diff --git "a/checkpoint-14632/trainer_state.json" "b/checkpoint-14632/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-14632/trainer_state.json"
@@ -0,0 +1,87888 @@
+{
+  "best_metric": 0.6621229648590088,
+  "best_model_checkpoint": "english_accents_classification/checkpoint-14632",
+  "epoch": 8.0,
+  "global_step": 14632,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0000000000000002e-07,
+      "loss": 1.6189,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.0000000000000003e-07,
+      "loss": 1.6213,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 1.6219,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 1.5799,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 1.6166,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 1.5887,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4000000000000001e-06,
+      "loss": 1.8102,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 1.7259,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8000000000000001e-06,
+      "loss": 1.616,
+      "step": 9
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 1.6127,
+      "step": 10
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2e-06,
+      "loss": 1.6029,
+      "step": 11
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 1.6301,
+      "step": 12
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6e-06,
+      "loss": 1.588,
+      "step": 13
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 1.6212,
+      "step": 14
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3e-06,
+      "loss": 1.6452,
+      "step": 15
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2000000000000003e-06,
+      "loss": 1.6128,
+      "step": 16
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4000000000000005e-06,
+      "loss": 1.635,
+      "step": 17
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6000000000000003e-06,
+      "loss": 1.6062,
+      "step": 18
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8000000000000005e-06,
+      "loss": 1.5864,
+      "step": 19
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 1.6005,
+      "step": 20
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2000000000000004e-06,
+      "loss": 1.6144,
+      "step": 21
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4e-06,
+      "loss": 1.6035,
+      "step": 22
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.600000000000001e-06,
+      "loss": 1.6002,
+      "step": 23
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 1.5853,
+      "step": 24
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5e-06,
+      "loss": 1.5985,
+      "step": 25
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.2e-06,
+      "loss": 1.6188,
+      "step": 26
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.400000000000001e-06,
+      "loss": 1.619,
+      "step": 27
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.600000000000001e-06,
+      "loss": 1.6021,
+      "step": 28
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.8e-06,
+      "loss": 1.6138,
+      "step": 29
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6e-06,
+      "loss": 1.6141,
+      "step": 30
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.200000000000001e-06,
+      "loss": 1.5941,
+      "step": 31
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.4000000000000006e-06,
+      "loss": 1.5924,
+      "step": 32
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.600000000000001e-06,
+      "loss": 1.6327,
+      "step": 33
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.800000000000001e-06,
+      "loss": 1.5812,
+      "step": 34
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7e-06,
+      "loss": 1.6086,
+      "step": 35
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.2000000000000005e-06,
+      "loss": 1.5793,
+      "step": 36
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.4e-06,
+      "loss": 1.6127,
+      "step": 37
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.600000000000001e-06,
+      "loss": 1.6738,
+      "step": 38
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.800000000000002e-06,
+      "loss": 1.6226,
+      "step": 39
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 1.6036,
+      "step": 40
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.2e-06,
+      "loss": 1.6289,
+      "step": 41
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.400000000000001e-06,
+      "loss": 1.6319,
+      "step": 42
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.6e-06,
+      "loss": 1.5766,
+      "step": 43
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.8e-06,
+      "loss": 1.5967,
+      "step": 44
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9e-06,
+      "loss": 1.5908,
+      "step": 45
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.200000000000002e-06,
+      "loss": 1.6621,
+      "step": 46
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.4e-06,
+      "loss": 1.6298,
+      "step": 47
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.600000000000001e-06,
+      "loss": 1.5815,
+      "step": 48
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.800000000000001e-06,
+      "loss": 1.6429,
+      "step": 49
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1e-05,
+      "loss": 1.6407,
+      "step": 50
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999451754385966e-06,
+      "loss": 1.6047,
+      "step": 51
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99890350877193e-06,
+      "loss": 1.6086,
+      "step": 52
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.998355263157894e-06,
+      "loss": 1.587,
+      "step": 53
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99780701754386e-06,
+      "loss": 1.5865,
+      "step": 54
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.997258771929826e-06,
+      "loss": 1.5956,
+      "step": 55
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99671052631579e-06,
+      "loss": 1.6297,
+      "step": 56
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.996162280701755e-06,
+      "loss": 1.586,
+      "step": 57
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.995614035087721e-06,
+      "loss": 1.64,
+      "step": 58
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.995065789473685e-06,
+      "loss": 1.5891,
+      "step": 59
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.994517543859649e-06,
+      "loss": 1.6219,
+      "step": 60
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.993969298245615e-06,
+      "loss": 1.6415,
+      "step": 61
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.99342105263158e-06,
+      "loss": 1.5934,
+      "step": 62
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.992872807017544e-06,
+      "loss": 1.6102,
+      "step": 63
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.992324561403508e-06,
+      "loss": 1.6195,
+      "step": 64
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.991776315789474e-06,
+      "loss": 1.6313,
+      "step": 65
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99122807017544e-06,
+      "loss": 1.5726,
+      "step": 66
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.990679824561404e-06,
+      "loss": 1.6192,
+      "step": 67
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99013157894737e-06,
+      "loss": 1.6137,
+      "step": 68
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.989583333333333e-06,
+      "loss": 1.5447,
+      "step": 69
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.989035087719299e-06,
+      "loss": 1.623,
+      "step": 70
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.988486842105265e-06,
+      "loss": 1.6013,
+      "step": 71
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.987938596491229e-06,
+      "loss": 1.5958,
+      "step": 72
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.987390350877194e-06,
+      "loss": 1.6498,
+      "step": 73
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.98684210526316e-06,
+      "loss": 1.6095,
+      "step": 74
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.986293859649124e-06,
+      "loss": 1.5938,
+      "step": 75
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.985745614035088e-06,
+      "loss": 1.6111,
+      "step": 76
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.985197368421054e-06,
+      "loss": 1.6085,
+      "step": 77
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.98464912280702e-06,
+      "loss": 1.6821,
+      "step": 78
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.984100877192983e-06,
+      "loss": 1.6746,
+      "step": 79
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.983552631578947e-06,
+      "loss": 1.6056,
+      "step": 80
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.983004385964913e-06,
+      "loss": 1.6372,
+      "step": 81
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.982456140350879e-06,
+      "loss": 1.6186,
+      "step": 82
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.981907894736843e-06,
+      "loss": 1.6331,
+      "step": 83
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.981359649122808e-06,
+      "loss": 1.5904,
+      "step": 84
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.980811403508772e-06,
+      "loss": 1.6199,
+      "step": 85
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.980263157894738e-06,
+      "loss": 1.6174,
+      "step": 86
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.979714912280702e-06,
+      "loss": 1.6147,
+      "step": 87
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.979166666666668e-06,
+      "loss": 1.6235,
+      "step": 88
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.978618421052633e-06,
+      "loss": 1.6243,
+      "step": 89
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.978070175438597e-06,
+      "loss": 1.596,
+      "step": 90
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.977521929824561e-06,
+      "loss": 1.6091,
+      "step": 91
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.976973684210527e-06,
+      "loss": 1.6225,
+      "step": 92
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.976425438596493e-06,
+      "loss": 1.594,
+      "step": 93
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.975877192982456e-06,
+      "loss": 1.6064,
+      "step": 94
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.975328947368422e-06,
+      "loss": 1.62,
+      "step": 95
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.974780701754386e-06,
+      "loss": 1.5988,
+      "step": 96
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.974232456140352e-06,
+      "loss": 1.5904,
+      "step": 97
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.973684210526316e-06,
+      "loss": 1.5998,
+      "step": 98
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.973135964912281e-06,
+      "loss": 1.6162,
+      "step": 99
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.972587719298247e-06,
+      "loss": 1.6053,
+      "step": 100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.972039473684211e-06,
+      "loss": 1.5921,
+      "step": 101
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.971491228070175e-06,
+      "loss": 1.5983,
+      "step": 102
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.97094298245614e-06,
+      "loss": 1.585,
+      "step": 103
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.970394736842106e-06,
+      "loss": 1.6181,
+      "step": 104
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96984649122807e-06,
+      "loss": 1.5644,
+      "step": 105
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.969298245614036e-06,
+      "loss": 1.604,
+      "step": 106
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96875e-06,
+      "loss": 1.6047,
+      "step": 107
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.968201754385966e-06,
+      "loss": 1.5833,
+      "step": 108
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.967653508771931e-06,
+      "loss": 1.5984,
+      "step": 109
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.967105263157895e-06,
+      "loss": 1.6055,
+      "step": 110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.966557017543861e-06,
+      "loss": 1.6131,
+      "step": 111
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.966008771929825e-06,
+      "loss": 1.6359,
+      "step": 112
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96546052631579e-06,
+      "loss": 1.6774,
+      "step": 113
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.964912280701755e-06,
+      "loss": 1.6113,
+      "step": 114
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96436403508772e-06,
+      "loss": 1.5937,
+      "step": 115
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.963815789473686e-06,
+      "loss": 1.6192,
+      "step": 116
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.96326754385965e-06,
+      "loss": 1.5889,
+      "step": 117
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.962719298245614e-06,
+      "loss": 1.6035,
+      "step": 118
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96217105263158e-06,
+      "loss": 1.6195,
+      "step": 119
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.961622807017545e-06,
+      "loss": 1.6099,
+      "step": 120
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.96107456140351e-06,
+      "loss": 1.6361,
+      "step": 121
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.960526315789475e-06,
+      "loss": 1.6067,
+      "step": 122
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959978070175439e-06,
+      "loss": 1.5825,
+      "step": 123
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.959429824561405e-06,
+      "loss": 1.627,
+      "step": 124
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958881578947369e-06,
+      "loss": 1.6084,
+      "step": 125
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.958333333333334e-06,
+      "loss": 1.604,
+      "step": 126
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.9577850877193e-06,
+      "loss": 1.6189,
+      "step": 127
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.957236842105264e-06,
+      "loss": 1.588,
+      "step": 128
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956688596491228e-06,
+      "loss": 1.5951,
+      "step": 129
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.956140350877194e-06,
+      "loss": 1.5881,
+      "step": 130
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.95559210526316e-06,
+      "loss": 1.5897,
+      "step": 131
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.955043859649123e-06,
+      "loss": 1.5934,
+      "step": 132
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.954495614035089e-06,
+      "loss": 1.5825,
+      "step": 133
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953947368421053e-06,
+      "loss": 1.6072,
+      "step": 134
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.953399122807019e-06,
+      "loss": 1.6099,
+      "step": 135
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952850877192983e-06,
+      "loss": 1.567,
+      "step": 136
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.952302631578948e-06,
+      "loss": 1.6098,
+      "step": 137
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.951754385964914e-06,
+      "loss": 1.601,
+      "step": 138
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.951206140350878e-06,
+      "loss": 1.6302,
+      "step": 139
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.950657894736842e-06,
+      "loss": 1.6368,
+      "step": 140
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.950109649122807e-06,
+      "loss": 1.5434,
+      "step": 141
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.949561403508773e-06,
+      "loss": 1.5861,
+      "step": 142
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.949013157894739e-06,
+      "loss": 1.6322,
+      "step": 143
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.948464912280703e-06,
+      "loss": 1.5827,
+      "step": 144
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.947916666666667e-06,
+      "loss": 1.5974,
+      "step": 145
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.947368421052632e-06,
+      "loss": 1.5989,
+      "step": 146
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946820175438598e-06,
+      "loss": 1.6222,
+      "step": 147
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.946271929824562e-06,
+      "loss": 1.6038,
+      "step": 148
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945723684210528e-06,
+      "loss": 1.6322,
+      "step": 149
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945175438596492e-06,
+      "loss": 1.5713,
+      "step": 150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944627192982457e-06,
+      "loss": 1.6067,
+      "step": 151
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.944078947368421e-06,
+      "loss": 1.6134,
+      "step": 152
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.943530701754387e-06,
+      "loss": 1.5978,
+      "step": 153
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942982456140353e-06,
+      "loss": 1.6349,
+      "step": 154
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.942434210526317e-06,
+      "loss": 1.5941,
+      "step": 155
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.94188596491228e-06,
+      "loss": 1.6092,
+      "step": 156
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.941337719298246e-06,
+      "loss": 1.609,
+      "step": 157
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.940789473684212e-06,
+      "loss": 1.5794,
+      "step": 158
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.940241228070176e-06,
+      "loss": 1.577,
+      "step": 159
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.939692982456142e-06,
+      "loss": 1.5796,
+      "step": 160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.939144736842106e-06,
+      "loss": 1.6105,
+      "step": 161
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.938596491228071e-06,
+      "loss": 1.5971,
+      "step": 162
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.938048245614035e-06,
+      "loss": 1.6442,
+      "step": 163
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.937500000000001e-06,
+      "loss": 1.5841,
+      "step": 164
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.936951754385967e-06,
+      "loss": 1.5883,
+      "step": 165
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.93640350877193e-06,
+      "loss": 1.6212,
+      "step": 166
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.935855263157895e-06,
+      "loss": 1.6358,
+      "step": 167
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.93530701754386e-06,
+      "loss": 1.5676,
+      "step": 168
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.934758771929826e-06,
+      "loss": 1.6249,
+      "step": 169
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.93421052631579e-06,
+      "loss": 1.6125,
+      "step": 170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.933662280701754e-06,
+      "loss": 1.5269,
+      "step": 171
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.93311403508772e-06,
+      "loss": 1.6336,
+      "step": 172
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.932565789473685e-06,
+      "loss": 1.6016,
+      "step": 173
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.93201754385965e-06,
+      "loss": 1.608,
+      "step": 174
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.931469298245615e-06,
+      "loss": 1.5853,
+      "step": 175
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.93092105263158e-06,
+      "loss": 1.6313,
+      "step": 176
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.930372807017545e-06,
+      "loss": 1.6094,
+      "step": 177
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.929824561403509e-06,
+      "loss": 1.5985,
+      "step": 178
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.929276315789474e-06,
+      "loss": 1.5984,
+      "step": 179
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.92872807017544e-06,
+      "loss": 1.5962,
+      "step": 180
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.928179824561406e-06,
+      "loss": 1.6055,
+      "step": 181
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.92763157894737e-06,
+      "loss": 1.5979,
+      "step": 182
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.927083333333334e-06,
+      "loss": 1.6045,
+      "step": 183
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.9265350877193e-06,
+      "loss": 1.5985,
+      "step": 184
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.925986842105265e-06,
+      "loss": 1.5909,
+      "step": 185
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.925438596491229e-06,
+      "loss": 1.6119,
+      "step": 186
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.924890350877195e-06,
+      "loss": 1.5809,
+      "step": 187
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.924342105263158e-06,
+      "loss": 1.6073,
+      "step": 188
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.923793859649124e-06,
+      "loss": 1.5992,
+      "step": 189
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.923245614035088e-06,
+      "loss": 1.6149,
+      "step": 190
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.922697368421054e-06,
+      "loss": 1.5965,
+      "step": 191
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.92214912280702e-06,
+      "loss": 1.5796,
+      "step": 192
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.921600877192983e-06,
+      "loss": 1.5895,
+      "step": 193
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.921052631578947e-06,
+      "loss": 1.6012,
+      "step": 194
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.920504385964913e-06,
+      "loss": 1.5559,
+      "step": 195
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.919956140350879e-06,
+      "loss": 1.6173,
+      "step": 196
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.919407894736843e-06,
+      "loss": 1.6242,
+      "step": 197
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.918859649122807e-06,
+      "loss": 1.6299,
+      "step": 198
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.918311403508772e-06,
+      "loss": 1.5936,
+      "step": 199
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.917763157894738e-06,
+      "loss": 1.5939,
+      "step": 200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.917214912280702e-06,
+      "loss": 1.5866,
+      "step": 201
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.916666666666668e-06,
+      "loss": 1.5676,
+      "step": 202
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.916118421052633e-06,
+      "loss": 1.5596,
+      "step": 203
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.915570175438597e-06,
+      "loss": 1.6123,
+      "step": 204
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.915021929824561e-06,
+      "loss": 1.5995,
+      "step": 205
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.914473684210527e-06,
+      "loss": 1.5952,
+      "step": 206
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.913925438596493e-06,
+      "loss": 1.5186,
+      "step": 207
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.913377192982457e-06,
+      "loss": 1.5879,
+      "step": 208
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.91282894736842e-06,
+      "loss": 1.573,
+      "step": 209
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.912280701754386e-06,
+      "loss": 1.5239,
+      "step": 210
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.911732456140352e-06,
+      "loss": 1.605,
+      "step": 211
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.911184210526316e-06,
+      "loss": 1.5578,
+      "step": 212
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.910635964912282e-06,
+      "loss": 1.5879,
+      "step": 213
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.910087719298247e-06,
+      "loss": 1.5877,
+      "step": 214
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.909539473684211e-06,
+      "loss": 1.6077,
+      "step": 215
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.908991228070175e-06,
+      "loss": 1.5793,
+      "step": 216
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.908442982456141e-06,
+      "loss": 1.6156,
+      "step": 217
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.907894736842107e-06,
+      "loss": 1.593,
+      "step": 218
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.907346491228072e-06,
+      "loss": 1.5699,
