869e-05,
+      "loss": 1.4796,
+      "step": 5461
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.775611002897792e-05,
+      "loss": 1.5717,
+      "step": 5462
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.7701209515570194e-05,
+      "loss": 1.5448,
+      "step": 5463
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.764634479225195e-05,
+      "loss": 1.4254,
+      "step": 5464
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.759151587112204e-05,
+      "loss": 1.373,
+      "step": 5465
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.753672276427129e-05,
+      "loss": 1.6291,
+      "step": 5466
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.7481965483782755e-05,
+      "loss": 1.4768,
+      "step": 5467
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.742724404173135e-05,
+      "loss": 1.3676,
+      "step": 5468
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.7372558450184256e-05,
+      "loss": 1.3877,
+      "step": 5469
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.731790872120089e-05,
+      "loss": 1.1739,
+      "step": 5470
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.726329486683242e-05,
+      "loss": 1.3639,
+      "step": 5471
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.720871689912242e-05,
+      "loss": 1.3027,
+      "step": 5472
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.715417483010619e-05,
+      "loss": 1.5576,
+      "step": 5473
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.7099668671811604e-05,
+      "loss": 1.4069,
+      "step": 5474
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.704519843625818e-05,
+      "loss": 1.5033,
+      "step": 5475
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.69907641354577e-05,
+      "loss": 1.4811,
+      "step": 5476
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.6936365781414015e-05,
+      "loss": 1.2377,
+      "step": 5477
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.688200338612311e-05,
+      "loss": 1.4544,
+      "step": 5478
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.6827676961572856e-05,
+      "loss": 1.3379,
+      "step": 5479
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.677338651974334e-05,
+      "loss": 1.2493,
+      "step": 5480
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.671913207260667e-05,
+      "loss": 1.393,
+      "step": 5481
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.666491363212712e-05,
+      "loss": 1.4065,
+      "step": 5482
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.6610731210260706e-05,
+      "loss": 1.469,
+      "step": 5483
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.6556584818956005e-05,
+      "loss": 1.3008,
+      "step": 5484
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.650247447015316e-05,
+      "loss": 1.2502,
+      "step": 5485
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.6448400175784614e-05,
+      "loss": 1.2989,
+      "step": 5486
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.6394361947774857e-05,
+      "loss": 1.3109,
+      "step": 5487
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.6340359798040355e-05,
+      "loss": 1.5836,
+      "step": 5488
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.628639373848974e-05,
+      "loss": 1.394,
+      "step": 5489
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.623246378102343e-05,
+      "loss": 1.2395,
+      "step": 5490
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.617856993753417e-05,
+      "loss": 1.7199,
+      "step": 5491
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.612471221990654e-05,
+      "loss": 1.5018,
+      "step": 5492
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.607089064001734e-05,
+      "loss": 1.1023,
+      "step": 5493
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.6017105209735114e-05,
+      "loss": 1.4821,
+      "step": 5494
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.5963355940920816e-05,
+      "loss": 1.279,
+      "step": 5495
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.590964284542708e-05,
+      "loss": 1.4727,
+      "step": 5496
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.585596593509877e-05,
+      "loss": 1.4236,
+      "step": 5497
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.580232522177265e-05,
+      "loss": 1.476,
+      "step": 5498
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.580232522177265e-05,
+      "loss": 1.4205,
+      "step": 5499
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.5748720717277664e-05,
+      "loss": 1.4192,
+      "step": 5500
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.569515243343453e-05,
+      "loss": 1.3757,
+      "step": 5501
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.5641620382056165e-05,
+      "loss": 1.3022,
+      "step": 5502
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.558812457494747e-05,
+      "loss": 1.4771,
+      "step": 5503
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.55346650239053e-05,
+      "loss": 1.1859,
+      "step": 5504
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.548124174071861e-05,
+      "loss": 1.4009,
+      "step": 5505
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.542785473716812e-05,
+      "loss": 1.3007,
+      "step": 5506
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.537450402502695e-05,
+      "loss": 1.4874,
+      "step": 5507
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.5321189616059815e-05,
+      "loss": 1.6036,
+      "step": 5508
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.5267911522023625e-05,
+      "loss": 1.3516,
+      "step": 5509
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.521466975466729e-05,
+      "loss": 1.5061,
+      "step": 5510
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.516146432573171e-05,
+      "loss": 1.2611,
+      "step": 5511
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.510829524694963e-05,
+      "loss": 1.3996,
+      "step": 5512
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.505516253004593e-05,
+      "loss": 1.5436,
+      "step": 5513
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.50020661867374e-05,
+      "loss": 1.4408,
+      "step": 5514
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.4949006228732936e-05,
+      "loss": 1.5978,
+      "step": 5515
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.489598266773313e-05,
+      "loss": 1.3537,
+      "step": 5516
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.484299551543091e-05,
+      "loss": 1.4177,
+      "step": 5517
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.479004478351087e-05,
+      "loss": 1.2167,
+      "step": 5518
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.473713048364975e-05,
+      "loss": 1.3855,
+      "step": 5519
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.468425262751609e-05,
+      "loss": 1.3592,
+      "step": 5520
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.463141122677063e-05,
+      "loss": 1.4483,
+      "step": 5521
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.457860629306595e-05,
+      "loss": 1.1791,
+      "step": 5522
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.4525837838046485e-05,
+      "loss": 1.2406,
+      "step": 5523
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.447310587334878e-05,
+      "loss": 1.4013,
+      "step": 5524
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.442041041060122e-05,
+      "loss": 1.4011,
+      "step": 5525
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.436775146142433e-05,
+      "loss": 1.3905,
+      "step": 5526
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.4315129037430236e-05,
+      "loss": 1.5478,
+      "step": 5527
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.426254315022344e-05,
+      "loss": 1.4231,
+      "step": 5528