+      "step": 219
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.906798245614036e-06,
+      "loss": 1.5523,
+      "step": 220
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.90625e-06,
+      "loss": 1.6115,
+      "step": 221
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.905701754385966e-06,
+      "loss": 1.5944,
+      "step": 222
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.905153508771932e-06,
+      "loss": 1.6021,
+      "step": 223
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.904605263157896e-06,
+      "loss": 1.5622,
+      "step": 224
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.90405701754386e-06,
+      "loss": 1.5955,
+      "step": 225
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.903508771929825e-06,
+      "loss": 1.5477,
+      "step": 226
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.902960526315791e-06,
+      "loss": 1.593,
+      "step": 227
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.902412280701755e-06,
+      "loss": 1.5624,
+      "step": 228
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.90186403508772e-06,
+      "loss": 1.6333,
+      "step": 229
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.901315789473686e-06,
+      "loss": 1.5733,
+      "step": 230
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.90076754385965e-06,
+      "loss": 1.6361,
+      "step": 231
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.900219298245614e-06,
+      "loss": 1.5771,
+      "step": 232
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.89967105263158e-06,
+      "loss": 1.5242,
+      "step": 233
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.899122807017545e-06,
+      "loss": 1.5986,
+      "step": 234
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.89857456140351e-06,
+      "loss": 1.5672,
+      "step": 235
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.898026315789473e-06,
+      "loss": 1.5914,
+      "step": 236
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.897478070175439e-06,
+      "loss": 1.603,
+      "step": 237
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.896929824561405e-06,
+      "loss": 1.5511,
+      "step": 238
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.896381578947369e-06,
+      "loss": 1.6112,
+      "step": 239
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.895833333333334e-06,
+      "loss": 1.5598,
+      "step": 240
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.895285087719298e-06,
+      "loss": 1.568,
+      "step": 241
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.894736842105264e-06,
+      "loss": 1.5751,
+      "step": 242
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.894188596491228e-06,
+      "loss": 1.5652,
+      "step": 243
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.893640350877194e-06,
+      "loss": 1.5987,
+      "step": 244
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.89309210526316e-06,
+      "loss": 1.6153,
+      "step": 245
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.892543859649123e-06,
+      "loss": 1.5603,
+      "step": 246
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.891995614035087e-06,
+      "loss": 1.5149,
+      "step": 247
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.891447368421053e-06,
+      "loss": 1.5917,
+      "step": 248
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.890899122807019e-06,
+      "loss": 1.6398,
+      "step": 249
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.890350877192983e-06,
+      "loss": 1.59,
+      "step": 250
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.889802631578948e-06,
+      "loss": 1.5871,
+      "step": 251
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.889254385964912e-06,
+      "loss": 1.5808,
+      "step": 252
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.888706140350878e-06,
+      "loss": 1.624,
+      "step": 253
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.888157894736844e-06,
+      "loss": 1.5729,
+      "step": 254
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.887609649122808e-06,
+      "loss": 1.6246,
+      "step": 255
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.887061403508773e-06,
+      "loss": 1.5807,
+      "step": 256
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.886513157894739e-06,
+      "loss": 1.5824,
+      "step": 257
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.885964912280703e-06,
+      "loss": 1.4914,
+      "step": 258
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.885416666666667e-06,
+      "loss": 1.5841,
+      "step": 259
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.884868421052633e-06,
+      "loss": 1.619,
+      "step": 260
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.884320175438598e-06,
+      "loss": 1.6144,
+      "step": 261
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.883771929824562e-06,
+      "loss": 1.6616,
+      "step": 262
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.883223684210526e-06,
+      "loss": 1.5969,
+      "step": 263
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.882675438596492e-06,
+      "loss": 1.5666,
+      "step": 264
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.882127192982458e-06,
+      "loss": 1.5852,
+      "step": 265
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.881578947368422e-06,
+      "loss": 1.5602,
+      "step": 266
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.881030701754387e-06,
+      "loss": 1.6016,
+      "step": 267
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.880482456140351e-06,
+      "loss": 1.5877,
+      "step": 268
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.879934210526317e-06,
+      "loss": 1.5396,
+      "step": 269
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.879385964912281e-06,
+      "loss": 1.5916,
+      "step": 270
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.878837719298247e-06,
+      "loss": 1.632,
+      "step": 271
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.878289473684212e-06,
+      "loss": 1.6093,
+      "step": 272
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.877741228070176e-06,
+      "loss": 1.5966,
+      "step": 273
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.87719298245614e-06,
+      "loss": 1.5751,
+      "step": 274
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.876644736842106e-06,
+      "loss": 1.5996,
+      "step": 275
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.876096491228072e-06,
+      "loss": 1.5231,
+      "step": 276
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.875548245614035e-06,
+      "loss": 1.5348,
+      "step": 277
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.875000000000001e-06,
+      "loss": 1.5797,
+      "step": 278
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.874451754385965e-06,
+      "loss": 1.5114,
+      "step": 279
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.87390350877193e-06,
+      "loss": 1.5919,
+      "step": 280
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.873355263157895e-06,
+      "loss": 1.6463,
+      "step": 281
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.87280701754386e-06,
+      "loss": 1.5899,
+      "step": 282
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.872258771929826e-06,
+      "loss": 1.6788,
+      "step": 283
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.87171052631579e-06,
+      "loss": 1.6046,
+      "step": 284
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.871162280701754e-06,
+      "loss": 1.5824,
+      "step": 285
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.87061403508772e-06,
+      "loss": 1.5908,
+      "step": 286
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.870065789473685e-06,
+      "loss": 1.5776,
+      "step": 287
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.86951754385965e-06,
+      "loss": 1.5718,
+      "step": 288
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.868969298245615e-06,
+      "loss": 1.6692,
+      "step": 289
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.868421052631579e-06,
+      "loss": 1.6026,
+      "step": 290
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.867872807017545e-06,
+      "loss": 1.5914,
+      "step": 291
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.86732456140351e-06,
+      "loss": 1.5816,
+      "step": 292
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.866776315789474e-06,
+      "loss": 1.5316,
+      "step": 293
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.86622807017544e-06,
+      "loss": 1.6292,
+      "step": 294
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.865679824561404e-06,
+      "loss": 1.6738,
+      "step": 295
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.86513157894737e-06,
+      "loss": 1.6109,
+      "step": 296
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.864583333333334e-06,
+      "loss": 1.5851,
+      "step": 297
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.8640350877193e-06,
+      "loss": 1.5541,
+      "step": 298
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.863486842105265e-06,
+      "loss": 1.5817,
+      "step": 299
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.862938596491229e-06,
+      "loss": 1.5494,
+      "step": 300
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.862390350877193e-06,
+      "loss": 1.5637,
+      "step": 301
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.861842105263159e-06,
+      "loss": 1.5549,
+      "step": 302
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.861293859649124e-06,
+      "loss": 1.5229,
+      "step": 303
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.860745614035088e-06,
+      "loss": 1.5469,
+      "step": 304
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.860197368421054e-06,
+      "loss": 1.6283,
+      "step": 305
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.859649122807018e-06,
+      "loss": 1.5837,
+      "step": 306
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.859100877192984e-06,
+      "loss": 1.5537,
+      "step": 307
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.858552631578948e-06,
+      "loss": 1.5011,
+      "step": 308
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.858004385964913e-06,
+      "loss": 1.6426,
+      "step": 309
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.857456140350879e-06,
+      "loss": 1.5462,
+      "step": 310
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.856907894736843e-06,
+      "loss": 1.5511,
+      "step": 311
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.856359649122807e-06,
+      "loss": 1.4909,
+      "step": 312
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.855811403508773e-06,
+      "loss": 1.599,
+      "step": 313
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.855263157894738e-06,
+      "loss": 1.4916,
+      "step": 314
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.854714912280702e-06,
+      "loss": 1.5756,
+      "step": 315
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.854166666666668e-06,
+      "loss": 1.6147,
+      "step": 316
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.853618421052632e-06,
+      "loss": 1.529,
+      "step": 317
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.853070175438598e-06,
+      "loss": 1.4614,
+      "step": 318
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.852521929824562e-06,
+      "loss": 1.5494,
+      "step": 319
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.851973684210527e-06,
+      "loss": 1.5083,
+      "step": 320
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.851425438596493e-06,
+      "loss": 1.5172,
+      "step": 321
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.850877192982457e-06,
+      "loss": 1.5229,
+      "step": 322
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.85032894736842e-06,
+      "loss": 1.5596,
+      "step": 323
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.849780701754386e-06,
+      "loss": 1.5504,
+      "step": 324
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.849232456140352e-06,
+      "loss": 1.6142,
+      "step": 325
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.848684210526316e-06,
+      "loss": 1.5691,
+      "step": 326
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.848135964912282e-06,
+      "loss": 1.5577,
+      "step": 327
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.847587719298246e-06,
+      "loss": 1.6538,
+      "step": 328
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.847039473684211e-06,
+      "loss": 1.5549,
+      "step": 329
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.846491228070177e-06,
+      "loss": 1.5759,
+      "step": 330
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.845942982456141e-06,
+      "loss": 1.5356,
+      "step": 331
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.845394736842107e-06,
+      "loss": 1.5075,
+      "step": 332
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.84484649122807e-06,
+      "loss": 1.4751,
+      "step": 333
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.844298245614036e-06,
+      "loss": 1.5318,
+      "step": 334
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.84375e-06,
+      "loss": 1.5324,
+      "step": 335
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.843201754385966e-06,
+      "loss": 1.5586,
+      "step": 336
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.842653508771932e-06,
+      "loss": 1.4586,
+      "step": 337
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.842105263157896e-06,
+      "loss": 1.6205,
+      "step": 338
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.84155701754386e-06,
+      "loss": 1.5067,
+      "step": 339
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.841008771929825e-06,
+      "loss": 1.5263,
+      "step": 340
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.840460526315791e-06,
+      "loss": 1.5298,
+      "step": 341
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.839912280701755e-06,
+      "loss": 1.5432,
+      "step": 342
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.83936403508772e-06,
+      "loss": 1.5567,
+      "step": 343
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.838815789473685e-06,
+      "loss": 1.536,
+      "step": 344
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.83826754385965e-06,
+      "loss": 1.4357,
+      "step": 345
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.837719298245614e-06,
+      "loss": 1.4871,
+      "step": 346
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.83717105263158e-06,
+      "loss": 1.6105,
+      "step": 347
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.836622807017546e-06,
+      "loss": 1.578,
+      "step": 348
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.83607456140351e-06,
+      "loss": 1.6441,
+      "step": 349
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.835526315789474e-06,
+      "loss": 1.5116,
+      "step": 350
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.83497807017544e-06,
+      "loss": 1.6071,
+      "step": 351
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.834429824561405e-06,
+      "loss": 1.4636,
+      "step": 352
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.833881578947369e-06,
+      "loss": 1.563,
+      "step": 353
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.833333333333333e-06,
+      "loss": 1.4124,
+      "step": 354
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.832785087719299e-06,
+      "loss": 1.5552,
+      "step": 355
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.832236842105264e-06,
+      "loss": 1.5073,
+      "step": 356
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.831688596491228e-06,
+      "loss": 1.551,
+      "step": 357
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.831140350877194e-06,
+      "loss": 1.5211,
+      "step": 358
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.83059210526316e-06,
+      "loss": 1.596,
+      "step": 359
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.830043859649124e-06,
+      "loss": 1.6232,
+      "step": 360
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.829495614035088e-06,
+      "loss": 1.4023,
+      "step": 361
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.828947368421053e-06,
+      "loss": 1.4514,
+      "step": 362
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.828399122807019e-06,
+      "loss": 1.545,
+      "step": 363
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.827850877192983e-06,
+      "loss": 1.515,
+      "step": 364
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.827302631578949e-06,
+      "loss": 1.6025,
+      "step": 365
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.826754385964913e-06,
+      "loss": 1.5233,
+      "step": 366
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.826206140350878e-06,
+      "loss": 1.5173,
+      "step": 367
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.825657894736844e-06,
+      "loss": 1.4637,
+      "step": 368
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.825109649122808e-06,
+      "loss": 1.3262,
+      "step": 369
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.824561403508772e-06,
+      "loss": 1.6584,
+      "step": 370
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.824013157894737e-06,
+      "loss": 1.5473,
+      "step": 371
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.823464912280703e-06,
+      "loss": 1.4957,
+      "step": 372
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.822916666666667e-06,
+      "loss": 1.5567,
+      "step": 373
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.822368421052633e-06,
+      "loss": 1.4123,
+      "step": 374
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.821820175438598e-06,
+      "loss": 1.5594,
+      "step": 375
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.821271929824562e-06,
+      "loss": 1.3713,
+      "step": 376
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.820723684210526e-06,
+      "loss": 1.5994,
+      "step": 377
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.820175438596492e-06,
+      "loss": 1.4309,
+      "step": 378
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.819627192982458e-06,
+      "loss": 1.4488,
+      "step": 379
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.819078947368422e-06,
+      "loss": 1.5622,
+      "step": 380
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.818530701754386e-06,
+      "loss": 1.5291,
+      "step": 381
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.817982456140351e-06,
+      "loss": 1.405,
+      "step": 382
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.817434210526317e-06,
+      "loss": 1.5611,
+      "step": 383
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.816885964912281e-06,
+      "loss": 1.4792,
+      "step": 384
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.816337719298247e-06,
+      "loss": 1.3547,
+      "step": 385
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.815789473684212e-06,
+      "loss": 1.5,
+      "step": 386
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.815241228070176e-06,
+      "loss": 1.3078,
+      "step": 387
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.81469298245614e-06,
+      "loss": 1.4779,
+      "step": 388
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.814144736842106e-06,
+      "loss": 1.4842,
+      "step": 389
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.813596491228072e-06,
+      "loss": 1.5058,
+      "step": 390
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.813048245614036e-06,
+      "loss": 1.479,
+      "step": 391
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.8125e-06,
+      "loss": 1.6603,
+      "step": 392
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.811951754385965e-06,
+      "loss": 1.5291,
+      "step": 393
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.811403508771931e-06,
+      "loss": 1.408,
+      "step": 394
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.810855263157895e-06,
+      "loss": 1.4016,
+      "step": 395
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.81030701754386e-06,
+      "loss": 1.334,
+      "step": 396
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.809758771929825e-06,
+      "loss": 1.4967,
+      "step": 397
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.80921052631579e-06,
+      "loss": 1.5721,
+      "step": 398
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.808662280701754e-06,
+      "loss": 1.4429,
+      "step": 399
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.80811403508772e-06,
+      "loss": 1.5245,
+      "step": 400
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.807565789473686e-06,
+      "loss": 1.4883,
+      "step": 401
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.80701754385965e-06,
+      "loss": 1.3374,
+      "step": 402
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.806469298245615e-06,
+      "loss": 1.3775,
+      "step": 403
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.80592105263158e-06,
+      "loss": 1.3767,
+      "step": 404
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.805372807017545e-06,
+      "loss": 1.4754,
+      "step": 405
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.80482456140351e-06,
+      "loss": 1.577,
+      "step": 406
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.804276315789475e-06,
+      "loss": 1.5544,
+      "step": 407
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.803728070175439e-06,
+      "loss": 1.5704,
+      "step": 408
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.803179824561404e-06,
+      "loss": 1.2796,
+      "step": 409
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.80263157894737e-06,
+      "loss": 1.5332,
+      "step": 410
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.802083333333334e-06,
+      "loss": 1.621,
+      "step": 411
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.8015350877193e-06,
+      "loss": 1.6218,
+      "step": 412
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.800986842105264e-06,
+      "loss": 1.4522,
+      "step": 413
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.80043859649123e-06,
+      "loss": 1.3109,
+      "step": 414
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.799890350877193e-06,
+      "loss": 1.3817,
+      "step": 415
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.799342105263159e-06,
+      "loss": 1.6711,
+      "step": 416
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.798793859649125e-06,
+      "loss": 1.5279,
+      "step": 417
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.798245614035088e-06,
+      "loss": 1.4405,
+      "step": 418
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.797697368421052e-06,
+      "loss": 1.5167,
+      "step": 419
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.797149122807018e-06,
+      "loss": 1.4759,
+      "step": 420
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.796600877192984e-06,
+      "loss": 1.5374,
+      "step": 421
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.796052631578948e-06,
+      "loss": 1.6285,
+      "step": 422
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.795504385964913e-06,
+      "loss": 1.4872,
+      "step": 423
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.794956140350877e-06,
+      "loss": 1.3333,
+      "step": 424
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.794407894736843e-06,
+      "loss": 1.5224,
+      "step": 425
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.793859649122807e-06,
+      "loss": 1.4502,
+      "step": 426
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.793311403508773e-06,
+      "loss": 1.5023,
+      "step": 427
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.792763157894738e-06,
+      "loss": 1.6177,
+      "step": 428
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.792214912280702e-06,
+      "loss": 1.4586,
+      "step": 429
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.791666666666666e-06,
+      "loss": 1.5751,
+      "step": 430
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.791118421052632e-06,
+      "loss": 1.3771,
+      "step": 431
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.790570175438598e-06,
+      "loss": 1.4612,
+      "step": 432
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.790021929824562e-06,
+      "loss": 1.4935,
+      "step": 433