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.420999381140002e-05,
+      "loss": 1.4064,
+      "step": 5529
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.415748103254821e-05,
+      "loss": 1.155,
+      "step": 5530
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.41050048252481e-05,
+      "loss": 1.3722,
+      "step": 5531
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.405256520107176e-05,
+      "loss": 1.6292,
+      "step": 5532
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.400016217158308e-05,
+      "loss": 1.429,
+      "step": 5533
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.394779574833804e-05,
+      "loss": 1.6849,
+      "step": 5534
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.389546594288444e-05,
+      "loss": 1.3938,
+      "step": 5535
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.384317276676208e-05,
+      "loss": 1.5461,
+      "step": 5536
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.379091623150257e-05,
+      "loss": 1.4347,
+      "step": 5537
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.3738696348629476e-05,
+      "loss": 1.4733,
+      "step": 5538
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.368651312965847e-05,
+      "loss": 1.5371,
+      "step": 5539
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.363436658609684e-05,
+      "loss": 1.5441,
+      "step": 5540
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.3582256729443996e-05,
+      "loss": 1.3827,
+      "step": 5541
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.353018357119118e-05,
+      "loss": 1.2546,
+      "step": 5542
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.347814712282158e-05,
+      "loss": 1.521,
+      "step": 5543
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.342614739581018e-05,
+      "loss": 1.5451,
+      "step": 5544
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.3374184401623985e-05,
+      "loss": 1.4052,
+      "step": 5545
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.332225815172187e-05,
+      "loss": 1.2439,
+      "step": 5546
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.3270368657554664e-05,
+      "loss": 1.2423,
+      "step": 5547
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.3218515930564905e-05,
+      "loss": 1.4271,
+      "step": 5548
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.31666999821872e-05,
+      "loss": 1.5257,
+      "step": 5549
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.3114920823847994e-05,
+      "loss": 1.4751,
+      "step": 5550
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.306317846696565e-05,
+      "loss": 1.4103,
+      "step": 5551
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.301147292295026e-05,
+      "loss": 1.5196,
+      "step": 5552
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.295980420320408e-05,
+      "loss": 1.3678,
+      "step": 5553
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.290817231912096e-05,
+      "loss": 1.5096,
+      "step": 5554
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.2856577282086776e-05,
+      "loss": 1.3245,
+      "step": 5555
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.2805019103479264e-05,
+      "loss": 1.5063,
+      "step": 5556
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.275349779466805e-05,
+      "loss": 1.3588,
+      "step": 5557
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.2702013367014594e-05,
+      "loss": 1.4109,
+      "step": 5558
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.2650565831872135e-05,
+      "loss": 1.3435,
+      "step": 5559
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.259915520058605e-05,
+      "loss": 1.5226,
+      "step": 5560
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.2547781484493225e-05,
+      "loss": 1.3636,
+      "step": 5561
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.2496444694922726e-05,
+      "loss": 1.3331,
+      "step": 5562
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.244514484319512e-05,
+      "loss": 1.5301,
+      "step": 5563
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.2393881940623294e-05,
+      "loss": 1.2854,
+      "step": 5564
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.2342655998511564e-05,
+      "loss": 1.4876,
+      "step": 5565
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.229146702815629e-05,
+      "loss": 1.4413,
+      "step": 5566
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.224031504084566e-05,
+      "loss": 1.3174,
+      "step": 5567
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.218920004785977e-05,
+      "loss": 1.2449,
+      "step": 5568
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.2138122060470356e-05,
+      "loss": 1.3858,
+      "step": 5569
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.2087081089941205e-05,
+      "loss": 1.5181,
+      "step": 5570
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.203607714752783e-05,
+      "loss": 1.3734,
+      "step": 5571
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.1985110244477636e-05,
+      "loss": 1.477,
+      "step": 5572
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.19341803920298e-05,
+      "loss": 1.4538,
+      "step": 5573
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.1883287601415366e-05,
+      "loss": 1.4095,
+      "step": 5574
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.183243188385729e-05,
+      "loss": 1.553,
+      "step": 5575
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.1781613250570116e-05,
+      "loss": 1.4122,
+      "step": 5576
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.173083171276041e-05,
+      "loss": 1.0933,
+      "step": 5577
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.168008728162655e-05,
+      "loss": 1.5779,
+      "step": 5578
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.162937996835869e-05,
+      "loss": 1.3753,
+      "step": 5579
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.1578709784138685e-05,
+      "loss": 1.5041,
+      "step": 5580
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.15280767401404e-05,
+      "loss": 1.3093,
+      "step": 5581
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.1477480847529414e-05,
+      "loss": 1.4258,
+      "step": 5582
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.1426922117463143e-05,
+      "loss": 1.1847,
+      "step": 5583
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.137640056109068e-05,
+      "loss": 1.3069,
+      "step": 5584
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.132591618955318e-05,
+      "loss": 1.4275,
+      "step": 5585
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.1275469013983326e-05,
+      "loss": 1.3718,
+      "step": 5586
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.122505904550577e-05,
+      "loss": 1.3596,
+      "step": 5587
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.117468629523688e-05,
+      "loss": 1.2283,
+      "step": 5588
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.112435077428484e-05,
+      "loss": 1.2672,
+      "step": 5589
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.1074052493749704e-05,
+      "loss": 1.3182,
+      "step": 5590
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.102379146472312e-05,
+      "loss": 1.4585,
+      "step": 5591
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.09735676982887e-05,
+      "loss": 1.4808,
+      "step": 5592
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.092338120552176e-05,
+      "loss": 1.2839,
+      "step": 5593
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.0873231997489484e-05,
+      "loss": 1.5489,
+      "step": 5594
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.082312008525059e-05,
+      "loss": 1.2879,
+      "step": 5595
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.077304547985596e-05,
+      "loss": 1.3761,
+      "step": 5596
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.072300819234788e-05,
+      "loss": 1.3642,
+      "step": 5597
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.06730082337606e-05,
+      "loss": 1.4671,
+      "step": 5598
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.062304561512011e-05,
+      "loss": 1.5943,
+      "step": 5599
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.0573120347444216e-05,
+      "loss": 1.4888,
+      "step": 5600
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.052323244174229e-05,
+      "loss": 1.4863,
+      "step": 5601
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.0473381909015698e-05,
+      "loss": 1.2278,
+      "step": 5602
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.0423568760257427e-05,
+      "loss": 1.2932,
+      "step": 5603
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.0373793006452332e-05,
+      "loss": 1.4778,
+      "step": 5604
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.0324054658576794e-05,
+      "loss": 1.3033,
+      "step": 5605
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.027435372759926e-05,
+      "loss": 1.4266,
+      "step": 5606
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.0224690224479735e-05,
+      "loss": 1.4435,
+      "step": 5607
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.0175064160169953e-05,
+      "loss": 1.5296,
+      "step": 5608
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.0125475545613447e-05,
+      "loss": 1.4119,
+      "step": 5609
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.007592439174549e-05,
+      "loss": 1.4414,
+      "step": 5610
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.0026410709493193e-05,
+      "loss": 1.344,
+      "step": 5611
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9976934509775125e-05,
+      "loss": 1.4368,
+      "step": 5612
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.992749580350187e-05,
+      "loss": 1.4594,
+      "step": 5613
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9878094601575625e-05,
+      "loss": 1.3589,
+      "step": 5614
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.982873091489038e-05,
+      "loss": 1.4032,
+      "step": 5615
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.977940475433163e-05,
+      "loss": 1.222,
+      "step": 5616
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9730116130777007e-05,
+      "loss": 1.4179,
+      "step": 5617
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9680865055095465e-05,
+      "loss": 1.3589,
+      "step": 5618
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9631651538147864e-05,
+      "loss": 1.3933,
+      "step": 5619
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9582475590786773e-05,
+      "loss": 1.5989,
+      "step": 5620
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9533337223856515e-05,
+      "loss": 1.63,
+      "step": 5621
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9484236448192983e-05,
+      "loss": 1.2496,
+      "step": 5622
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9435173274623886e-05,
+      "loss": 1.2991,
+      "step": 5623
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.938614771396866e-05,
+      "loss": 1.4532,
+      "step": 5624
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9337159777038368e-05,
+      "loss": 1.4734,
+      "step": 5625
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9288209474635888e-05,
+      "loss": 1.4451,
+      "step": 5626
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9239296817555596e-05,
+      "loss": 1.2753,
+      "step": 5627
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.919042181658387e-05,
+      "loss": 1.4777,
+      "step": 5628
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9141584482498487e-05,
+      "loss": 1.4128,
+      "step": 5629
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.9092784826069075e-05,
+      "loss": 1.2176,
+      "step": 5630
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.904402285805694e-05,
+      "loss": 1.6544,
+      "step": 5631
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.899529858921508e-05,
+      "loss": 1.3355,
+      "step": 5632
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.89466120302881e-05,
+      "loss": 1.5183,
+      "step": 5633
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.8897963192012366e-05,
+      "loss": 1.4567,
+      "step": 5634
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.8849352085115922e-05,
+      "loss": 1.2068,
+      "step": 5635
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.88007787203185e-05,
+      "loss": 1.3477,
+      "step": 5636
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.8752243108331445e-05,
+      "loss": 1.2755,
+      "step": 5637
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.8703745259857817e-05,
+      "loss": 1.5816,
+      "step": 5638
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.8655285185592375e-05,
+      "loss": 1.1035,
+      "step": 5639
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.8606862896221497e-05,
+      "loss": 1.5514,
+      "step": 5640
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.8558478402423294e-05,
+      "loss": 1.4164,
+      "step": 5641
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.851013171486747e-05,
+      "loss": 1.3477,
+      "step": 5642
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.846182284421548e-05,
+      "loss": 1.6802,
+      "step": 5643
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.8413551801120286e-05,
+      "loss": 1.5233,
+      "step": 5644
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.8365318596226685e-05,
+      "loss": 1.4039,
+      "step": 5645
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.8317123240170995e-05,
+      "loss": 1.3734,
+      "step": 5646
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.826896574358133e-05,
+      "loss": 1.3079,
+      "step": 5647
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.8220846117077248e-05,
+      "loss": 1.3617,
+      "step": 5648
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.817276437127021e-05,
+      "loss": 1.2281,
+      "step": 5649
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.8124720516763092e-05,
+      "loss": 1.4767,
+      "step": 5650
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.807671456415053e-05,
+      "loss": 1.4368,
+      "step": 5651
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.802874652401881e-05,
+      "loss": 1.5231,
+      "step": 5652
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7980816406945875e-05,
+      "loss": 1.4303,
+      "step": 5653
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7932924223501157e-05,
+      "loss": 1.5416,
+      "step": 5654
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.788506998424589e-05,
+      "loss": 1.4784,
+      "step": 5655
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7837253699732867e-05,
+      "loss": 1.3019,
+      "step": 5656
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7789475380506536e-05,
+      "loss": 1.5988,
+      "step": 5657
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.774173503710298e-05,
+      "loss": 1.5256,
+      "step": 5658
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7694032680049774e-05,
+      "loss": 1.3046,
+      "step": 5659
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7646368319866423e-05,
+      "loss": 1.3386,
+      "step": 5660
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7598741967063692e-05,
+      "loss": 1.4228,
+      "step": 5661
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.75511536321442e-05,
+      "loss": 1.3942,
+      "step": 5662
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.750360332560211e-05,
+      "loss": 1.4614,
+      "step": 5663
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7456091057923262e-05,