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.789473684210527e-06,
+      "loss": 1.3496,
+      "step": 434
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.788925438596491e-06,
+      "loss": 1.3496,
+      "step": 435
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.788377192982457e-06,
+      "loss": 1.502,
+      "step": 436
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.787828947368421e-06,
+      "loss": 1.3634,
+      "step": 437
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.787280701754387e-06,
+      "loss": 1.6359,
+      "step": 438
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.786732456140352e-06,
+      "loss": 1.4702,
+      "step": 439
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.786184210526316e-06,
+      "loss": 1.1785,
+      "step": 440
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.785635964912282e-06,
+      "loss": 1.5294,
+      "step": 441
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.785087719298246e-06,
+      "loss": 1.6261,
+      "step": 442
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.784539473684212e-06,
+      "loss": 1.2982,
+      "step": 443
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.783991228070177e-06,
+      "loss": 1.4772,
+      "step": 444
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.783442982456141e-06,
+      "loss": 1.3814,
+      "step": 445
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.782894736842105e-06,
+      "loss": 1.4567,
+      "step": 446
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.782346491228071e-06,
+      "loss": 1.5755,
+      "step": 447
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.781798245614037e-06,
+      "loss": 1.5008,
+      "step": 448
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.78125e-06,
+      "loss": 1.5462,
+      "step": 449
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.780701754385966e-06,
+      "loss": 1.5404,
+      "step": 450
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.78015350877193e-06,
+      "loss": 1.4801,
+      "step": 451
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.779605263157896e-06,
+      "loss": 1.5141,
+      "step": 452
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.77905701754386e-06,
+      "loss": 1.316,
+      "step": 453
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.778508771929826e-06,
+      "loss": 1.3051,
+      "step": 454
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.777960526315791e-06,
+      "loss": 1.6948,
+      "step": 455
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.777412280701755e-06,
+      "loss": 1.2102,
+      "step": 456
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.77686403508772e-06,
+      "loss": 1.4565,
+      "step": 457
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.776315789473685e-06,
+      "loss": 1.373,
+      "step": 458
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.77576754385965e-06,
+      "loss": 1.3683,
+      "step": 459
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.775219298245615e-06,
+      "loss": 1.2011,
+      "step": 460
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.77467105263158e-06,
+      "loss": 1.357,
+      "step": 461
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.774122807017544e-06,
+      "loss": 1.3649,
+      "step": 462
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.77357456140351e-06,
+      "loss": 1.4238,
+      "step": 463
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.773026315789474e-06,
+      "loss": 1.4322,
+      "step": 464
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.77247807017544e-06,
+      "loss": 1.5225,
+      "step": 465
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.771929824561405e-06,
+      "loss": 1.3478,
+      "step": 466
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.771381578947369e-06,
+      "loss": 1.385,
+      "step": 467
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.770833333333333e-06,
+      "loss": 1.3698,
+      "step": 468
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.770285087719299e-06,
+      "loss": 1.5302,
+      "step": 469
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.769736842105264e-06,
+      "loss": 1.5617,
+      "step": 470
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.769188596491228e-06,
+      "loss": 1.5257,
+      "step": 471
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.768640350877194e-06,
+      "loss": 1.5281,
+      "step": 472
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.768092105263158e-06,
+      "loss": 1.2836,
+      "step": 473
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.767543859649124e-06,
+      "loss": 1.5181,
+      "step": 474
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.766995614035088e-06,
+      "loss": 1.4838,
+      "step": 475
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.766447368421053e-06,
+      "loss": 1.4532,
+      "step": 476
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.765899122807019e-06,
+      "loss": 1.4188,
+      "step": 477
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.765350877192983e-06,
+      "loss": 1.4062,
+      "step": 478
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.764802631578949e-06,
+      "loss": 1.3363,
+      "step": 479
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.764254385964913e-06,
+      "loss": 1.3906,
+      "step": 480
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.763706140350878e-06,
+      "loss": 1.4046,
+      "step": 481
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.763157894736844e-06,
+      "loss": 1.1599,
+      "step": 482
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.762609649122808e-06,
+      "loss": 1.3574,
+      "step": 483
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.762061403508772e-06,
+      "loss": 1.3339,
+      "step": 484
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.761513157894738e-06,
+      "loss": 1.7126,
+      "step": 485
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.760964912280703e-06,
+      "loss": 1.3703,
+      "step": 486
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.760416666666667e-06,
+      "loss": 1.5016,
+      "step": 487
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.759868421052633e-06,
+      "loss": 1.4981,
+      "step": 488
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.759320175438597e-06,
+      "loss": 1.5738,
+      "step": 489
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.758771929824563e-06,
+      "loss": 1.3995,
+      "step": 490
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.758223684210527e-06,
+      "loss": 1.5722,
+      "step": 491
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.757675438596492e-06,
+      "loss": 1.4068,
+      "step": 492
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.757127192982458e-06,
+      "loss": 1.253,
+      "step": 493
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.756578947368422e-06,
+      "loss": 1.6311,
+      "step": 494
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.756030701754386e-06,
+      "loss": 1.47,
+      "step": 495
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.755482456140352e-06,
+      "loss": 1.3125,
+      "step": 496
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.754934210526317e-06,
+      "loss": 1.5251,
+      "step": 497
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.754385964912281e-06,
+      "loss": 1.4275,
+      "step": 498
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.753837719298247e-06,
+      "loss": 1.384,
+      "step": 499
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.753289473684211e-06,
+      "loss": 1.4488,
+      "step": 500
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.752741228070177e-06,
+      "loss": 1.3114,
+      "step": 501
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.75219298245614e-06,
+      "loss": 1.4273,
+      "step": 502
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.751644736842106e-06,
+      "loss": 1.5851,
+      "step": 503
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.751096491228072e-06,
+      "loss": 1.4323,
+      "step": 504
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.750548245614036e-06,
+      "loss": 1.6404,
+      "step": 505
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.75e-06,
+      "loss": 1.5603,
+      "step": 506
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.749451754385965e-06,
+      "loss": 1.6297,
+      "step": 507
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.748903508771931e-06,
+      "loss": 1.2822,
+      "step": 508
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.748355263157895e-06,
+      "loss": 1.2603,
+      "step": 509
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.747807017543859e-06,
+      "loss": 1.5481,
+      "step": 510
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.747258771929825e-06,
+      "loss": 1.3965,
+      "step": 511
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.74671052631579e-06,
+      "loss": 1.2515,
+      "step": 512
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.746162280701756e-06,
+      "loss": 1.4196,
+      "step": 513
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.74561403508772e-06,
+      "loss": 1.5923,
+      "step": 514
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.745065789473686e-06,
+      "loss": 1.4826,
+      "step": 515
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.74451754385965e-06,
+      "loss": 1.4207,
+      "step": 516
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.743969298245615e-06,
+      "loss": 1.3618,
+      "step": 517
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.74342105263158e-06,
+      "loss": 1.5572,
+      "step": 518
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.742872807017545e-06,
+      "loss": 1.4831,
+      "step": 519
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.74232456140351e-06,
+      "loss": 1.3672,
+      "step": 520
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.741776315789475e-06,
+      "loss": 1.5016,
+      "step": 521
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.741228070175439e-06,
+      "loss": 1.7451,
+      "step": 522
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.740679824561404e-06,
+      "loss": 1.1284,
+      "step": 523
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.74013157894737e-06,
+      "loss": 1.381,
+      "step": 524
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.739583333333334e-06,
+      "loss": 1.4554,
+      "step": 525
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.739035087719298e-06,
+      "loss": 1.2923,
+      "step": 526
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.738486842105264e-06,
+      "loss": 1.4403,
+      "step": 527
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.73793859649123e-06,
+      "loss": 1.2883,
+      "step": 528
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.737390350877193e-06,
+      "loss": 1.5477,
+      "step": 529
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.736842105263159e-06,
+      "loss": 1.4205,
+      "step": 530
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.736293859649125e-06,
+      "loss": 1.0822,
+      "step": 531
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.735745614035089e-06,
+      "loss": 1.4949,
+      "step": 532
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.735197368421053e-06,
+      "loss": 1.5024,
+      "step": 533
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.734649122807018e-06,
+      "loss": 1.3362,
+      "step": 534
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.734100877192984e-06,
+      "loss": 1.4924,
+      "step": 535
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.733552631578948e-06,
+      "loss": 1.5822,
+      "step": 536
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.733004385964912e-06,
+      "loss": 1.4553,
+      "step": 537
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.732456140350878e-06,
+      "loss": 1.5296,
+      "step": 538
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.731907894736843e-06,
+      "loss": 1.2877,
+      "step": 539
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.731359649122807e-06,
+      "loss": 1.4462,
+      "step": 540
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.730811403508773e-06,
+      "loss": 1.4504,
+      "step": 541
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.730263157894739e-06,
+      "loss": 1.4341,
+      "step": 542
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.729714912280703e-06,
+      "loss": 1.5238,
+      "step": 543
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.729166666666667e-06,
+      "loss": 1.3775,
+      "step": 544
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.728618421052632e-06,
+      "loss": 1.2012,
+      "step": 545
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.728070175438598e-06,
+      "loss": 1.3915,
+      "step": 546
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.727521929824562e-06,
+      "loss": 1.4723,
+      "step": 547
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.726973684210526e-06,
+      "loss": 1.379,
+      "step": 548
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.726425438596492e-06,
+      "loss": 1.3831,
+      "step": 549
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.725877192982457e-06,
+      "loss": 1.448,
+      "step": 550
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.725328947368423e-06,
+      "loss": 1.1887,
+      "step": 551
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.724780701754387e-06,
+      "loss": 1.2772,
+      "step": 552
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.72423245614035e-06,
+      "loss": 1.4557,
+      "step": 553
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.723684210526316e-06,
+      "loss": 1.5756,
+      "step": 554
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.723135964912282e-06,
+      "loss": 1.3622,
+      "step": 555
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.722587719298246e-06,
+      "loss": 1.1393,
+      "step": 556
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.722039473684212e-06,
+      "loss": 1.3914,
+      "step": 557
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.721491228070177e-06,
+      "loss": 1.6535,
+      "step": 558
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.720942982456141e-06,
+      "loss": 1.3738,
+      "step": 559
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.720394736842105e-06,
+      "loss": 1.606,
+      "step": 560
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.719846491228071e-06,
+      "loss": 1.3295,
+      "step": 561
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.719298245614037e-06,
+      "loss": 1.5963,
+      "step": 562
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.71875e-06,
+      "loss": 1.4318,
+      "step": 563
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.718201754385965e-06,
+      "loss": 1.542,
+      "step": 564
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.71765350877193e-06,
+      "loss": 1.4125,
+      "step": 565
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.717105263157896e-06,
+      "loss": 1.3374,
+      "step": 566
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.71655701754386e-06,
+      "loss": 1.3186,
+      "step": 567
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.716008771929826e-06,
+      "loss": 1.3262,
+      "step": 568
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.71546052631579e-06,
+      "loss": 1.6003,
+      "step": 569
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.714912280701755e-06,
+      "loss": 1.49,
+      "step": 570
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.71436403508772e-06,
+      "loss": 1.3189,
+      "step": 571
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.713815789473685e-06,
+      "loss": 1.4813,
+      "step": 572
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.71326754385965e-06,
+      "loss": 1.3382,
+      "step": 573
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.712719298245615e-06,
+      "loss": 1.5008,
+      "step": 574
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.712171052631579e-06,
+      "loss": 1.5237,
+      "step": 575
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.711622807017544e-06,
+      "loss": 1.5785,
+      "step": 576
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.71107456140351e-06,
+      "loss": 1.2756,
+      "step": 577
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.710526315789474e-06,
+      "loss": 1.5562,
+      "step": 578
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.70997807017544e-06,
+      "loss": 1.183,
+      "step": 579
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.709429824561404e-06,
+      "loss": 1.475,
+      "step": 580
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.70888157894737e-06,
+      "loss": 1.4098,
+      "step": 581
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.708333333333333e-06,
+      "loss": 1.6858,
+      "step": 582
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.707785087719299e-06,
+      "loss": 1.592,
+      "step": 583
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.707236842105265e-06,
+      "loss": 1.5074,
+      "step": 584
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.706688596491229e-06,
+      "loss": 1.3963,
+      "step": 585
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.706140350877193e-06,
+      "loss": 1.2549,
+      "step": 586
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.705592105263158e-06,
+      "loss": 1.4393,
+      "step": 587
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.705043859649124e-06,
+      "loss": 1.345,
+      "step": 588
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.70449561403509e-06,
+      "loss": 1.4728,
+      "step": 589
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.703947368421054e-06,
+      "loss": 1.371,
+      "step": 590
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.703399122807018e-06,
+      "loss": 1.5284,
+      "step": 591
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.702850877192983e-06,
+      "loss": 1.4141,
+      "step": 592
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.702302631578949e-06,
+      "loss": 1.4495,
+      "step": 593
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.701754385964913e-06,
+      "loss": 1.2138,
+      "step": 594
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.701206140350879e-06,
+      "loss": 1.6397,
+      "step": 595
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.700657894736843e-06,
+      "loss": 1.2456,
+      "step": 596
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.700109649122808e-06,
+      "loss": 1.2655,
+      "step": 597
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.699561403508772e-06,
+      "loss": 1.3153,
+      "step": 598
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.699013157894738e-06,
+      "loss": 0.9932,
+      "step": 599
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.698464912280704e-06,
+      "loss": 1.1332,
+      "step": 600
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.697916666666667e-06,
+      "loss": 1.4094,
+      "step": 601
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.697368421052631e-06,
+      "loss": 1.5591,
+      "step": 602
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.696820175438597e-06,
+      "loss": 1.677,
+      "step": 603
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.696271929824563e-06,
+      "loss": 1.3945,
+      "step": 604
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.695723684210527e-06,
+      "loss": 1.3573,
+      "step": 605
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.695175438596492e-06,
+      "loss": 1.2397,
+      "step": 606
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.694627192982456e-06,
+      "loss": 1.0611,
+      "step": 607
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.694078947368422e-06,
+      "loss": 1.3981,
+      "step": 608
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.693530701754386e-06,
+      "loss": 1.3096,
+      "step": 609
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.692982456140352e-06,
+      "loss": 1.1827,
+      "step": 610
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.692434210526317e-06,
+      "loss": 1.4968,
+      "step": 611
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.691885964912281e-06,
+      "loss": 1.4429,
+      "step": 612
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.691337719298245e-06,
+      "loss": 1.4739,
+      "step": 613
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.690789473684211e-06,
+      "loss": 1.4338,
+      "step": 614
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.690241228070177e-06,
+      "loss": 1.3714,
+      "step": 615
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.68969298245614e-06,
+      "loss": 1.3216,
+      "step": 616
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.689144736842106e-06,
+      "loss": 1.3759,
+      "step": 617
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.68859649122807e-06,
+      "loss": 1.1748,
+      "step": 618
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.688048245614036e-06,
+      "loss": 1.5408,
+      "step": 619
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.6875e-06,
+      "loss": 1.1484,
+      "step": 620
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.686951754385966e-06,
+      "loss": 1.2402,
+      "step": 621
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.686403508771931e-06,
+      "loss": 1.228,
+      "step": 622
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.685855263157895e-06,
+      "loss": 1.6224,
+      "step": 623
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.685307017543861e-06,
+      "loss": 1.286,
+      "step": 624
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.684758771929825e-06,
+      "loss": 1.5365,
+      "step": 625
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.68421052631579e-06,
+      "loss": 1.3259,
+      "step": 626
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.683662280701756e-06,
+      "loss": 1.5149,
+      "step": 627
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.68311403508772e-06,
+      "loss": 1.3851,
+      "step": 628
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.682565789473684e-06,
+      "loss": 1.6213,
+      "step": 629
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.68201754385965e-06,
+      "loss": 1.34,
+      "step": 630
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.681469298245616e-06,
+      "loss": 1.4035,
+      "step": 631
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.68092105263158e-06,
+      "loss": 1.3066,
+      "step": 632
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.680372807017545e-06,
+      "loss": 1.2879,
+      "step": 633
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.67982456140351e-06,
+      "loss": 1.3165,
+      "step": 634
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.679276315789475e-06,
+      "loss": 1.5089,
+      "step": 635
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.678728070175439e-06,
+      "loss": 1.4722,
+      "step": 636
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.678179824561405e-06,
+      "loss": 1.5241,
+      "step": 637
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.67763157894737e-06,
+      "loss": 1.217,
+      "step": 638
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.677083333333334e-06,
+      "loss": 1.4934,
+      "step": 639
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.676535087719298e-06,
+      "loss": 1.2004,
+      "step": 640
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.675986842105264e-06,
+      "loss": 1.4093,
+      "step": 641
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.67543859649123e-06,
+      "loss": 1.3824,
+      "step": 642
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.674890350877194e-06,
+      "loss": 1.4095,
+      "step": 643
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.67434210526316e-06,
+      "loss": 1.1532,
+      "step": 644
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.673793859649123e-06,
+      "loss": 1.5697,
+      "step": 645
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.673245614035089e-06,
+      "loss": 1.3113,
+      "step": 646
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.672697368421053e-06,
+      "loss": 1.495,
+      "step": 647
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.672149122807018e-06,
+      "loss": 1.2961,
+      "step": 648
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.671600877192984e-06,
+      "loss": 1.5511,
+      "step": 649
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.671052631578948e-06,
+      "loss": 1.5397,
+      "step": 650
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.670504385964912e-06,
+      "loss": 1.4059,
+      "step": 651
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.669956140350878e-06,
+      "loss": 1.3017,
+      "step": 652
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.669407894736843e-06,
+      "loss": 1.4164,
+      "step": 653
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.668859649122807e-06,
+      "loss": 1.4454,
+      "step": 654
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.668311403508771e-06,
+      "loss": 1.3519,
+      "step": 655
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.667763157894737e-06,
+      "loss": 1.7366,
+      "step": 656
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.667214912280703e-06,
+      "loss": 1.0731,
+      "step": 657
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.666666666666667e-06,
+      "loss": 1.4981,
+      "step": 658
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.666118421052632e-06,
+      "loss": 1.3079,
+      "step": 659
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.665570175438598e-06,
+      "loss": 1.507,
+      "step": 660
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.665021929824562e-06,
+      "loss": 1.5545,
+      "step": 661
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.664473684210528e-06,
+      "loss": 1.465,
+      "step": 662
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.663925438596492e-06,
+      "loss": 1.4141,
+      "step": 663
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.663377192982457e-06,
+      "loss": 1.7057,
+      "step": 664
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.662828947368423e-06,
+      "loss": 1.5708,
+      "step": 665
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.662280701754387e-06,
+      "loss": 1.1865,
+      "step": 666
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.661732456140351e-06,
+      "loss": 1.2074,
+      "step": 667
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.661184210526317e-06,
+      "loss": 1.415,
+      "step": 668
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.660635964912282e-06,
+      "loss": 1.432,
+      "step": 669
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.660087719298246e-06,
+      "loss": 1.2124,
+      "step": 670
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.659539473684212e-06,
+      "loss": 1.5203,
+      "step": 671
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.658991228070176e-06,
+      "loss": 1.5524,
+      "step": 672
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.658442982456142e-06,
+      "loss": 1.2892,
+      "step": 673
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.657894736842106e-06,
+      "loss": 1.449,
+      "step": 674
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.657346491228071e-06,
+      "loss": 1.5454,
+      "step": 675
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.656798245614037e-06,
+      "loss": 1.3266,
+      "step": 676
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.656250000000001e-06,
+      "loss": 1.583,
+      "step": 677
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.655701754385965e-06,
+      "loss": 1.3023,
+      "step": 678
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.65515350877193e-06,
+      "loss": 1.3627,
+      "step": 679
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.654605263157896e-06,
+      "loss": 1.7626,
+      "step": 680
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.65405701754386e-06,
+      "loss": 1.2912,
+      "step": 681
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.653508771929824e-06,
+      "loss": 1.1882,
+      "step": 682
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.65296052631579e-06,
+      "loss": 1.2706,
+      "step": 683
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.652412280701756e-06,
+      "loss": 1.2711,
+      "step": 684
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.65186403508772e-06,
+      "loss": 1.4684,
+      "step": 685
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.651315789473685e-06,
+      "loss": 1.2182,
+      "step": 686
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.650767543859651e-06,
+      "loss": 1.2898,
+      "step": 687
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.650219298245615e-06,
+      "loss": 1.4698,
+      "step": 688
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.649671052631579e-06,
+      "loss": 1.5078,
+      "step": 689
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.649122807017545e-06,
+      "loss": 1.2905,
+      "step": 690
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.64857456140351e-06,
+      "loss": 1.3313,
+      "step": 691
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.648026315789474e-06,
+      "loss": 1.6415,
+      "step": 692
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.647478070175438e-06,
+      "loss": 1.2708,
+      "step": 693
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.646929824561404e-06,
+      "loss": 1.2519,
+      "step": 694
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.64638157894737e-06,
+      "loss": 1.4228,
+      "step": 695
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.645833333333333e-06,
+      "loss": 1.4238,
+      "step": 696
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.645285087719299e-06,
+      "loss": 1.2829,
+      "step": 697
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.644736842105263e-06,
+      "loss": 1.4791,
+      "step": 698
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.644188596491229e-06,
+      "loss": 1.3054,
+      "step": 699
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.643640350877194e-06,
+      "loss": 1.2393,
+      "step": 700
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.643092105263158e-06,
+      "loss": 1.4569,
+      "step": 701
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.642543859649124e-06,
+      "loss": 1.5314,
+      "step": 702
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.64199561403509e-06,
+      "loss": 1.4429,
+      "step": 703
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.641447368421054e-06,
+      "loss": 1.4001,
+      "step": 704
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.640899122807018e-06,
+      "loss": 1.4711,
+      "step": 705
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.640350877192983e-06,
+      "loss": 1.5105,
+      "step": 706
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.639802631578949e-06,
+      "loss": 1.5712,
+      "step": 707
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.639254385964913e-06,
+      "loss": 1.1879,
+      "step": 708
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.638706140350877e-06,
+      "loss": 1.4232,
+      "step": 709
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.638157894736843e-06,
+      "loss": 1.3908,
+      "step": 710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.637609649122808e-06,
+      "loss": 1.3746,
+      "step": 711
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.637061403508772e-06,
+      "loss": 1.3365,
+      "step": 712
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.636513157894738e-06,
+      "loss": 1.332,
+      "step": 713
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.635964912280704e-06,
+      "loss": 1.3599,
+      "step": 714
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.635416666666668e-06,
+      "loss": 1.3906,
+      "step": 715
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.634868421052632e-06,
+      "loss": 1.347,
+      "step": 716
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.634320175438597e-06,
+      "loss": 1.6044,
+      "step": 717
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.633771929824563e-06,
+      "loss": 1.43,
+      "step": 718
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.633223684210527e-06,
+      "loss": 1.6385,
+      "step": 719
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.632675438596491e-06,
+      "loss": 1.2482,
+      "step": 720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.632127192982457e-06,
+      "loss": 1.5222,
+      "step": 721
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.631578947368422e-06,
+      "loss": 1.5039,
+      "step": 722
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.631030701754386e-06,
+      "loss": 1.3621,
+      "step": 723
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.630482456140352e-06,
+      "loss": 1.0233,
+      "step": 724
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.629934210526316e-06,
+      "loss": 1.5586,
+      "step": 725
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.629385964912282e-06,
+      "loss": 1.3294,
+      "step": 726
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.628837719298246e-06,
+      "loss": 1.312,
+      "step": 727
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.628289473684211e-06,
+      "loss": 1.8133,
+      "step": 728
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.627741228070177e-06,
+      "loss": 1.4477,
+      "step": 729
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.627192982456141e-06,
+      "loss": 1.2599,
+      "step": 730
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.626644736842105e-06,
+      "loss": 1.4847,
+      "step": 731
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.62609649122807e-06,
+      "loss": 1.3055,
+      "step": 732
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.625548245614036e-06,
+      "loss": 1.4451,
+      "step": 733
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.625e-06,
+      "loss": 1.3406,
+      "step": 734
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.624451754385966e-06,
+      "loss": 1.4059,
+      "step": 735
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.62390350877193e-06,
+      "loss": 1.4206,
+      "step": 736
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.623355263157895e-06,
+      "loss": 1.2521,
+      "step": 737
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.622807017543861e-06,
+      "loss": 1.1541,
+      "step": 738
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.622258771929825e-06,
+      "loss": 1.4745,
+      "step": 739
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.62171052631579e-06,
+      "loss": 1.2929,
+      "step": 740
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.621162280701755e-06,
+      "loss": 1.4212,
+      "step": 741
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.62061403508772e-06,
+      "loss": 1.2687,
+      "step": 742
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.620065789473684e-06,
+      "loss": 1.4033,
+      "step": 743
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.61951754385965e-06,
+      "loss": 1.3846,
+      "step": 744
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.618969298245616e-06,
+      "loss": 1.1781,
+      "step": 745
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.61842105263158e-06,
+      "loss": 1.2305,
+      "step": 746
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.617872807017544e-06,
+      "loss": 1.4191,
+      "step": 747
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.61732456140351e-06,
+      "loss": 1.2596,
+      "step": 748
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.616776315789475e-06,
+      "loss": 1.2573,
+      "step": 749
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.616228070175439e-06,
+      "loss": 1.1952,
+      "step": 750
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.615679824561405e-06,
+      "loss": 1.5325,
+      "step": 751
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.615131578947369e-06,
+      "loss": 1.5547,
+      "step": 752
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.614583333333334e-06,
+      "loss": 1.3568,
+      "step": 753
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.614035087719298e-06,
+      "loss": 1.4996,
+      "step": 754
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.613486842105264e-06,
+      "loss": 0.9884,
+      "step": 755
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.61293859649123e-06,
+      "loss": 1.4376,
+      "step": 756
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.612390350877194e-06,
+      "loss": 1.3306,
+      "step": 757
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.611842105263158e-06,
+      "loss": 1.4464,
+      "step": 758
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.611293859649123e-06,
+      "loss": 1.4811,
+      "step": 759
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.610745614035089e-06,
+      "loss": 1.1567,
+      "step": 760
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.610197368421053e-06,
+      "loss": 1.1047,
+      "step": 761
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.609649122807019e-06,
+      "loss": 1.544,
+      "step": 762
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.609100877192983e-06,
+      "loss": 1.4364,
+      "step": 763
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.608552631578948e-06,
+      "loss": 1.4023,
+      "step": 764
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.608004385964912e-06,
+      "loss": 1.1875,
+      "step": 765
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.607456140350878e-06,
+      "loss": 1.3958,
+      "step": 766
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.606907894736844e-06,
+      "loss": 1.4744,
+      "step": 767
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.606359649122808e-06,
+      "loss": 1.2566,
+      "step": 768
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.605811403508772e-06,
+      "loss": 1.1136,
+      "step": 769
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.605263157894737e-06,
+      "loss": 1.3993,
+      "step": 770
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.604714912280703e-06,
+      "loss": 1.46,
+      "step": 771
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.604166666666669e-06,
+      "loss": 1.1783,
+      "step": 772
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.603618421052633e-06,
+      "loss": 1.2221,
+      "step": 773
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.603070175438597e-06,
+      "loss": 1.0447,
+      "step": 774
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.602521929824562e-06,
+      "loss": 0.8983,
+      "step": 775
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.601973684210528e-06,
+      "loss": 1.6789,
+      "step": 776
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.601425438596492e-06,
+      "loss": 1.3283,
+      "step": 777
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.600877192982458e-06,
+      "loss": 1.2531,
+      "step": 778
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.600328947368422e-06,
+      "loss": 1.3151,
+      "step": 779
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.599780701754387e-06,
+      "loss": 1.6015,
+      "step": 780
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.599232456140351e-06,
+      "loss": 1.0854,
+      "step": 781
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.598684210526317e-06,
+      "loss": 0.9978,
+      "step": 782
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.598135964912283e-06,
+      "loss": 1.1765,
+      "step": 783
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.597587719298246e-06,
+      "loss": 1.1884,
+      "step": 784
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.59703947368421e-06,
+      "loss": 1.287,
+      "step": 785
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.596491228070176e-06,
+      "loss": 1.4139,
+      "step": 786
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.595942982456142e-06,
+      "loss": 1.3023,
+      "step": 787
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.595394736842106e-06,
+      "loss": 1.3849,
+      "step": 788
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.594846491228071e-06,
+      "loss": 0.9455,
+      "step": 789
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.594298245614035e-06,
+      "loss": 1.2492,
+      "step": 790
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.593750000000001e-06,
+      "loss": 1.0087,
+      "step": 791
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.593201754385965e-06,
+      "loss": 1.4873,
+      "step": 792
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.59265350877193e-06,
+      "loss": 1.3162,
+      "step": 793
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.592105263157896e-06,
+      "loss": 1.3656,
+      "step": 794
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.59155701754386e-06,
+      "loss": 1.5118,
+      "step": 795
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.591008771929824e-06,
+      "loss": 1.5127,
+      "step": 796
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.59046052631579e-06,
+      "loss": 1.3377,
+      "step": 797
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.589912280701756e-06,
+      "loss": 1.2662,
+      "step": 798
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.58936403508772e-06,
+      "loss": 1.4259,
+      "step": 799
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.588815789473685e-06,
+      "loss": 1.4435,
+      "step": 800
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.58826754385965e-06,
+      "loss": 1.4352,
+      "step": 801
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.587719298245615e-06,
+      "loss": 0.8642,
+      "step": 802
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.587171052631579e-06,
+      "loss": 1.2873,
+      "step": 803
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.586622807017545e-06,
+      "loss": 0.9103,
+      "step": 804
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.58607456140351e-06,
+      "loss": 1.6016,
+      "step": 805
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.585526315789474e-06,
+      "loss": 1.1292,
+      "step": 806
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.584978070175438e-06,
+      "loss": 1.4338,
+      "step": 807
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.584429824561404e-06,
+      "loss": 1.4652,
+      "step": 808
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.58388157894737e-06,
+      "loss": 1.1851,
+      "step": 809
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.583333333333335e-06,
+      "loss": 1.3938,
+      "step": 810
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.5827850877193e-06,
+      "loss": 1.3557,
+      "step": 811
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.582236842105263e-06,
+      "loss": 1.0327,
+      "step": 812
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.581688596491229e-06,
+      "loss": 1.3039,
+      "step": 813
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.581140350877195e-06,
+      "loss": 1.122,
+      "step": 814
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.580592105263159e-06,
+      "loss": 1.1399,
+      "step": 815
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.580043859649124e-06,
+      "loss": 1.3883,
+      "step": 816
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.579495614035088e-06,
+      "loss": 1.4518,
+      "step": 817
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.578947368421054e-06,
+      "loss": 1.702,
+      "step": 818
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.578399122807018e-06,
+      "loss": 1.4784,
+      "step": 819
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.577850877192984e-06,
+      "loss": 1.1803,
+      "step": 820
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.57730263157895e-06,
+      "loss": 1.2715,
+      "step": 821
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.576754385964913e-06,
+      "loss": 1.0676,
+      "step": 822
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.576206140350877e-06,
+      "loss": 1.1541,
+      "step": 823
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.575657894736843e-06,
+      "loss": 1.427,
+      "step": 824
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.575109649122809e-06,
+      "loss": 1.1821,
+      "step": 825
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.574561403508773e-06,
+      "loss": 1.2425,
+      "step": 826
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.574013157894738e-06,
+      "loss": 1.2502,
+      "step": 827
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.573464912280702e-06,
+      "loss": 1.0769,
+      "step": 828
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.572916666666668e-06,
+      "loss": 1.4767,
+      "step": 829
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.572368421052632e-06,
+      "loss": 1.0642,
+      "step": 830
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.571820175438597e-06,
+      "loss": 1.5595,
+      "step": 831
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.571271929824563e-06,
+      "loss": 1.28,
+      "step": 832
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.570723684210527e-06,
+      "loss": 1.8001,
+      "step": 833
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.570175438596491e-06,
+      "loss": 1.2081,
+      "step": 834
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.569627192982457e-06,
+      "loss": 1.3587,
+      "step": 835
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.569078947368422e-06,
+      "loss": 1.521,
+      "step": 836
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.568530701754386e-06,
+      "loss": 1.2132,
+      "step": 837
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.56798245614035e-06,
+      "loss": 1.1024,
+      "step": 838
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.567434210526316e-06,
+      "loss": 1.1069,
+      "step": 839
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.566885964912282e-06,
+      "loss": 1.2447,
+      "step": 840
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.566337719298246e-06,
+      "loss": 1.2609,
+      "step": 841
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.565789473684211e-06,
+      "loss": 1.1249,
+      "step": 842
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.565241228070177e-06,
+      "loss": 1.2653,
+      "step": 843
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.564692982456141e-06,
+      "loss": 1.1234,
+      "step": 844
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.564144736842105e-06,
+      "loss": 1.0573,
+      "step": 845
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.56359649122807e-06,
+      "loss": 1.4703,
+      "step": 846
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.563048245614036e-06,
+      "loss": 1.2867,
+      "step": 847
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.562500000000002e-06,
+      "loss": 1.6266,
+      "step": 848
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.561951754385966e-06,
+      "loss": 1.8115,
+      "step": 849
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.56140350877193e-06,
+      "loss": 1.0887,
+      "step": 850
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.560855263157896e-06,
+      "loss": 2.0986,
+      "step": 851
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.560307017543861e-06,
+      "loss": 1.3037,
+      "step": 852
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.559758771929825e-06,
+      "loss": 1.0526,
+      "step": 853
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.55921052631579e-06,
+      "loss": 1.5428,
+      "step": 854
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.558662280701755e-06,
+      "loss": 1.209,
+      "step": 855
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.55811403508772e-06,
+      "loss": 1.326,
+      "step": 856
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.557565789473685e-06,
+      "loss": 1.105,
+      "step": 857
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.55701754385965e-06,
+      "loss": 1.2755,
+      "step": 858
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.556469298245616e-06,
+      "loss": 1.3139,
+      "step": 859
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.55592105263158e-06,
+      "loss": 1.3869,
+      "step": 860
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.555372807017544e-06,
+      "loss": 1.4445,
+      "step": 861