+      "loss": 1.7335,
+      "step": 5664
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.740861683958491e-05,
+      "loss": 1.3091,
+      "step": 5665
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7361180681056174e-05,
+      "loss": 1.2964,
+      "step": 5666
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7313782592797597e-05,
+      "loss": 1.5973,
+      "step": 5667
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7266422585261485e-05,
+      "loss": 1.2554,
+      "step": 5668
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.721910066889155e-05,
+      "loss": 1.1746,
+      "step": 5669
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7171816854123267e-05,
+      "loss": 1.4222,
+      "step": 5670
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7124571151383628e-05,
+      "loss": 1.2485,
+      "step": 5671
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7077363571091295e-05,
+      "loss": 1.2763,
+      "step": 5672
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.703019412365633e-05,
+      "loss": 1.2597,
+      "step": 5673
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.6983062819480685e-05,
+      "loss": 1.3214,
+      "step": 5674
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.693596966895775e-05,
+      "loss": 1.3359,
+      "step": 5675
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.6888914682472366e-05,
+      "loss": 1.2984,
+      "step": 5676
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.6841897870401166e-05,
+      "loss": 1.3314,
+      "step": 5677
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.679491924311226e-05,
+      "loss": 1.3865,
+      "step": 5678
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.6747978810965446e-05,
+      "loss": 1.3369,
+      "step": 5679
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.6701076584311847e-05,
+      "loss": 1.2286,
+      "step": 5680
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.665421257349452e-05,
+      "loss": 1.4352,
+      "step": 5681
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.6607386788847754e-05,
+      "loss": 1.4738,
+      "step": 5682
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.6560599240697692e-05,
+      "loss": 1.4879,
+      "step": 5683
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.6513849939361745e-05,
+      "loss": 1.4231,
+      "step": 5684
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.6467138895149247e-05,
+      "loss": 1.2884,
+      "step": 5685
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.6420466118360755e-05,
+      "loss": 1.2765,
+      "step": 5686
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.6373831619288613e-05,
+      "loss": 1.5129,
+      "step": 5687
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.6327235408216643e-05,
+      "loss": 1.3225,
+      "step": 5688
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.6280677495420248e-05,
+      "loss": 1.3392,
+      "step": 5689
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.62341578911663e-05,
+      "loss": 1.3996,
+      "step": 5690
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.6187676605713285e-05,
+      "loss": 1.3114,
+      "step": 5691
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.61412336493114e-05,
+      "loss": 1.3109,
+      "step": 5692
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.609482903220206e-05,
+      "loss": 1.3778,
+      "step": 5693
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.6048462764618477e-05,
+      "loss": 1.3621,
+      "step": 5694
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.600213485678531e-05,
+      "loss": 1.2901,
+      "step": 5695
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.5955845318918838e-05,
+      "loss": 1.3234,
+      "step": 5696
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.5909594161226715e-05,
+      "loss": 1.5922,
+      "step": 5697
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.586338139390827e-05,
+      "loss": 1.3163,
+      "step": 5698
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.581720702715438e-05,
+      "loss": 1.2307,
+      "step": 5699
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.5771071071147402e-05,
+      "loss": 1.2797,
+      "step": 5700
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.5724973536061136e-05,
+      "loss": 1.4036,
+      "step": 5701
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.5678914432061075e-05,
+      "loss": 1.3041,
+      "step": 5702
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.5632893769304133e-05,
+      "loss": 1.257,
+      "step": 5703
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.558691155793884e-05,
+      "loss": 1.3189,
+      "step": 5704
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.5540967808105042e-05,
+      "loss": 1.3799,
+      "step": 5705
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.5495062529934432e-05,
+      "loss": 1.4018,
+      "step": 5706
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.5449195733549868e-05,
+      "loss": 1.4688,
+      "step": 5707
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.540336742906595e-05,
+      "loss": 1.4057,
+      "step": 5708
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.5357577626588726e-05,
+      "loss": 1.4562,
+      "step": 5709
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.5311826336215737e-05,
+      "loss": 1.602,
+      "step": 5710
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.526611356803612e-05,
+      "loss": 1.2746,
+      "step": 5711
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.522043933213032e-05,
+      "loss": 1.4424,
+      "step": 5712
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.5174803638570478e-05,
+      "loss": 1.4371,
+      "step": 5713
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.5129206497420142e-05,
+      "loss": 1.4948,
+      "step": 5714
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.5083647918734456e-05,
+      "loss": 1.621,
+      "step": 5715
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.5038127912559816e-05,
+      "loss": 1.3329,
+      "step": 5716
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.499264648893449e-05,
+      "loss": 1.4879,
+      "step": 5717
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.4947203657887897e-05,
+      "loss": 1.4777,
+      "step": 5718
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.490179942944113e-05,
+      "loss": 1.4584,
+      "step": 5719
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.4856433813606694e-05,
+      "loss": 1.3167,
+      "step": 5720
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.4811106820388652e-05,
+      "loss": 1.367,
+      "step": 5721
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.4765818459782408e-05,
+      "loss": 1.5315,
+      "step": 5722
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.472056874177502e-05,
+      "loss": 1.1017,
+      "step": 5723
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.4675357676344902e-05,
+      "loss": 1.4866,
+      "step": 5724
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.4630185273462014e-05,
+      "loss": 1.4014,
+      "step": 5725
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.458505154308779e-05,
+      "loss": 1.2086,
+      "step": 5726
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.4539956495175e-05,
+      "loss": 1.331,
+      "step": 5727
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.4494900139668152e-05,
+      "loss": 1.295,
+      "step": 5728
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.444988248650293e-05,
+      "loss": 1.3493,
+      "step": 5729
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.440490354560665e-05,
+      "loss": 1.2728,
+      "step": 5730
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.4359963326898094e-05,
+      "loss": 1.5224,
+      "step": 5731