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.55482456140351e-06,
+      "loss": 1.455,
+      "step": 862
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.554276315789475e-06,
+      "loss": 1.1506,
+      "step": 863
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.55372807017544e-06,
+      "loss": 1.4058,
+      "step": 864
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.553179824561403e-06,
+      "loss": 1.0908,
+      "step": 865
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.552631578947369e-06,
+      "loss": 1.1565,
+      "step": 866
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.552083333333335e-06,
+      "loss": 1.6048,
+      "step": 867
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.551535087719299e-06,
+      "loss": 1.4852,
+      "step": 868
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.550986842105264e-06,
+      "loss": 1.2066,
+      "step": 869
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.55043859649123e-06,
+      "loss": 1.506,
+      "step": 870
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.549890350877194e-06,
+      "loss": 1.6714,
+      "step": 871
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.549342105263158e-06,
+      "loss": 1.2948,
+      "step": 872
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.548793859649124e-06,
+      "loss": 1.1778,
+      "step": 873
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.54824561403509e-06,
+      "loss": 1.4338,
+      "step": 874
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.547697368421053e-06,
+      "loss": 1.4389,
+      "step": 875
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.547149122807017e-06,
+      "loss": 1.4787,
+      "step": 876
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.546600877192983e-06,
+      "loss": 1.2358,
+      "step": 877
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.546052631578948e-06,
+      "loss": 1.2375,
+      "step": 878
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.545504385964912e-06,
+      "loss": 1.2057,
+      "step": 879
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.544956140350878e-06,
+      "loss": 1.0663,
+      "step": 880
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.544407894736842e-06,
+      "loss": 1.4306,
+      "step": 881
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.543859649122808e-06,
+      "loss": 1.4681,
+      "step": 882
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.543311403508773e-06,
+      "loss": 1.0961,
+      "step": 883
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.542763157894737e-06,
+      "loss": 1.228,
+      "step": 884
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.542214912280703e-06,
+      "loss": 1.3536,
+      "step": 885
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.541666666666669e-06,
+      "loss": 1.1626,
+      "step": 886
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.541118421052633e-06,
+      "loss": 1.4037,
+      "step": 887
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.540570175438597e-06,
+      "loss": 1.2844,
+      "step": 888
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.540021929824562e-06,
+      "loss": 1.3613,
+      "step": 889
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.539473684210528e-06,
+      "loss": 1.329,
+      "step": 890
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.538925438596492e-06,
+      "loss": 1.6204,
+      "step": 891
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.538377192982456e-06,
+      "loss": 1.4036,
+      "step": 892
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.537828947368422e-06,
+      "loss": 1.3056,
+      "step": 893
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.537280701754387e-06,
+      "loss": 1.4053,
+      "step": 894
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.536732456140351e-06,
+      "loss": 1.1841,
+      "step": 895
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.536184210526317e-06,
+      "loss": 1.227,
+      "step": 896
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.535635964912281e-06,
+      "loss": 1.1819,
+      "step": 897
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.535087719298247e-06,
+      "loss": 1.2576,
+      "step": 898
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.53453947368421e-06,
+      "loss": 1.4012,
+      "step": 899
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.533991228070176e-06,
+      "loss": 1.5662,
+      "step": 900
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.533442982456142e-06,
+      "loss": 1.1016,
+      "step": 901
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.532894736842106e-06,
+      "loss": 1.028,
+      "step": 902
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.53234649122807e-06,
+      "loss": 1.3574,
+      "step": 903
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.531798245614036e-06,
+      "loss": 1.2259,
+      "step": 904
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.531250000000001e-06,
+      "loss": 1.2077,
+      "step": 905
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.530701754385965e-06,
+      "loss": 1.3781,
+      "step": 906
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.530153508771931e-06,
+      "loss": 0.9238,
+      "step": 907
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.529605263157895e-06,
+      "loss": 1.3721,
+      "step": 908
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.52905701754386e-06,
+      "loss": 1.4381,
+      "step": 909
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.528508771929825e-06,
+      "loss": 1.2853,
+      "step": 910
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.52796052631579e-06,
+      "loss": 0.9883,
+      "step": 911
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.527412280701756e-06,
+      "loss": 1.2537,
+      "step": 912
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.52686403508772e-06,
+      "loss": 1.4594,
+      "step": 913
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.526315789473684e-06,
+      "loss": 1.1862,
+      "step": 914
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.52576754385965e-06,
+      "loss": 1.4263,
+      "step": 915
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.525219298245615e-06,
+      "loss": 0.9834,
+      "step": 916
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.52467105263158e-06,
+      "loss": 1.1458,
+      "step": 917
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.524122807017545e-06,
+      "loss": 1.3862,
+      "step": 918
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.523574561403509e-06,
+      "loss": 1.2631,
+      "step": 919
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.523026315789474e-06,
+      "loss": 1.4251,
+      "step": 920
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.52247807017544e-06,
+      "loss": 1.642,
+      "step": 921
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.521929824561404e-06,
+      "loss": 1.3956,
+      "step": 922
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.52138157894737e-06,
+      "loss": 1.5426,
+      "step": 923
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.520833333333334e-06,
+      "loss": 1.2544,
+      "step": 924
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.5202850877193e-06,
+      "loss": 1.311,
+      "step": 925
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.519736842105263e-06,
+      "loss": 1.2479,
+      "step": 926
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.519188596491229e-06,
+      "loss": 1.4056,
+      "step": 927
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.518640350877195e-06,
+      "loss": 1.2737,
+      "step": 928
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.518092105263159e-06,
+      "loss": 0.9877,
+      "step": 929
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.517543859649123e-06,
+      "loss": 1.6645,
+      "step": 930
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.516995614035088e-06,
+      "loss": 1.6771,
+      "step": 931
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.516447368421054e-06,
+      "loss": 1.0832,
+      "step": 932
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.515899122807018e-06,
+      "loss": 1.3777,
+      "step": 933
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.515350877192984e-06,
+      "loss": 1.2021,
+      "step": 934
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.514802631578948e-06,
+      "loss": 1.2174,
+      "step": 935
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.514254385964913e-06,
+      "loss": 1.372,
+      "step": 936
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.513706140350877e-06,
+      "loss": 1.3445,
+      "step": 937
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.513157894736843e-06,
+      "loss": 1.1818,
+      "step": 938
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.512609649122809e-06,
+      "loss": 1.5237,
+      "step": 939
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.512061403508773e-06,
+      "loss": 1.2687,
+      "step": 940
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.511513157894737e-06,
+      "loss": 1.5573,
+      "step": 941
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.510964912280702e-06,
+      "loss": 1.1572,
+      "step": 942
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.510416666666668e-06,
+      "loss": 1.3226,
+      "step": 943
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.509868421052632e-06,
+      "loss": 1.1578,
+      "step": 944
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.509320175438598e-06,
+      "loss": 1.225,
+      "step": 945
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.508771929824562e-06,
+      "loss": 1.3393,
+      "step": 946
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.508223684210527e-06,
+      "loss": 1.4376,
+      "step": 947
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.507675438596491e-06,
+      "loss": 1.2321,
+      "step": 948
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.507127192982457e-06,
+      "loss": 1.4334,
+      "step": 949
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.506578947368423e-06,
+      "loss": 1.3327,
+      "step": 950
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.506030701754387e-06,
+      "loss": 1.1657,
+      "step": 951
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.50548245614035e-06,
+      "loss": 1.2792,
+      "step": 952
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.504934210526316e-06,
+      "loss": 1.3712,
+      "step": 953
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.504385964912282e-06,
+      "loss": 1.5853,
+      "step": 954
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.503837719298246e-06,
+      "loss": 0.9404,
+      "step": 955
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.503289473684212e-06,
+      "loss": 1.4239,
+      "step": 956
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.502741228070176e-06,
+      "loss": 1.4563,
+      "step": 957
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.502192982456141e-06,
+      "loss": 1.1962,
+      "step": 958
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.501644736842107e-06,
+      "loss": 1.1443,
+      "step": 959
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.501096491228071e-06,
+      "loss": 1.3478,
+      "step": 960
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.500548245614037e-06,
+      "loss": 1.6141,
+      "step": 961
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.5e-06,
+      "loss": 1.1788,
+      "step": 962
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.499451754385966e-06,
+      "loss": 1.3632,
+      "step": 963
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49890350877193e-06,
+      "loss": 1.2453,
+      "step": 964
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.498355263157896e-06,
+      "loss": 1.0943,
+      "step": 965
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.497807017543862e-06,
+      "loss": 1.4352,
+      "step": 966
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.497258771929825e-06,
+      "loss": 1.3513,
+      "step": 967
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49671052631579e-06,
+      "loss": 1.0286,
+      "step": 968
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.496162280701755e-06,
+      "loss": 1.2683,
+      "step": 969
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49561403508772e-06,
+      "loss": 1.3916,
+      "step": 970
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.495065789473685e-06,
+      "loss": 1.2464,
+      "step": 971
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49451754385965e-06,
+      "loss": 1.4436,
+      "step": 972
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.493969298245614e-06,
+      "loss": 1.2135,
+      "step": 973
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49342105263158e-06,
+      "loss": 1.3874,
+      "step": 974
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.492872807017544e-06,
+      "loss": 0.8462,
+      "step": 975
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49232456140351e-06,
+      "loss": 1.0459,
+      "step": 976
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.491776315789475e-06,
+      "loss": 1.3269,
+      "step": 977
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49122807017544e-06,
+      "loss": 1.4514,
+      "step": 978
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.490679824561403e-06,
+      "loss": 0.9894,
+      "step": 979
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.490131578947369e-06,
+      "loss": 0.9301,
+      "step": 980
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.489583333333335e-06,
+      "loss": 1.2316,
+      "step": 981
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.489035087719299e-06,
+      "loss": 1.0792,
+      "step": 982
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.488486842105263e-06,
+      "loss": 1.2167,
+      "step": 983
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.487938596491228e-06,
+      "loss": 1.3855,
+      "step": 984
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.487390350877194e-06,
+      "loss": 1.3536,
+      "step": 985
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.486842105263158e-06,
+      "loss": 1.4812,
+      "step": 986
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.486293859649124e-06,
+      "loss": 1.4868,
+      "step": 987
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.48574561403509e-06,
+      "loss": 1.293,
+      "step": 988
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.485197368421053e-06,
+      "loss": 1.2855,
+      "step": 989
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.484649122807017e-06,
+      "loss": 1.4199,
+      "step": 990
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.484100877192983e-06,
+      "loss": 1.3903,
+      "step": 991
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.483552631578949e-06,
+      "loss": 1.1865,
+      "step": 992
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.483004385964913e-06,
+      "loss": 1.3868,
+      "step": 993
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.482456140350878e-06,
+      "loss": 1.1726,
+      "step": 994
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.481907894736842e-06,
+      "loss": 1.7414,
+      "step": 995
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.481359649122808e-06,
+      "loss": 1.2585,
+      "step": 996
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.480811403508774e-06,
+      "loss": 0.9731,
+      "step": 997
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.480263157894738e-06,
+      "loss": 1.5375,
+      "step": 998
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.479714912280703e-06,
+      "loss": 1.0876,
+      "step": 999
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.479166666666667e-06,
+      "loss": 1.4471,
+      "step": 1000
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.478618421052633e-06,
+      "loss": 1.0642,
+      "step": 1001
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.478070175438597e-06,
+      "loss": 1.5109,
+      "step": 1002
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.477521929824563e-06,
+      "loss": 1.423,
+      "step": 1003
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.476973684210528e-06,
+      "loss": 0.8495,
+      "step": 1004
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.476425438596492e-06,
+      "loss": 0.9094,
+      "step": 1005
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.475877192982456e-06,
+      "loss": 1.1805,
+      "step": 1006
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.475328947368422e-06,
+      "loss": 1.3601,
+      "step": 1007
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.474780701754388e-06,
+      "loss": 1.3316,
+      "step": 1008
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.474232456140352e-06,
+      "loss": 1.2604,
+      "step": 1009
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.473684210526315e-06,
+      "loss": 1.2909,
+      "step": 1010
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.473135964912281e-06,
+      "loss": 0.9428,
+      "step": 1011
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.472587719298247e-06,
+      "loss": 1.4013,
+      "step": 1012
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.47203947368421e-06,
+      "loss": 1.319,
+      "step": 1013
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.471491228070176e-06,
+      "loss": 1.306,
+      "step": 1014
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.470942982456142e-06,
+      "loss": 1.1069,
+      "step": 1015
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.470394736842106e-06,
+      "loss": 1.4556,
+      "step": 1016
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.46984649122807e-06,
+      "loss": 1.1509,
+      "step": 1017
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.469298245614036e-06,
+      "loss": 1.1137,
+      "step": 1018
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.468750000000001e-06,
+      "loss": 1.0947,
+      "step": 1019
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.468201754385965e-06,
+      "loss": 1.4003,
+      "step": 1020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.46765350877193e-06,
+      "loss": 1.2034,
+      "step": 1021
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.467105263157895e-06,
+      "loss": 1.0069,
+      "step": 1022
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.46655701754386e-06,
+      "loss": 1.2635,
+      "step": 1023
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.466008771929825e-06,
+      "loss": 1.0455,
+      "step": 1024
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.46546052631579e-06,
+      "loss": 1.3723,
+      "step": 1025
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.464912280701754e-06,
+      "loss": 1.6388,
+      "step": 1026
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.46436403508772e-06,
+      "loss": 1.0357,
+      "step": 1027
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.463815789473684e-06,
+      "loss": 1.1893,
+      "step": 1028
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.46326754385965e-06,
+      "loss": 1.1694,
+      "step": 1029
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.462719298245615e-06,
+      "loss": 1.0924,
+      "step": 1030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.462171052631581e-06,
+      "loss": 1.3887,
+      "step": 1031
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.461622807017545e-06,
+      "loss": 1.2261,
+      "step": 1032
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.461074561403509e-06,
+      "loss": 1.0361,
+      "step": 1033
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.460526315789475e-06,
+      "loss": 0.9178,
+      "step": 1034
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.45997807017544e-06,
+      "loss": 1.3836,
+      "step": 1035
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.459429824561404e-06,
+      "loss": 1.2547,
+      "step": 1036
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.458881578947368e-06,
+      "loss": 1.2003,
+      "step": 1037
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.458333333333334e-06,
+      "loss": 1.3708,
+      "step": 1038
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.4577850877193e-06,
+      "loss": 1.4567,
+      "step": 1039
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.457236842105264e-06,
+      "loss": 0.9283,
+      "step": 1040
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.45668859649123e-06,
+      "loss": 1.196,
+      "step": 1041
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.456140350877195e-06,
+      "loss": 1.185,
+      "step": 1042
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.455592105263159e-06,
+      "loss": 0.9903,
+      "step": 1043
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.455043859649123e-06,
+      "loss": 1.0872,
+      "step": 1044
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.454495614035089e-06,
+      "loss": 0.9278,
+      "step": 1045
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.453947368421054e-06,
+      "loss": 1.3758,
+      "step": 1046
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.453399122807018e-06,
+      "loss": 1.5365,
+      "step": 1047
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.452850877192982e-06,
+      "loss": 1.5406,
+      "step": 1048
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.452302631578948e-06,
+      "loss": 1.1549,
+      "step": 1049
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.451754385964914e-06,
+      "loss": 0.8726,
+      "step": 1050
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.451206140350878e-06,
+      "loss": 1.4188,
+      "step": 1051
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.450657894736843e-06,
+      "loss": 1.0862,
+      "step": 1052
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.450109649122807e-06,
+      "loss": 1.1149,
+      "step": 1053
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.449561403508773e-06,
+      "loss": 0.8799,
+      "step": 1054
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.449013157894737e-06,
+      "loss": 0.9237,
+      "step": 1055
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.448464912280703e-06,
+      "loss": 1.1251,
+      "step": 1056
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.447916666666668e-06,
+      "loss": 1.4392,
+      "step": 1057
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.447368421052632e-06,
+      "loss": 1.3797,
+      "step": 1058
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.446820175438596e-06,
+      "loss": 1.6107,
+      "step": 1059
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.446271929824562e-06,
+      "loss": 1.2902,
+      "step": 1060
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.445723684210527e-06,
+      "loss": 1.4204,
+      "step": 1061
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.445175438596491e-06,
+      "loss": 0.851,
+      "step": 1062
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.444627192982457e-06,
+      "loss": 1.1136,
+      "step": 1063
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.444078947368421e-06,
+      "loss": 1.3759,
+      "step": 1064
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.443530701754387e-06,
+      "loss": 0.9931,
+      "step": 1065
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.44298245614035e-06,
+      "loss": 1.3122,
+      "step": 1066
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.442434210526316e-06,
+      "loss": 1.3691,
+      "step": 1067
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.441885964912282e-06,
+      "loss": 1.139,
+      "step": 1068
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.441337719298248e-06,
+      "loss": 1.2136,
+      "step": 1069
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.440789473684212e-06,
+      "loss": 1.1116,
+      "step": 1070
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.440241228070176e-06,
+      "loss": 0.9741,
+      "step": 1071
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.439692982456141e-06,
+      "loss": 1.2975,
+      "step": 1072
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.439144736842107e-06,
+      "loss": 1.0928,
+      "step": 1073
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.438596491228071e-06,
+      "loss": 1.2689,
+      "step": 1074
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.438048245614035e-06,