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.431506184028747e-05,
+      "loss": 1.432,
+      "step": 5732
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.4270199095676384e-05,
+      "loss": 1.4744,
+      "step": 5733
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.422537510295799e-05,
+      "loss": 1.3768,
+      "step": 5734
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.4180589872016855e-05,
+      "loss": 1.4671,
+      "step": 5735
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.4135843412729077e-05,
+      "loss": 1.2906,
+      "step": 5736
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.4091135734961977e-05,
+      "loss": 1.2128,
+      "step": 5737
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.4046466848574657e-05,
+      "loss": 1.4377,
+      "step": 5738
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.4001836763417385e-05,
+      "loss": 1.2902,
+      "step": 5739
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.3957245489331982e-05,
+      "loss": 1.479,
+      "step": 5740
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.3912693036151735e-05,
+      "loss": 1.57,
+      "step": 5741
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.386817941370132e-05,
+      "loss": 1.4542,
+      "step": 5742
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.3823704631796927e-05,
+      "loss": 1.6197,
+      "step": 5743
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.3779268700246048e-05,
+      "loss": 1.3269,
+      "step": 5744
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.3734871628847688e-05,
+      "loss": 1.4762,
+      "step": 5745
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.3690513427392325e-05,
+      "loss": 1.4862,
+      "step": 5746
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.364619410566182e-05,
+      "loss": 1.3103,
+      "step": 5747
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.3601913673429364e-05,
+      "loss": 1.4912,
+      "step": 5748
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.3557672140459808e-05,
+      "loss": 1.5579,
+      "step": 5749
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.351346951650919e-05,
+      "loss": 1.2179,
+      "step": 5750
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.346930581132507e-05,
+      "loss": 1.6732,
+      "step": 5751
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.3425181034646438e-05,
+      "loss": 1.5588,
+      "step": 5752
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.338109519620373e-05,
+      "loss": 1.4159,
+      "step": 5753
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.3337048305718634e-05,
+      "loss": 1.334,
+      "step": 5754
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.3293040372904407e-05,
+      "loss": 1.1805,
+      "step": 5755
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.3249071407465682e-05,
+      "loss": 1.4347,
+      "step": 5756
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.320514141909853e-05,
+      "loss": 1.2724,
+      "step": 5757
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.316125041749029e-05,
+      "loss": 1.4889,
+      "step": 5758
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.311739841231977e-05,
+      "loss": 1.2955,
+      "step": 5759
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.307358541325737e-05,
+      "loss": 1.4477,
+      "step": 5760
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.3029811429964564e-05,
+      "loss": 1.2804,
+      "step": 5761
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.298607647209443e-05,
+      "loss": 1.6093,
+      "step": 5762
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2942380549291408e-05,
+      "loss": 1.1698,
+      "step": 5763
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2898723671191348e-05,
+      "loss": 1.3219,
+      "step": 5764
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2855105847421343e-05,
+      "loss": 1.314,
+      "step": 5765
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.281152708760006e-05,
+      "loss": 1.4982,
+      "step": 5766
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2767987401337455e-05,
+      "loss": 1.1462,
+      "step": 5767
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2724486798234956e-05,
+      "loss": 1.3682,
+      "step": 5768
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.268102528788516e-05,
+      "loss": 1.3894,
+      "step": 5769
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2637602879872354e-05,
+      "loss": 1.3269,
+      "step": 5770
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2594219583771902e-05,
+      "loss": 1.5558,
+      "step": 5771
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2550875409150794e-05,
+      "loss": 1.5287,
+      "step": 5772
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2507570365567144e-05,
+      "loss": 1.493,
+      "step": 5773
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2464304462570705e-05,
+      "loss": 1.4676,
+      "step": 5774
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2421077709702366e-05,
+      "loss": 1.3787,
+      "step": 5775
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2377890116494515e-05,
+      "loss": 1.4344,
+      "step": 5776
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.233474169247087e-05,
+      "loss": 1.5035,
+      "step": 5777
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2291632447146517e-05,
+      "loss": 1.1926,
+      "step": 5778
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2248562390027906e-05,
+      "loss": 1.157,
+      "step": 5779
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2205531530612756e-05,
+      "loss": 1.3813,
+      "step": 5780
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2162539878390364e-05,
+      "loss": 1.3655,
+      "step": 5781
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.211958744284113e-05,
+      "loss": 1.4617,
+      "step": 5782
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2076674233436934e-05,
+      "loss": 1.5107,
+      "step": 5783
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.203380025964099e-05,
+      "loss": 1.6129,
+      "step": 5784
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1990965530907913e-05,
+      "loss": 1.2733,
+      "step": 5785
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1948170056683525e-05,
+      "loss": 1.2772,
+      "step": 5786
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1905413846405077e-05,
+      "loss": 1.4288,
+      "step": 5787
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1862696909501203e-05,
+      "loss": 1.5518,
+      "step": 5788
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1820019255391855e-05,
+      "loss": 1.4533,
+      "step": 5789
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.177738089348822e-05,
+      "loss": 1.4815,
+      "step": 5790
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.173478183319293e-05,
+      "loss": 1.4901,
+      "step": 5791
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1692222083899937e-05,
+      "loss": 1.4061,
+      "step": 5792
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.164970165499449e-05,
+      "loss": 1.6294,
+      "step": 5793
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.16072205558532e-05,
+      "loss": 1.3223,
+      "step": 5794
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1564778795843955e-05,
+      "loss": 1.2952,
+      "step": 5795
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1522376384326082e-05,
+      "loss": 1.4372,
+      "step": 5796
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.148001333065006e-05,
+      "loss": 1.4794,
+      "step": 5797
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.14376896441578e-05,
+      "loss": 1.3639,
+      "step": 5798
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1395405334182516e-05,
+      "loss": 1.3271,
+      "step": 5799
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.13531604100488e-05,
+      "loss": 1.3759,
+      "step": 5800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1310954881072376e-05,
+      "loss": 1.5447,
+      "step": 5801
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1268788756560466e-05,
+      "loss": 1.4888,
+      "step": 5802
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.122666204581152e-05,
+      "loss": 1.3658,
+      "step": 5803
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.118457475811535e-05,
+      "loss": 1.571,
+      "step": 5804
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1142526902752933e-05,
+      "loss": 1.3688,
+      "step": 5805
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1100518488996767e-05,
+      "loss": 1.4804,
+      "step": 5806
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1058549526110462e-05,
+      "loss": 1.4462,
+      "step": 5807
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1016620023349032e-05,
+      "loss": 1.4171,
+      "step": 5808
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.097472998995873e-05,
+      "loss": 1.4541,
+      "step": 5809
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.093287943517719e-05,
+      "loss": 1.4924,
+      "step": 5810
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.089106836823329e-05,
+      "loss": 1.1834,
+      "step": 5811
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.084929679834706e-05,
+      "loss": 1.5365,
+      "step": 5812
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.0807564734730177e-05,
+      "loss": 1.3627,
+      "step": 5813
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.076587218658521e-05,
+      "loss": 1.2397,
+      "step": 5814
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.0724219163106253e-05,
+      "loss": 1.4001,
+      "step": 5815
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.0682605673478594e-05,
+      "loss": 1.3519,
+      "step": 5816
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.0641031726878924e-05,
+      "loss": 1.3713,
+      "step": 5817
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.059949733247497e-05,
+      "loss": 1.4798,
+      "step": 5818
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.055800249942599e-05,
+      "loss": 1.287,
+      "step": 5819
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.0516547236882387e-05,
+      "loss": 1.302,
+      "step": 5820
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.0475131553985905e-05,
+      "loss": 1.3592,
+      "step": 5821
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.043375545986943e-05,
+      "loss": 1.2779,
+      "step": 5822
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.0392418963657265e-05,
+      "loss": 1.4448,
+      "step": 5823
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.0351122074464922e-05,
+      "loss": 1.2602,
+      "step": 5824
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.0309864801399205e-05,
+      "loss": 1.2557,
+      "step": 5825
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.0268647153558074e-05,
+      "loss": 1.5478,
+      "step": 5826
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.022746914003091e-05,
+      "loss": 1.4129,
+      "step": 5827
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.018633076989831e-05,
+      "loss": 1.5611,
+      "step": 5828
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.0145232052231998e-05,
+      "loss": 1.5384,
+      "step": 5829
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.0104172996095106e-05,
+      "loss": 1.5918,
+      "step": 5830
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.006315361054196e-05,
+      "loss": 1.1963,
+      "step": 5831
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.002217390461818e-05,
+      "loss": 1.5069,
+      "step": 5832
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.998123388736053e-05,
+      "loss": 1.3144,
+      "step": 5833
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.994033356779714e-05,
+      "loss": 1.1599,
+      "step": 5834
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.9899472954947342e-05,
+      "loss": 1.3776,
+      "step": 5835
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.9858652057821714e-05,
+      "loss": 1.5684,
+      "step": 5836
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.9817870885421975e-05,
+      "loss": 1.6551,
+      "step": 5837
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.9777129446741328e-05,
+      "loss": 1.4947,
+      "step": 5838
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.9736427750763963e-05,
+      "loss": 1.3453,
+      "step": 5839
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.969576580646544e-05,
+      "loss": 1.1923,
+      "step": 5840
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.9655143622812533e-05,
+      "loss": 1.3905,
+      "step": 5841
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.9614561208763237e-05,
+      "loss": 1.4116,
+      "step": 5842
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.9574018573266727e-05,
+      "loss": 1.1953,
+      "step": 5843
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.9533515725263453e-05,
+      "loss": 1.3414,
+      "step": 5844
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.949305267368522e-05,
+      "loss": 1.4846,
+      "step": 5845
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.9452629427454784e-05,
+      "loss": 1.4958,
+      "step": 5846
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.9412245995486368e-05,
+      "loss": 1.3126,
+      "step": 5847
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.9371902386685203e-05,
+      "loss": 1.4982,
+      "step": 5848
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.9331598609947976e-05,
+      "loss": 1.6045,
+      "step": 5849
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.92913346741624e-05,
+      "loss": 1.3899,
+      "step": 5850
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.9251110588207478e-05,
+      "loss": 1.1182,
+      "step": 5851
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.9210926360953408e-05,
+      "loss": 1.3297,
+      "step": 5852
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.9170782001261656e-05,
+      "loss": 1.4556,
+      "step": 5853
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.9130677517984764e-05,
+      "loss": 1.4955,
+      "step": 5854
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.909061291996661e-05,
+      "loss": 1.5893,
+      "step": 5855
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.905058821604222e-05,
+      "loss": 1.472,
+      "step": 5856
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.9010603415037886e-05,
+      "loss": 1.3449,
+      "step": 5857
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.8970658525770914e-05,
+      "loss": 1.1833,
+      "step": 5858
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.8930753557050096e-05,
+      "loss": 1.6214,
+      "step": 5859
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.8890888517675153e-05,
+      "loss": 1.403,
+      "step": 5860
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.8851063416437164e-05,
+      "loss": 1.3798,
+      "step": 5861
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.8811278262118347e-05,
+      "loss": 1.3978,
+      "step": 5862
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.8771533063492107e-05,
+      "loss": 1.257,
+      "step": 5863
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.8731827829323077e-05,
+      "loss": 1.4393,
+      "step": 5864
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.8692162568366968e-05,
+      "loss": 1.5147,
+      "step": 5865
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.8652537289370818e-05,
+      "loss": 1.3774,
+      "step": 5866