+      "loss": 0.9586,
+      "step": 1075
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.4375e-06,
+      "loss": 1.1927,
+      "step": 1076
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.436951754385966e-06,
+      "loss": 1.1833,
+      "step": 1077
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.43640350877193e-06,
+      "loss": 0.9895,
+      "step": 1078
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.435855263157896e-06,
+      "loss": 0.8224,
+      "step": 1079
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.43530701754386e-06,
+      "loss": 1.4687,
+      "step": 1080
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.434758771929826e-06,
+      "loss": 0.8144,
+      "step": 1081
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.43421052631579e-06,
+      "loss": 1.1118,
+      "step": 1082
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.433662280701755e-06,
+      "loss": 1.1123,
+      "step": 1083
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.433114035087721e-06,
+      "loss": 1.2221,
+      "step": 1084
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.432565789473685e-06,
+      "loss": 0.9424,
+      "step": 1085
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.432017543859649e-06,
+      "loss": 0.6738,
+      "step": 1086
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.431469298245615e-06,
+      "loss": 1.1483,
+      "step": 1087
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.43092105263158e-06,
+      "loss": 1.3352,
+      "step": 1088
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.430372807017544e-06,
+      "loss": 1.1146,
+      "step": 1089
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.42982456140351e-06,
+      "loss": 1.7168,
+      "step": 1090
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.429276315789474e-06,
+      "loss": 1.4468,
+      "step": 1091
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.42872807017544e-06,
+      "loss": 0.7348,
+      "step": 1092
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.428179824561404e-06,
+      "loss": 1.146,
+      "step": 1093
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.42763157894737e-06,
+      "loss": 1.3433,
+      "step": 1094
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.427083333333335e-06,
+      "loss": 1.4538,
+      "step": 1095
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.426535087719299e-06,
+      "loss": 0.9371,
+      "step": 1096
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.425986842105263e-06,
+      "loss": 1.0263,
+      "step": 1097
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.425438596491229e-06,
+      "loss": 1.4267,
+      "step": 1098
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.424890350877194e-06,
+      "loss": 1.0411,
+      "step": 1099
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.424342105263158e-06,
+      "loss": 1.0814,
+      "step": 1100
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.423793859649124e-06,
+      "loss": 1.2131,
+      "step": 1101
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.423245614035088e-06,
+      "loss": 0.8933,
+      "step": 1102
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.422697368421054e-06,
+      "loss": 1.0558,
+      "step": 1103
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.422149122807017e-06,
+      "loss": 1.104,
+      "step": 1104
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.421600877192983e-06,
+      "loss": 1.3473,
+      "step": 1105
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.421052631578949e-06,
+      "loss": 1.2908,
+      "step": 1106
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.420504385964913e-06,
+      "loss": 1.2004,
+      "step": 1107
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.419956140350878e-06,
+      "loss": 1.3328,
+      "step": 1108
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.419407894736842e-06,
+      "loss": 0.9187,
+      "step": 1109
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.418859649122808e-06,
+      "loss": 1.3277,
+      "step": 1110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.418311403508774e-06,
+      "loss": 1.0139,
+      "step": 1111
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.417763157894738e-06,
+      "loss": 1.022,
+      "step": 1112
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.417214912280702e-06,
+      "loss": 1.2201,
+      "step": 1113
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.416666666666667e-06,
+      "loss": 0.9551,
+      "step": 1114
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.416118421052633e-06,
+      "loss": 1.287,
+      "step": 1115
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.415570175438597e-06,
+      "loss": 1.2536,
+      "step": 1116
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.415021929824563e-06,
+      "loss": 1.3323,
+      "step": 1117
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.414473684210527e-06,
+      "loss": 0.9316,
+      "step": 1118
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.413925438596492e-06,
+      "loss": 1.0979,
+      "step": 1119
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.413377192982456e-06,
+      "loss": 1.0865,
+      "step": 1120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.412828947368422e-06,
+      "loss": 1.0014,
+      "step": 1121
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.412280701754388e-06,
+      "loss": 0.9333,
+      "step": 1122
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.411732456140352e-06,
+      "loss": 0.9565,
+      "step": 1123
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.411184210526316e-06,
+      "loss": 1.4697,
+      "step": 1124
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.410635964912281e-06,
+      "loss": 1.1529,
+      "step": 1125
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.410087719298247e-06,
+      "loss": 1.3792,
+      "step": 1126
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.409539473684211e-06,
+      "loss": 1.2724,
+      "step": 1127
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.408991228070177e-06,
+      "loss": 1.2922,
+      "step": 1128
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.40844298245614e-06,
+      "loss": 1.3374,
+      "step": 1129
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.407894736842106e-06,
+      "loss": 1.503,
+      "step": 1130
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.40734649122807e-06,
+      "loss": 1.3521,
+      "step": 1131
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.406798245614036e-06,
+      "loss": 1.0812,
+      "step": 1132
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.406250000000002e-06,
+      "loss": 1.0761,
+      "step": 1133
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.405701754385966e-06,
+      "loss": 1.4163,
+      "step": 1134
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.40515350877193e-06,
+      "loss": 0.9293,
+      "step": 1135
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.404605263157895e-06,
+      "loss": 1.9861,
+      "step": 1136
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.404057017543861e-06,
+      "loss": 1.3682,
+      "step": 1137
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.403508771929825e-06,
+      "loss": 1.0497,
+      "step": 1138
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.402960526315789e-06,
+      "loss": 0.9352,
+      "step": 1139
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.402412280701755e-06,
+      "loss": 1.3254,
+      "step": 1140
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.40186403508772e-06,
+      "loss": 1.3585,
+      "step": 1141
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.401315789473686e-06,
+      "loss": 1.2365,
+      "step": 1142
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.40076754385965e-06,
+      "loss": 0.8543,
+      "step": 1143
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.400219298245616e-06,
+      "loss": 1.3627,
+      "step": 1144
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.39967105263158e-06,
+      "loss": 1.3248,
+      "step": 1145
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.399122807017545e-06,
+      "loss": 1.1748,
+      "step": 1146
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.39857456140351e-06,
+      "loss": 1.2155,
+      "step": 1147
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.398026315789475e-06,
+      "loss": 0.8959,
+      "step": 1148
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.39747807017544e-06,
+      "loss": 1.3035,
+      "step": 1149
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.396929824561404e-06,
+      "loss": 1.3292,
+      "step": 1150
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.396381578947368e-06,
+      "loss": 1.3903,
+      "step": 1151
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.395833333333334e-06,
+      "loss": 1.2972,
+      "step": 1152
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.3952850877193e-06,
+      "loss": 1.4332,
+      "step": 1153
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.394736842105264e-06,
+      "loss": 1.2317,
+      "step": 1154
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.39418859649123e-06,
+      "loss": 1.2168,
+      "step": 1155
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.393640350877193e-06,
+      "loss": 1.1372,
+      "step": 1156
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.393092105263159e-06,
+      "loss": 0.8778,
+      "step": 1157
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.392543859649123e-06,
+      "loss": 0.9526,
+      "step": 1158
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.391995614035089e-06,
+      "loss": 1.3145,
+      "step": 1159
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.391447368421054e-06,
+      "loss": 1.2983,
+      "step": 1160
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.390899122807018e-06,
+      "loss": 1.1094,
+      "step": 1161
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.390350877192982e-06,
+      "loss": 1.1275,
+      "step": 1162
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.389802631578948e-06,
+      "loss": 2.2679,
+      "step": 1163
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.389254385964914e-06,
+      "loss": 0.9996,
+      "step": 1164
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.388706140350878e-06,
+      "loss": 1.3495,
+      "step": 1165
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.388157894736842e-06,
+      "loss": 0.9964,
+      "step": 1166
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.387609649122807e-06,
+      "loss": 1.3647,
+      "step": 1167
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.387061403508773e-06,
+      "loss": 1.4005,
+      "step": 1168
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.386513157894737e-06,
+      "loss": 1.0003,
+      "step": 1169
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.385964912280703e-06,
+      "loss": 1.3954,
+      "step": 1170
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.385416666666668e-06,
+      "loss": 0.9433,
+      "step": 1171
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.384868421052632e-06,
+      "loss": 0.9953,
+      "step": 1172
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.384320175438596e-06,
+      "loss": 1.2347,
+      "step": 1173
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.383771929824562e-06,
+      "loss": 1.2636,
+      "step": 1174
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.383223684210528e-06,
+      "loss": 1.2076,
+      "step": 1175
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.382675438596492e-06,
+      "loss": 1.3091,
+      "step": 1176
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.382127192982456e-06,
+      "loss": 1.0928,
+      "step": 1177
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.381578947368421e-06,
+      "loss": 1.0965,
+      "step": 1178
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.381030701754387e-06,
+      "loss": 1.1728,
+      "step": 1179
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.380482456140353e-06,
+      "loss": 1.4928,
+      "step": 1180
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.379934210526317e-06,
+      "loss": 1.5945,
+      "step": 1181
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.37938596491228e-06,
+      "loss": 0.8925,
+      "step": 1182
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.378837719298246e-06,
+      "loss": 1.0697,
+      "step": 1183
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.378289473684212e-06,
+      "loss": 1.3688,
+      "step": 1184
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.377741228070176e-06,
+      "loss": 0.9182,
+      "step": 1185
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.377192982456142e-06,
+      "loss": 1.0381,
+      "step": 1186
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.376644736842107e-06,
+      "loss": 1.0532,
+      "step": 1187
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.376096491228071e-06,
+      "loss": 1.0711,
+      "step": 1188
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.375548245614035e-06,
+      "loss": 1.1647,
+      "step": 1189
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.375000000000001e-06,
+      "loss": 1.2727,
+      "step": 1190
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.374451754385967e-06,
+      "loss": 1.3611,
+      "step": 1191
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.37390350877193e-06,
+      "loss": 1.3153,
+      "step": 1192
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.373355263157894e-06,
+      "loss": 1.4279,
+      "step": 1193
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.37280701754386e-06,
+      "loss": 1.3926,
+      "step": 1194
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.372258771929826e-06,
+      "loss": 1.231,
+      "step": 1195
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.37171052631579e-06,
+      "loss": 1.1295,
+      "step": 1196
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.371162280701755e-06,
+      "loss": 1.4405,
+      "step": 1197
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.370614035087721e-06,
+      "loss": 1.3009,
+      "step": 1198
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.370065789473685e-06,
+      "loss": 1.4016,
+      "step": 1199
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.369517543859649e-06,
+      "loss": 1.1549,
+      "step": 1200
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.368969298245615e-06,
+      "loss": 0.8532,
+      "step": 1201
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.36842105263158e-06,
+      "loss": 1.3606,
+      "step": 1202
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.367872807017544e-06,
+      "loss": 1.1874,
+      "step": 1203
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.367324561403508e-06,
+      "loss": 0.8429,
+      "step": 1204
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.366776315789474e-06,
+      "loss": 1.2779,
+      "step": 1205
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.36622807017544e-06,
+      "loss": 1.5,
+      "step": 1206
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.365679824561404e-06,
+      "loss": 0.8334,
+      "step": 1207
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.36513157894737e-06,
+      "loss": 1.1442,
+      "step": 1208
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.364583333333333e-06,
+      "loss": 1.4285,
+      "step": 1209
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.364035087719299e-06,
+      "loss": 1.1351,
+      "step": 1210
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.363486842105263e-06,
+      "loss": 1.2052,
+      "step": 1211
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.362938596491229e-06,
+      "loss": 0.7195,
+      "step": 1212
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.362390350877194e-06,
+      "loss": 1.244,
+      "step": 1213
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.361842105263158e-06,
+      "loss": 0.978,
+      "step": 1214
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.361293859649124e-06,
+      "loss": 1.2951,
+      "step": 1215
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.360745614035088e-06,
+      "loss": 1.3697,
+      "step": 1216
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.360197368421054e-06,
+      "loss": 1.2978,
+      "step": 1217
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.35964912280702e-06,
+      "loss": 0.8597,
+      "step": 1218
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.359100877192983e-06,
+      "loss": 1.4995,
+      "step": 1219
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.358552631578947e-06,
+      "loss": 1.3036,
+      "step": 1220
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.358004385964913e-06,
+      "loss": 0.7991,
+      "step": 1221
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.357456140350879e-06,
+      "loss": 0.9751,
+      "step": 1222
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.356907894736843e-06,
+      "loss": 1.4683,
+      "step": 1223
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.356359649122808e-06,
+      "loss": 0.7806,
+      "step": 1224
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.355811403508772e-06,
+      "loss": 1.5712,
+      "step": 1225
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.355263157894738e-06,
+      "loss": 1.2091,
+      "step": 1226
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.354714912280702e-06,
+      "loss": 1.0449,
+      "step": 1227
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.354166666666668e-06,
+      "loss": 1.1194,
+      "step": 1228
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.353618421052633e-06,
+      "loss": 1.1346,
+      "step": 1229
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.353070175438597e-06,
+      "loss": 1.2989,
+      "step": 1230
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.352521929824561e-06,
+      "loss": 1.0517,
+      "step": 1231
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.351973684210527e-06,
+      "loss": 1.3284,
+      "step": 1232
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.351425438596493e-06,
+      "loss": 1.4296,
+      "step": 1233
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.350877192982457e-06,
+      "loss": 1.4763,
+      "step": 1234
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.350328947368422e-06,
+      "loss": 1.4444,
+      "step": 1235
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.349780701754386e-06,
+      "loss": 1.2285,
+      "step": 1236
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.349232456140352e-06,
+      "loss": 1.6859,
+      "step": 1237
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.348684210526316e-06,
+      "loss": 1.5346,
+      "step": 1238
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.348135964912282e-06,
+      "loss": 1.7957,
+      "step": 1239
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.347587719298247e-06,
+      "loss": 1.1699,
+      "step": 1240
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.347039473684211e-06,
+      "loss": 1.3251,
+      "step": 1241
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.346491228070175e-06,
+      "loss": 1.2648,
+      "step": 1242
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.34594298245614e-06,
+      "loss": 1.4093,
+      "step": 1243
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.345394736842106e-06,
+      "loss": 1.4639,
+      "step": 1244
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.34484649122807e-06,
+      "loss": 1.3016,
+      "step": 1245
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.344298245614036e-06,
+      "loss": 0.9403,
+      "step": 1246
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.34375e-06,
+      "loss": 1.3665,
+      "step": 1247
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.343201754385966e-06,
+      "loss": 1.2579,
+      "step": 1248
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.34265350877193e-06,
+      "loss": 1.0434,
+      "step": 1249
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.342105263157895e-06,
+      "loss": 1.2797,
+      "step": 1250
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.341557017543861e-06,
+      "loss": 1.2245,
+      "step": 1251
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.341008771929825e-06,
+      "loss": 1.0126,
+      "step": 1252
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.34046052631579e-06,
+      "loss": 1.6172,
+      "step": 1253
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.339912280701755e-06,
+      "loss": 1.3291,
+      "step": 1254
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.33936403508772e-06,
+      "loss": 1.6026,
+      "step": 1255
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.338815789473686e-06,
+      "loss": 1.1072,
+      "step": 1256
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.33826754385965e-06,
+      "loss": 1.0074,
+      "step": 1257
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.337719298245614e-06,
+      "loss": 1.2725,
+      "step": 1258
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.33717105263158e-06,
+      "loss": 1.0511,
+      "step": 1259
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.336622807017545e-06,
+      "loss": 0.9787,
+      "step": 1260
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.33607456140351e-06,
+      "loss": 1.2891,
+      "step": 1261
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.335526315789475e-06,
+      "loss": 1.1626,
+      "step": 1262
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.334978070175439e-06,
+      "loss": 1.2339,
+      "step": 1263
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.334429824561405e-06,
+      "loss": 0.7681,
+      "step": 1264
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.333881578947369e-06,
+      "loss": 1.3096,
+      "step": 1265
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.333333333333334e-06,
+      "loss": 1.5535,
+      "step": 1266
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.3327850877193e-06,
+      "loss": 0.8929,
+      "step": 1267
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.332236842105264e-06,
+      "loss": 1.4391,
+      "step": 1268
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.331688596491228e-06,
+      "loss": 1.0387,
+      "step": 1269
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.331140350877194e-06,
+      "loss": 1.4616,
+      "step": 1270
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.33059210526316e-06,
+      "loss": 1.2473,
+      "step": 1271
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.330043859649123e-06,
+      "loss": 1.0871,
+      "step": 1272
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.329495614035089e-06,
+      "loss": 1.4076,
+      "step": 1273
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.328947368421053e-06,
+      "loss": 0.863,
+      "step": 1274
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.328399122807019e-06,
+      "loss": 1.1719,
+      "step": 1275
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.327850877192983e-06,
+      "loss": 1.3886,
+      "step": 1276
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.327302631578948e-06,
+      "loss": 1.123,
+      "step": 1277
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.326754385964914e-06,
+      "loss": 0.9162,
+      "step": 1278
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.326206140350878e-06,
+      "loss": 1.1479,
+      "step": 1279
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.325657894736842e-06,
+      "loss": 1.3832,
+      "step": 1280
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.325109649122808e-06,
+      "loss": 1.5941,
+      "step": 1281
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.324561403508773e-06,
+      "loss": 0.7984,
+      "step": 1282
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.324013157894737e-06,
+      "loss": 1.1849,
+      "step": 1283
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.323464912280703e-06,
+      "loss": 1.1415,
+      "step": 1284
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.322916666666667e-06,
+      "loss": 1.2339,
+      "step": 1285
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.322368421052633e-06,
+      "loss": 0.7566,
+      "step": 1286
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.321820175438596e-06,
+      "loss": 1.6407,