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.861295200107276e-05,
+      "loss": 1.4369,
+      "step": 5867
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.857340671220218e-05,
+      "loss": 1.4581,
+      "step": 5868
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.8533901431479483e-05,
+      "loss": 1.473,
+      "step": 5869
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.8494436167616503e-05,
+      "loss": 1.3003,
+      "step": 5870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.8455010929316007e-05,
+      "loss": 1.4353,
+      "step": 5871
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.841562572527209e-05,
+      "loss": 1.4846,
+      "step": 5872
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.8376280564169934e-05,
+      "loss": 1.2563,
+      "step": 5873
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.8336975454686e-05,
+      "loss": 1.371,
+      "step": 5874
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.8297710405487733e-05,
+      "loss": 1.597,
+      "step": 5875
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.8258485425233894e-05,
+      "loss": 1.5732,
+      "step": 5876
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.821930052257439e-05,
+      "loss": 1.2593,
+      "step": 5877
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.8180155706150236e-05,
+      "loss": 1.3768,
+      "step": 5878
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.814105098459371e-05,
+      "loss": 1.4153,
+      "step": 5879
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.8101986366528025e-05,
+      "loss": 1.5059,
+      "step": 5880
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.8062961860567883e-05,
+      "loss": 1.5483,
+      "step": 5881
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.8023977475318833e-05,
+      "loss": 1.4757,
+      "step": 5882
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.7985033219377723e-05,
+      "loss": 1.4506,
+      "step": 5883
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.7946129101332554e-05,
+      "loss": 1.504,
+      "step": 5884
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.7907265129762484e-05,
+      "loss": 1.3314,
+      "step": 5885
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.7868441313237706e-05,
+      "loss": 1.2945,
+      "step": 5886
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.7829657660319698e-05,
+      "loss": 1.2983,
+      "step": 5887
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.779091417956098e-05,
+      "loss": 1.3965,
+      "step": 5888
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.7752210879505314e-05,
+      "loss": 1.379,
+      "step": 5889
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.77135477686875e-05,
+      "loss": 1.3216,
+      "step": 5890
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.767492485563349e-05,
+      "loss": 1.2726,
+      "step": 5891
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.763634214886045e-05,
+      "loss": 1.4149,
+      "step": 5892
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.7597799656876666e-05,
+      "loss": 1.4651,
+      "step": 5893
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.755929738818136e-05,
+      "loss": 1.3915,
+      "step": 5894
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.752083535126521e-05,
+      "loss": 1.4676,
+      "step": 5895
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.748241355460982e-05,
+      "loss": 1.3886,
+      "step": 5896
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.7444032006687893e-05,
+      "loss": 1.4454,
+      "step": 5897
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.740569071596336e-05,
+      "loss": 1.4109,
+      "step": 5898
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.736738969089122e-05,
+      "loss": 1.2293,
+      "step": 5899
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.7329128939917648e-05,
+      "loss": 1.4001,
+      "step": 5900
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.7290908471479805e-05,
+      "loss": 1.4678,
+      "step": 5901
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.725272829400617e-05,
+      "loss": 1.3676,
+      "step": 5902
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.7214588415916145e-05,
+      "loss": 1.2687,
+      "step": 5903
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.717648884562035e-05,
+      "loss": 1.3937,
+      "step": 5904
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.71384295915205e-05,
+      "loss": 1.1829,
+      "step": 5905
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.7100410662009446e-05,
+      "loss": 1.3879,
+      "step": 5906
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.7062432065471024e-05,
+      "loss": 1.6636,
+      "step": 5907
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.702449381028033e-05,
+      "loss": 1.3425,
+      "step": 5908
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.698659590480347e-05,
+      "loss": 1.4045,
+      "step": 5909
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6948738357397742e-05,
+      "loss": 1.3939,
+      "step": 5910
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.691092117641142e-05,
+      "loss": 1.5215,
+      "step": 5911
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.68731443701839e-05,
+      "loss": 1.5312,
+      "step": 5912
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6835407947045835e-05,
+      "loss": 1.4814,
+      "step": 5913
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6797711915318758e-05,
+      "loss": 1.6534,
+      "step": 5914
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.676005628331543e-05,
+      "loss": 1.3479,
+      "step": 5915
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6722441059339645e-05,
+      "loss": 1.6795,
+      "step": 5916
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6684866251686327e-05,
+      "loss": 1.2779,
+      "step": 5917
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6647331868641426e-05,
+      "loss": 1.4723,
+      "step": 5918
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.660983791848203e-05,
+      "loss": 1.3327,
+      "step": 5919
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6572384409476283e-05,
+      "loss": 1.4798,
+      "step": 5920
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6534971349883487e-05,
+      "loss": 1.4876,
+      "step": 5921
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.649759874795389e-05,
+      "loss": 1.4908,
+      "step": 5922
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.64602666119289e-05,
+      "loss": 1.3237,
+      "step": 5923
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6422974950040994e-05,
+      "loss": 1.4854,
+      "step": 5924
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6385723770513774e-05,
+      "loss": 1.5217,
+      "step": 5925
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6348513081561734e-05,
+      "loss": 1.3901,
+      "step": 5926
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6311342891390734e-05,
+      "loss": 1.3492,
+      "step": 5927
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.627421320819742e-05,
+      "loss": 1.4374,
+      "step": 5928
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.623712404016964e-05,
+      "loss": 1.4313,
+      "step": 5929
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6200075395486293e-05,
+      "loss": 1.4363,
+      "step": 5930
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6163067282317336e-05,
+      "loss": 1.3864,
+      "step": 5931
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.612609970882384e-05,
+      "loss": 1.4192,
+      "step": 5932
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6089172683157773e-05,
+      "loss": 1.4048,
+      "step": 5933
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.605228621346242e-05,
+      "loss": 1.4325,
+      "step": 5934