+      "step": 1287
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.321271929824562e-06,
+      "loss": 1.405,
+      "step": 1288
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.320723684210528e-06,
+      "loss": 1.1408,
+      "step": 1289
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.320175438596492e-06,
+      "loss": 1.0088,
+      "step": 1290
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.319627192982457e-06,
+      "loss": 1.8424,
+      "step": 1291
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.319078947368421e-06,
+      "loss": 1.3108,
+      "step": 1292
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.318530701754387e-06,
+      "loss": 1.1241,
+      "step": 1293
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.317982456140353e-06,
+      "loss": 0.9525,
+      "step": 1294
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.317434210526317e-06,
+      "loss": 1.2644,
+      "step": 1295
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.31688596491228e-06,
+      "loss": 0.9938,
+      "step": 1296
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.316337719298246e-06,
+      "loss": 1.0159,
+      "step": 1297
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.315789473684212e-06,
+      "loss": 1.2474,
+      "step": 1298
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.315241228070176e-06,
+      "loss": 0.812,
+      "step": 1299
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.314692982456142e-06,
+      "loss": 1.1927,
+      "step": 1300
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.314144736842106e-06,
+      "loss": 1.0025,
+      "step": 1301
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.313596491228071e-06,
+      "loss": 0.9498,
+      "step": 1302
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.313048245614035e-06,
+      "loss": 1.1001,
+      "step": 1303
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.312500000000001e-06,
+      "loss": 1.0212,
+      "step": 1304
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.311951754385967e-06,
+      "loss": 1.0529,
+      "step": 1305
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.31140350877193e-06,
+      "loss": 1.0336,
+      "step": 1306
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.310855263157895e-06,
+      "loss": 1.0851,
+      "step": 1307
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.31030701754386e-06,
+      "loss": 1.402,
+      "step": 1308
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.309758771929826e-06,
+      "loss": 0.8297,
+      "step": 1309
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.30921052631579e-06,
+      "loss": 0.9341,
+      "step": 1310
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.308662280701754e-06,
+      "loss": 0.9813,
+      "step": 1311
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.30811403508772e-06,
+      "loss": 1.6664,
+      "step": 1312
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.307565789473685e-06,
+      "loss": 1.3339,
+      "step": 1313
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.30701754385965e-06,
+      "loss": 1.1539,
+      "step": 1314
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.306469298245615e-06,
+      "loss": 1.1825,
+      "step": 1315
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.30592105263158e-06,
+      "loss": 0.8676,
+      "step": 1316
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.305372807017545e-06,
+      "loss": 1.2109,
+      "step": 1317
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.304824561403509e-06,
+      "loss": 0.8776,
+      "step": 1318
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.304276315789474e-06,
+      "loss": 1.2378,
+      "step": 1319
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.30372807017544e-06,
+      "loss": 1.5578,
+      "step": 1320
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.303179824561404e-06,
+      "loss": 1.1551,
+      "step": 1321
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.302631578947368e-06,
+      "loss": 1.3116,
+      "step": 1322
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.302083333333334e-06,
+      "loss": 1.2238,
+      "step": 1323
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.3015350877193e-06,
+      "loss": 0.9159,
+      "step": 1324
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.300986842105263e-06,
+      "loss": 1.4038,
+      "step": 1325
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.300438596491229e-06,
+      "loss": 1.5643,
+      "step": 1326
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.299890350877195e-06,
+      "loss": 0.8969,
+      "step": 1327
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.299342105263159e-06,
+      "loss": 0.98,
+      "step": 1328
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.298793859649124e-06,
+      "loss": 1.3241,
+      "step": 1329
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.298245614035088e-06,
+      "loss": 0.978,
+      "step": 1330
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.297697368421054e-06,
+      "loss": 1.2626,
+      "step": 1331
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.29714912280702e-06,
+      "loss": 1.4727,
+      "step": 1332
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.296600877192984e-06,
+      "loss": 1.1,
+      "step": 1333
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.296052631578947e-06,
+      "loss": 1.7227,
+      "step": 1334
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.295504385964913e-06,
+      "loss": 1.175,
+      "step": 1335
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.294956140350879e-06,
+      "loss": 1.0246,
+      "step": 1336
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.294407894736843e-06,
+      "loss": 1.2136,
+      "step": 1337
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.293859649122807e-06,
+      "loss": 1.2425,
+      "step": 1338
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.293311403508772e-06,
+      "loss": 1.5903,
+      "step": 1339
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.292763157894738e-06,
+      "loss": 1.1001,
+      "step": 1340
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.292214912280702e-06,
+      "loss": 0.7359,
+      "step": 1341
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.291666666666668e-06,
+      "loss": 1.1143,
+      "step": 1342
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.291118421052633e-06,
+      "loss": 1.0185,
+      "step": 1343
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.290570175438597e-06,
+      "loss": 2.0014,
+      "step": 1344
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.290021929824561e-06,
+      "loss": 1.2041,
+      "step": 1345
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.289473684210527e-06,
+      "loss": 0.9598,
+      "step": 1346
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.288925438596493e-06,
+      "loss": 0.8631,
+      "step": 1347
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.288377192982457e-06,
+      "loss": 0.8991,
+      "step": 1348
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.28782894736842e-06,
+      "loss": 1.6891,
+      "step": 1349
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.287280701754386e-06,
+      "loss": 0.9816,
+      "step": 1350
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.286732456140352e-06,
+      "loss": 0.9632,
+      "step": 1351
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.286184210526316e-06,
+      "loss": 1.263,
+      "step": 1352
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.285635964912282e-06,
+      "loss": 1.6317,
+      "step": 1353
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.285087719298247e-06,
+      "loss": 1.6011,
+      "step": 1354
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.284539473684211e-06,
+      "loss": 0.8387,
+      "step": 1355
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.283991228070175e-06,
+      "loss": 0.9544,
+      "step": 1356
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.283442982456141e-06,
+      "loss": 0.7587,
+      "step": 1357
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.282894736842107e-06,
+      "loss": 1.1616,
+      "step": 1358
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.28234649122807e-06,
+      "loss": 1.0514,
+      "step": 1359
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.281798245614035e-06,
+      "loss": 1.0484,
+      "step": 1360
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.28125e-06,
+      "loss": 0.8764,
+      "step": 1361
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.280701754385966e-06,
+      "loss": 1.0567,
+      "step": 1362
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.28015350877193e-06,
+      "loss": 0.9521,
+      "step": 1363
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.279605263157896e-06,
+      "loss": 1.0174,
+      "step": 1364
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.27905701754386e-06,
+      "loss": 0.8289,
+      "step": 1365
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.278508771929825e-06,
+      "loss": 1.5319,
+      "step": 1366
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.277960526315791e-06,
+      "loss": 1.0531,
+      "step": 1367
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.277412280701755e-06,
+      "loss": 1.4674,
+      "step": 1368
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.27686403508772e-06,
+      "loss": 1.1776,
+      "step": 1369
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.276315789473686e-06,
+      "loss": 1.3526,
+      "step": 1370
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.27576754385965e-06,
+      "loss": 1.158,
+      "step": 1371
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.275219298245614e-06,
+      "loss": 0.9871,
+      "step": 1372
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.27467105263158e-06,
+      "loss": 1.1526,
+      "step": 1373
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.274122807017546e-06,
+      "loss": 1.242,
+      "step": 1374
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.27357456140351e-06,
+      "loss": 1.0126,
+      "step": 1375
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.273026315789474e-06,
+      "loss": 1.1522,
+      "step": 1376
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.27247807017544e-06,
+      "loss": 1.1602,
+      "step": 1377
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.271929824561405e-06,
+      "loss": 1.1455,
+      "step": 1378
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.271381578947369e-06,
+      "loss": 1.692,
+      "step": 1379
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.270833333333334e-06,
+      "loss": 0.9547,
+      "step": 1380
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.270285087719298e-06,
+      "loss": 1.0863,
+      "step": 1381
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.269736842105264e-06,
+      "loss": 0.9448,
+      "step": 1382
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.269188596491228e-06,
+      "loss": 1.2907,
+      "step": 1383
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.268640350877194e-06,
+      "loss": 1.1102,
+      "step": 1384
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.26809210526316e-06,
+      "loss": 1.0869,
+      "step": 1385
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.267543859649123e-06,
+      "loss": 1.2259,
+      "step": 1386
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.266995614035087e-06,
+      "loss": 1.1781,
+      "step": 1387
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.266447368421053e-06,
+      "loss": 1.0502,
+      "step": 1388
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.265899122807019e-06,
+      "loss": 1.3089,
+      "step": 1389
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.265350877192983e-06,
+      "loss": 1.2559,
+      "step": 1390
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.264802631578948e-06,
+      "loss": 1.3721,
+      "step": 1391
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.264254385964912e-06,
+      "loss": 1.0953,
+      "step": 1392
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.263706140350878e-06,
+      "loss": 1.1067,
+      "step": 1393
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.263157894736842e-06,
+      "loss": 1.3758,
+      "step": 1394
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.262609649122808e-06,
+      "loss": 0.7584,
+      "step": 1395
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.262061403508773e-06,
+      "loss": 1.0315,
+      "step": 1396
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.261513157894737e-06,
+      "loss": 0.7629,
+      "step": 1397
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.260964912280701e-06,
+      "loss": 1.7739,
+      "step": 1398
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.260416666666667e-06,
+      "loss": 0.9944,
+      "step": 1399
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.259868421052633e-06,
+      "loss": 0.8441,
+      "step": 1400
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.259320175438598e-06,
+      "loss": 1.4191,
+      "step": 1401
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.258771929824562e-06,
+      "loss": 1.3696,
+      "step": 1402
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.258223684210526e-06,
+      "loss": 0.5777,
+      "step": 1403
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.257675438596492e-06,
+      "loss": 1.498,
+      "step": 1404
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.257127192982458e-06,
+      "loss": 1.5032,
+      "step": 1405
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.256578947368422e-06,
+      "loss": 1.0476,
+      "step": 1406
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.256030701754387e-06,
+      "loss": 0.9014,
+      "step": 1407
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.255482456140351e-06,
+      "loss": 1.1065,
+      "step": 1408
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.254934210526317e-06,
+      "loss": 1.1248,
+      "step": 1409
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.254385964912281e-06,
+      "loss": 1.0573,
+      "step": 1410
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.253837719298247e-06,
+      "loss": 0.8697,
+      "step": 1411
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.253289473684212e-06,
+      "loss": 0.7925,
+      "step": 1412
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.252741228070176e-06,
+      "loss": 1.3427,
+      "step": 1413
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.25219298245614e-06,
+      "loss": 1.0504,
+      "step": 1414
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.251644736842106e-06,
+      "loss": 1.8093,
+      "step": 1415
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.251096491228072e-06,
+      "loss": 1.1698,
+      "step": 1416
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.250548245614036e-06,
+      "loss": 1.7879,
+      "step": 1417
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.250000000000001e-06,
+      "loss": 1.0235,
+      "step": 1418
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.249451754385965e-06,
+      "loss": 1.6141,
+      "step": 1419
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.248903508771931e-06,
+      "loss": 1.3342,
+      "step": 1420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.248355263157895e-06,
+      "loss": 0.9491,
+      "step": 1421
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.24780701754386e-06,
+      "loss": 0.8826,
+      "step": 1422
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.247258771929826e-06,
+      "loss": 1.4669,
+      "step": 1423
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.24671052631579e-06,
+      "loss": 0.8098,
+      "step": 1424
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.246162280701754e-06,
+      "loss": 1.5417,
+      "step": 1425
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.24561403508772e-06,
+      "loss": 1.072,
+      "step": 1426
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.245065789473685e-06,
+      "loss": 1.7743,
+      "step": 1427
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.24451754385965e-06,
+      "loss": 1.1459,
+      "step": 1428
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.243969298245615e-06,
+      "loss": 1.6599,
+      "step": 1429
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.243421052631579e-06,
+      "loss": 1.1403,
+      "step": 1430
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.242872807017545e-06,
+      "loss": 1.328,
+      "step": 1431
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.242324561403509e-06,
+      "loss": 1.1637,
+      "step": 1432
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.241776315789474e-06,
+      "loss": 0.769,
+      "step": 1433
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.24122807017544e-06,
+      "loss": 1.2736,
+      "step": 1434
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.240679824561404e-06,
+      "loss": 0.9858,
+      "step": 1435
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.240131578947368e-06,
+      "loss": 1.5293,
+      "step": 1436
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.239583333333334e-06,
+      "loss": 1.3206,
+      "step": 1437
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.2390350877193e-06,
+      "loss": 1.1731,
+      "step": 1438
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.238486842105265e-06,
+      "loss": 1.0929,
+      "step": 1439
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.237938596491229e-06,
+      "loss": 1.0258,
+      "step": 1440
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.237390350877193e-06,
+      "loss": 1.7288,
+      "step": 1441
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.236842105263159e-06,
+      "loss": 0.9944,
+      "step": 1442
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.236293859649124e-06,
+      "loss": 1.6123,
+      "step": 1443
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.235745614035088e-06,
+      "loss": 1.3686,
+      "step": 1444
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.235197368421054e-06,
+      "loss": 0.88,
+      "step": 1445
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.234649122807018e-06,
+      "loss": 0.9311,
+      "step": 1446
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.234100877192984e-06,
+      "loss": 0.6611,
+      "step": 1447
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.233552631578948e-06,
+      "loss": 1.2245,
+      "step": 1448
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.233004385964913e-06,
+      "loss": 1.0569,
+      "step": 1449
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.232456140350879e-06,
+      "loss": 1.7502,
+      "step": 1450
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.231907894736843e-06,
+      "loss": 0.8562,
+      "step": 1451
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.231359649122807e-06,
+      "loss": 1.3962,
+      "step": 1452
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.230811403508773e-06,
+      "loss": 1.4035,
+      "step": 1453
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.230263157894738e-06,
+      "loss": 1.3651,
+      "step": 1454
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.229714912280702e-06,
+      "loss": 1.233,
+      "step": 1455
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.229166666666668e-06,
+      "loss": 0.9204,
+      "step": 1456
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.228618421052632e-06,
+      "loss": 1.1521,
+      "step": 1457
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.228070175438598e-06,
+      "loss": 0.826,
+      "step": 1458
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.227521929824562e-06,
+      "loss": 1.4339,
+      "step": 1459
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.226973684210527e-06,
+      "loss": 1.2589,
+      "step": 1460
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.226425438596493e-06,
+      "loss": 1.4699,
+      "step": 1461
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.225877192982457e-06,
+      "loss": 0.9702,
+      "step": 1462
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.225328947368421e-06,
+      "loss": 0.8809,
+      "step": 1463
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.224780701754387e-06,
+      "loss": 1.3064,
+      "step": 1464
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.224232456140352e-06,
+      "loss": 1.0029,
+      "step": 1465
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.223684210526316e-06,
+      "loss": 1.2937,
+      "step": 1466
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.22313596491228e-06,
+      "loss": 1.2693,
+      "step": 1467
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.222587719298246e-06,
+      "loss": 0.8106,
+      "step": 1468
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.222039473684212e-06,
+      "loss": 1.0028,
+      "step": 1469
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.221491228070175e-06,
+      "loss": 1.1536,
+      "step": 1470
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.220942982456141e-06,
+      "loss": 1.0471,
+      "step": 1471
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.220394736842107e-06,
+      "loss": 1.303,
+      "step": 1472
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.21984649122807e-06,
+      "loss": 0.5747,
+      "step": 1473
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.219298245614035e-06,
+      "loss": 1.4345,
+      "step": 1474
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.21875e-06,
+      "loss": 1.1474,
+      "step": 1475
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.218201754385966e-06,
+      "loss": 1.9462,
+      "step": 1476
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.217653508771932e-06,
+      "loss": 0.9062,
+      "step": 1477
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.217105263157896e-06,
+      "loss": 1.166,
+      "step": 1478
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.21655701754386e-06,
+      "loss": 0.6263,
+      "step": 1479
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.216008771929825e-06,
+      "loss": 1.411,
+      "step": 1480
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.215460526315791e-06,
+      "loss": 1.3703,
+      "step": 1481
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.214912280701755e-06,
+      "loss": 0.9505,
+      "step": 1482
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.21436403508772e-06,
+      "loss": 1.2346,
+      "step": 1483
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.213815789473685e-06,
+      "loss": 0.9993,
+      "step": 1484
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.21326754385965e-06,
+      "loss": 1.35,
+      "step": 1485
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.212719298245614e-06,
+      "loss": 1.0425,
+      "step": 1486
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.21217105263158e-06,
+      "loss": 1.3112,
+      "step": 1487
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.211622807017546e-06,
+      "loss": 1.1507,
+      "step": 1488
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.21107456140351e-06,
+      "loss": 1.1185,
+      "step": 1489
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.210526315789474e-06,
+      "loss": 1.6362,
+      "step": 1490
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.20997807017544e-06,
+      "loss": 1.5506,
+      "step": 1491
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.209429824561405e-06,
+      "loss": 0.6661,
+      "step": 1492
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.208881578947369e-06,
+      "loss": 1.2201,
+      "step": 1493
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.208333333333333e-06,
+      "loss": 0.8768,
+      "step": 1494
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.207785087719299e-06,
+      "loss": 1.0375,
+      "step": 1495
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.207236842105264e-06,
+      "loss": 0.7469,
+      "step": 1496
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.206688596491228e-06,
+      "loss": 0.858,
+      "step": 1497
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.206140350877194e-06,
+      "loss": 0.685,
+      "step": 1498
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.20559210526316e-06,
+      "loss": 1.0901,
+      "step": 1499