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.6015440307871833e-05,
+      "loss": 1.5093,
+      "step": 5935
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.5978634974511375e-05,
+      "loss": 1.4523,
+      "step": 5936
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.5941870221497247e-05,
+      "loss": 1.4478,
+      "step": 5937
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.5905146056936915e-05,
+      "loss": 1.5958,
+      "step": 5938
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.5868462488928683e-05,
+      "loss": 1.4258,
+      "step": 5939
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.583181952556205e-05,
+      "loss": 1.4432,
+      "step": 5940
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.5795217174917498e-05,
+      "loss": 1.3328,
+      "step": 5941
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5758655445066627e-05,
+      "loss": 1.4439,
+      "step": 5942
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5722134344071925e-05,
+      "loss": 1.2801,
+      "step": 5943
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5685653879987082e-05,
+      "loss": 1.5429,
+      "step": 5944
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5649214060856733e-05,
+      "loss": 1.6657,
+      "step": 5945
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5612814894716578e-05,
+      "loss": 1.4585,
+      "step": 5946
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.557645638959335e-05,
+      "loss": 1.4307,
+      "step": 5947
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5540138553504844e-05,
+      "loss": 1.2484,
+      "step": 5948
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5503861394459895e-05,
+      "loss": 1.3572,
+      "step": 5949
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5467624920458235e-05,
+      "loss": 1.2259,
+      "step": 5950
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.543142913949076e-05,
+      "loss": 1.4477,
+      "step": 5951
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5395274059539376e-05,
+      "loss": 1.4909,
+      "step": 5952
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5359159688577018e-05,
+      "loss": 1.4971,
+      "step": 5953
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5323086034567534e-05,
+      "loss": 1.2357,
+      "step": 5954
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5287053105465942e-05,
+      "loss": 1.2506,
+      "step": 5955
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5251060909218196e-05,
+      "loss": 1.3387,
+      "step": 5956
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5215109453761322e-05,
+      "loss": 1.4986,
+      "step": 5957
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5179198747023205e-05,
+      "loss": 1.2513,
+      "step": 5958
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.514332879692304e-05,
+      "loss": 1.3861,
+      "step": 5959
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5107499611370746e-05,
+      "loss": 1.3542,
+      "step": 5960
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.507171119826738e-05,
+      "loss": 1.372,
+      "step": 5961
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5035963565505007e-05,
+      "loss": 1.3055,
+      "step": 5962
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.5000256720966676e-05,
+      "loss": 1.2843,
+      "step": 5963
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4964590672526513e-05,
+      "loss": 1.3492,
+      "step": 5964
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.492896542804949e-05,
+      "loss": 1.2736,
+      "step": 5965
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4893380995391726e-05,
+      "loss": 1.4671,
+      "step": 5966
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4857837382400274e-05,
+      "loss": 1.4363,
+      "step": 5967
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4822334596913268e-05,
+      "loss": 1.4722,
+      "step": 5968
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4786872646759665e-05,
+      "loss": 1.4143,
+      "step": 5969
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4751451539759632e-05,
+      "loss": 1.4867,
+      "step": 5970
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4716071283724142e-05,
+      "loss": 1.4658,
+      "step": 5971
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4680731886455268e-05,
+      "loss": 1.6191,
+      "step": 5972
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.464543335574604e-05,
+      "loss": 1.4352,
+      "step": 5973
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4610175699380547e-05,
+      "loss": 1.2301,
+      "step": 5974
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4574958925133697e-05,
+      "loss": 1.186,
+      "step": 5975
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4539783040771525e-05,
+      "loss": 1.3003,
+      "step": 5976
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4504648054051007e-05,
+      "loss": 1.248,
+      "step": 5977
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4469553972720118e-05,
+      "loss": 1.2855,
+      "step": 5978
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4434500804517803e-05,
+      "loss": 1.2987,
+      "step": 5979
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4399488557173947e-05,
+      "loss": 1.4866,
+      "step": 5980
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.436451723840948e-05,
+      "loss": 1.4374,
+      "step": 5981
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4329586855936216e-05,
+      "loss": 1.5525,
+      "step": 5982
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4294697417457014e-05,
+      "loss": 1.4084,
+      "step": 5983
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4259848930665721e-05,
+      "loss": 1.2846,
+      "step": 5984
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4225041403247096e-05,
+      "loss": 1.5169,
+      "step": 5985
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4190274842876849e-05,
+      "loss": 1.6052,
+      "step": 5986
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4155549257221733e-05,
+      "loss": 1.4655,
+      "step": 5987
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4120864653939426e-05,
+      "loss": 1.3089,
+      "step": 5988
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4086221040678583e-05,
+      "loss": 1.6142,
+      "step": 5989
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4051618425078694e-05,
+      "loss": 1.4214,
+      "step": 5990
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4017056814770502e-05,
+      "loss": 1.4159,
+      "step": 5991
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.3982536217375375e-05,
+      "loss": 1.325,
+      "step": 5992
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.3948056640505868e-05,
+      "loss": 1.2539,
+      "step": 5993
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.3913618091765347e-05,
+      "loss": 1.619,
+      "step": 5994
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.3879220578748264e-05,
+      "loss": 1.442,
+      "step": 5995
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.3844864109039912e-05,
+      "loss": 1.4759,
+      "step": 5996
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.3810548690216563e-05,
+      "loss": 1.3652,
+      "step": 5997
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.377627432984545e-05,
+      "loss": 1.3823,
+      "step": 5998
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.3742041035484731e-05,
+      "loss": 1.1931,
+      "step": 5999
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.3707848814683611e-05,
+      "loss": 1.296,
+      "step": 6000
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 6790,
+  "num_train_epochs": 1,
+  "save_steps": 1000,
+  "total_flos": 1.0854690183509443e+18,
+  "trial_name": null,
+  "trial_params": null
+}