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.205043859649124e-06,
+      "loss": 1.3182,
+      "step": 1500
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.204495614035088e-06,
+      "loss": 0.8294,
+      "step": 1501
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.203947368421053e-06,
+      "loss": 0.943,
+      "step": 1502
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.203399122807019e-06,
+      "loss": 1.2669,
+      "step": 1503
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.202850877192983e-06,
+      "loss": 1.1411,
+      "step": 1504
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.202302631578947e-06,
+      "loss": 1.0086,
+      "step": 1505
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.201754385964913e-06,
+      "loss": 1.3222,
+      "step": 1506
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.201206140350878e-06,
+      "loss": 0.962,
+      "step": 1507
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.200657894736842e-06,
+      "loss": 0.9153,
+      "step": 1508
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.200109649122808e-06,
+      "loss": 1.4224,
+      "step": 1509
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.199561403508772e-06,
+      "loss": 1.0408,
+      "step": 1510
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.199013157894738e-06,
+      "loss": 0.7368,
+      "step": 1511
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.198464912280703e-06,
+      "loss": 0.7224,
+      "step": 1512
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.197916666666667e-06,
+      "loss": 1.0745,
+      "step": 1513
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.197368421052633e-06,
+      "loss": 1.2979,
+      "step": 1514
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.196820175438599e-06,
+      "loss": 1.4069,
+      "step": 1515
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.196271929824563e-06,
+      "loss": 1.3879,
+      "step": 1516
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.195723684210526e-06,
+      "loss": 1.403,
+      "step": 1517
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.195175438596492e-06,
+      "loss": 1.6171,
+      "step": 1518
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.194627192982458e-06,
+      "loss": 1.4009,
+      "step": 1519
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.194078947368422e-06,
+      "loss": 1.1313,
+      "step": 1520
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.193530701754386e-06,
+      "loss": 0.8152,
+      "step": 1521
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.192982456140351e-06,
+      "loss": 1.0986,
+      "step": 1522
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.192434210526317e-06,
+      "loss": 1.1748,
+      "step": 1523
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.191885964912281e-06,
+      "loss": 1.2528,
+      "step": 1524
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.191337719298247e-06,
+      "loss": 0.6482,
+      "step": 1525
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.190789473684212e-06,
+      "loss": 1.0179,
+      "step": 1526
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.190241228070176e-06,
+      "loss": 0.654,
+      "step": 1527
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.18969298245614e-06,
+      "loss": 1.1459,
+      "step": 1528
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.189144736842106e-06,
+      "loss": 1.3981,
+      "step": 1529
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.188596491228072e-06,
+      "loss": 1.0335,
+      "step": 1530
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.188048245614036e-06,
+      "loss": 1.1272,
+      "step": 1531
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.1875e-06,
+      "loss": 0.8301,
+      "step": 1532
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.186951754385965e-06,
+      "loss": 1.3914,
+      "step": 1533
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.186403508771931e-06,
+      "loss": 1.3058,
+      "step": 1534
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.185855263157895e-06,
+      "loss": 0.8226,
+      "step": 1535
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.18530701754386e-06,
+      "loss": 1.1225,
+      "step": 1536
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.184758771929825e-06,
+      "loss": 1.5595,
+      "step": 1537
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.18421052631579e-06,
+      "loss": 0.9996,
+      "step": 1538
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.183662280701754e-06,
+      "loss": 1.3103,
+      "step": 1539
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.18311403508772e-06,
+      "loss": 1.0318,
+      "step": 1540
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.182565789473686e-06,
+      "loss": 0.8449,
+      "step": 1541
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.18201754385965e-06,
+      "loss": 1.3513,
+      "step": 1542
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.181469298245614e-06,
+      "loss": 1.3667,
+      "step": 1543
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.18092105263158e-06,
+      "loss": 0.8286,
+      "step": 1544
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.180372807017545e-06,
+      "loss": 1.0254,
+      "step": 1545
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.179824561403509e-06,
+      "loss": 1.0852,
+      "step": 1546
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.179276315789475e-06,
+      "loss": 1.0948,
+      "step": 1547
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.178728070175439e-06,
+      "loss": 1.1069,
+      "step": 1548
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.178179824561404e-06,
+      "loss": 1.3166,
+      "step": 1549
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.17763157894737e-06,
+      "loss": 0.9801,
+      "step": 1550
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.177083333333334e-06,
+      "loss": 0.9043,
+      "step": 1551
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.1765350877193e-06,
+      "loss": 1.2215,
+      "step": 1552
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.175986842105264e-06,
+      "loss": 0.8749,
+      "step": 1553
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.17543859649123e-06,
+      "loss": 1.7246,
+      "step": 1554
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.174890350877193e-06,
+      "loss": 1.1586,
+      "step": 1555
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.174342105263159e-06,
+      "loss": 0.9473,
+      "step": 1556
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.173793859649125e-06,
+      "loss": 1.0711,
+      "step": 1557
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.173245614035089e-06,
+      "loss": 1.3365,
+      "step": 1558
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.172697368421053e-06,
+      "loss": 1.5191,
+      "step": 1559
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.172149122807018e-06,
+      "loss": 0.965,
+      "step": 1560
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.171600877192984e-06,
+      "loss": 1.0459,
+      "step": 1561
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.171052631578948e-06,
+      "loss": 1.4467,
+      "step": 1562
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.170504385964914e-06,
+      "loss": 1.1501,
+      "step": 1563
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.169956140350877e-06,
+      "loss": 0.9849,
+      "step": 1564
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.169407894736843e-06,
+      "loss": 1.7106,
+      "step": 1565
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.168859649122807e-06,
+      "loss": 1.257,
+      "step": 1566
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.168311403508773e-06,
+      "loss": 0.7997,
+      "step": 1567
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.167763157894738e-06,
+      "loss": 1.166,
+      "step": 1568
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.167214912280702e-06,
+      "loss": 1.2542,
+      "step": 1569
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.166666666666666e-06,
+      "loss": 1.1116,
+      "step": 1570
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.166118421052632e-06,
+      "loss": 1.2098,
+      "step": 1571
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.165570175438598e-06,
+      "loss": 0.7681,
+      "step": 1572
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.165021929824562e-06,
+      "loss": 1.2585,
+      "step": 1573
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.164473684210527e-06,
+      "loss": 0.9305,
+      "step": 1574
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.163925438596491e-06,
+      "loss": 1.3061,
+      "step": 1575
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.163377192982457e-06,
+      "loss": 1.1783,
+      "step": 1576
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.162828947368421e-06,
+      "loss": 1.0579,
+      "step": 1577
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.162280701754387e-06,
+      "loss": 0.9789,
+      "step": 1578
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.161732456140352e-06,
+      "loss": 1.387,
+      "step": 1579
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.161184210526316e-06,
+      "loss": 1.0374,
+      "step": 1580
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.16063596491228e-06,
+      "loss": 1.1769,
+      "step": 1581
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.160087719298246e-06,
+      "loss": 0.981,
+      "step": 1582
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.159539473684212e-06,
+      "loss": 1.0804,
+      "step": 1583
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.158991228070176e-06,
+      "loss": 1.7583,
+      "step": 1584
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.158442982456141e-06,
+      "loss": 0.9901,
+      "step": 1585
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.157894736842105e-06,
+      "loss": 1.2938,
+      "step": 1586
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.157346491228071e-06,
+      "loss": 1.3399,
+      "step": 1587
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.156798245614037e-06,
+      "loss": 1.027,
+      "step": 1588
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.15625e-06,
+      "loss": 0.9799,
+      "step": 1589
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.155701754385966e-06,
+      "loss": 1.6178,
+      "step": 1590
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.15515350877193e-06,
+      "loss": 1.2141,
+      "step": 1591
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.154605263157896e-06,
+      "loss": 1.3368,
+      "step": 1592
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.15405701754386e-06,
+      "loss": 1.0728,
+      "step": 1593
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.153508771929826e-06,
+      "loss": 0.9912,
+      "step": 1594
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.152960526315791e-06,
+      "loss": 0.87,
+      "step": 1595
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.152412280701755e-06,
+      "loss": 1.1577,
+      "step": 1596
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.15186403508772e-06,
+      "loss": 1.3109,
+      "step": 1597
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.151315789473685e-06,
+      "loss": 1.0694,
+      "step": 1598
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.15076754385965e-06,
+      "loss": 1.2763,
+      "step": 1599
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.150219298245615e-06,
+      "loss": 1.6378,
+      "step": 1600
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.14967105263158e-06,
+      "loss": 1.3756,
+      "step": 1601
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.149122807017544e-06,
+      "loss": 0.7313,
+      "step": 1602
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.14857456140351e-06,
+      "loss": 1.0868,
+      "step": 1603
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.148026315789474e-06,
+      "loss": 1.1997,
+      "step": 1604
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.14747807017544e-06,
+      "loss": 1.0816,
+      "step": 1605
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.146929824561405e-06,
+      "loss": 1.386,
+      "step": 1606
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.14638157894737e-06,
+      "loss": 1.076,
+      "step": 1607
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.145833333333333e-06,
+      "loss": 1.3938,
+      "step": 1608
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.145285087719299e-06,
+      "loss": 1.1123,
+      "step": 1609
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.144736842105264e-06,
+      "loss": 1.1789,
+      "step": 1610
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.144188596491228e-06,
+      "loss": 1.3312,
+      "step": 1611
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.143640350877194e-06,
+      "loss": 1.5751,
+      "step": 1612
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.143092105263158e-06,
+      "loss": 1.5112,
+      "step": 1613
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.142543859649124e-06,
+      "loss": 0.7303,
+      "step": 1614
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.141995614035088e-06,
+      "loss": 0.913,
+      "step": 1615
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.141447368421053e-06,
+      "loss": 1.8944,
+      "step": 1616
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.140899122807019e-06,
+      "loss": 0.8747,
+      "step": 1617
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.140350877192983e-06,
+      "loss": 1.4807,
+      "step": 1618
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.139802631578947e-06,
+      "loss": 1.0457,
+      "step": 1619
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.139254385964913e-06,
+      "loss": 1.0467,
+      "step": 1620
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.138706140350878e-06,
+      "loss": 0.6365,
+      "step": 1621
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.138157894736842e-06,
+      "loss": 0.7077,
+      "step": 1622
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.137609649122808e-06,
+      "loss": 1.0921,
+      "step": 1623
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.137061403508772e-06,
+      "loss": 0.605,
+      "step": 1624
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.136513157894738e-06,
+      "loss": 1.6211,
+      "step": 1625
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.135964912280703e-06,
+      "loss": 1.4223,
+      "step": 1626
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.135416666666667e-06,
+      "loss": 0.8883,
+      "step": 1627
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.134868421052633e-06,
+      "loss": 1.039,
+      "step": 1628
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.134320175438597e-06,
+      "loss": 0.9505,
+      "step": 1629
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.133771929824563e-06,
+      "loss": 1.2434,
+      "step": 1630
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.133223684210527e-06,
+      "loss": 0.8477,
+      "step": 1631
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.132675438596492e-06,
+      "loss": 1.3471,
+      "step": 1632
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.132127192982458e-06,
+      "loss": 1.1322,
+      "step": 1633
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.131578947368422e-06,
+      "loss": 1.465,
+      "step": 1634
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.131030701754386e-06,
+      "loss": 0.8017,
+      "step": 1635
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.130482456140352e-06,
+      "loss": 1.1232,
+      "step": 1636
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.129934210526317e-06,
+      "loss": 0.8563,
+      "step": 1637
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.129385964912281e-06,
+      "loss": 1.6043,
+      "step": 1638
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.128837719298247e-06,
+      "loss": 1.6069,
+      "step": 1639
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.128289473684211e-06,
+      "loss": 1.0519,
+      "step": 1640
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.127741228070177e-06,
+      "loss": 1.1955,
+      "step": 1641
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.12719298245614e-06,
+      "loss": 1.2037,
+      "step": 1642
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.126644736842106e-06,
+      "loss": 1.1574,
+      "step": 1643
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.126096491228072e-06,
+      "loss": 0.7217,
+      "step": 1644
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.125548245614036e-06,
+      "loss": 1.0539,
+      "step": 1645
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.125e-06,
+      "loss": 1.0666,
+      "step": 1646
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.124451754385966e-06,
+      "loss": 0.7011,
+      "step": 1647
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.123903508771931e-06,
+      "loss": 1.1146,
+      "step": 1648
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.123355263157895e-06,
+      "loss": 1.3585,
+      "step": 1649
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.12280701754386e-06,
+      "loss": 1.2969,
+      "step": 1650
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.122258771929825e-06,
+      "loss": 1.4808,
+      "step": 1651
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.12171052631579e-06,
+      "loss": 1.1544,
+      "step": 1652
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.121162280701754e-06,
+      "loss": 1.2895,
+      "step": 1653
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.12061403508772e-06,
+      "loss": 1.2508,
+      "step": 1654
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.120065789473686e-06,
+      "loss": 0.9552,
+      "step": 1655
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.11951754385965e-06,
+      "loss": 1.0063,
+      "step": 1656
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.118969298245614e-06,
+      "loss": 0.7896,
+      "step": 1657
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.11842105263158e-06,
+      "loss": 1.0933,
+      "step": 1658
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.117872807017545e-06,
+      "loss": 1.1853,
+      "step": 1659
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.11732456140351e-06,
+      "loss": 0.7587,
+      "step": 1660
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.116776315789475e-06,
+      "loss": 0.8038,
+      "step": 1661
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.116228070175439e-06,
+      "loss": 0.7677,
+      "step": 1662
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.115679824561404e-06,
+      "loss": 1.1078,
+      "step": 1663
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.11513157894737e-06,
+      "loss": 1.233,
+      "step": 1664
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.114583333333334e-06,
+      "loss": 1.2196,
+      "step": 1665
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.114035087719298e-06,
+      "loss": 1.1837,
+      "step": 1666
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.113486842105264e-06,
+      "loss": 0.4497,
+      "step": 1667
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.11293859649123e-06,
+      "loss": 0.7942,
+      "step": 1668
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.112390350877193e-06,
+      "loss": 1.323,
+      "step": 1669
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.111842105263159e-06,
+      "loss": 1.3368,
+      "step": 1670
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.111293859649125e-06,
+      "loss": 1.354,
+      "step": 1671
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.110745614035089e-06,
+      "loss": 1.5253,
+      "step": 1672
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.110197368421053e-06,
+      "loss": 1.5332,
+      "step": 1673
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.109649122807018e-06,
+      "loss": 0.7388,
+      "step": 1674
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.109100877192984e-06,
+      "loss": 1.4899,
+      "step": 1675
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.108552631578948e-06,
+      "loss": 0.9589,
+      "step": 1676
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.108004385964912e-06,
+      "loss": 0.9128,
+      "step": 1677
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.107456140350878e-06,
+      "loss": 1.2692,
+      "step": 1678
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.106907894736843e-06,
+      "loss": 0.9825,
+      "step": 1679
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.106359649122807e-06,
+      "loss": 0.9528,
+      "step": 1680
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.105811403508773e-06,
+      "loss": 0.4866,
+      "step": 1681
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.105263157894739e-06,
+      "loss": 0.7748,
+      "step": 1682
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.104714912280703e-06,
+      "loss": 1.3558,
+      "step": 1683
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.104166666666667e-06,
+      "loss": 0.8936,
+      "step": 1684
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.103618421052632e-06,
+      "loss": 1.0935,
+      "step": 1685
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.103070175438598e-06,
+      "loss": 0.9897,
+      "step": 1686
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.102521929824562e-06,
+      "loss": 1.5849,
+      "step": 1687
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.101973684210526e-06,
+      "loss": 0.6241,
+      "step": 1688
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.101425438596492e-06,
+      "loss": 1.3195,
+      "step": 1689
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.100877192982457e-06,
+      "loss": 1.131,
+      "step": 1690
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.100328947368421e-06,
+      "loss": 0.8792,
+      "step": 1691
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.099780701754387e-06,
+      "loss": 0.94,
+      "step": 1692
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.099232456140351e-06,
+      "loss": 1.523,
+      "step": 1693
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.098684210526317e-06,
+      "loss": 1.0464,
+      "step": 1694
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.09813596491228e-06,
+      "loss": 0.7898,
+      "step": 1695
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.097587719298246e-06,
+      "loss": 1.6305,
+      "step": 1696
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.097039473684212e-06,
+      "loss": 1.0737,
+      "step": 1697
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.096491228070178e-06,
+      "loss": 1.5529,
+      "step": 1698
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.095942982456142e-06,
+      "loss": 1.2245,
+      "step": 1699
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.095394736842105e-06,
+      "loss": 1.37,
+      "step": 1700
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.094846491228071e-06,
+      "loss": 1.4564,
+      "step": 1701
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.094298245614037e-06,
+      "loss": 0.4927,
+      "step": 1702
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.09375e-06,
+      "loss": 1.1257,
+      "step": 1703
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.093201754385965e-06,
+      "loss": 1.3784,
+      "step": 1704
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.09265350877193e-06,
+      "loss": 1.2112,
+      "step": 1705
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.092105263157896e-06,
+      "loss": 0.7852,
+      "step": 1706
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.09155701754386e-06,
+      "loss": 0.5313,
+      "step": 1707
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.091008771929826e-06,
+      "loss": 1.5335,
+      "step": 1708
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.09046052631579e-06,
+      "loss": 0.7435,
+      "step": 1709
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.089912280701755e-06,
+      "loss": 1.0108,
+      "step": 1710
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.08936403508772e-06,
+      "loss": 1.0628,
+      "step": 1711
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.088815789473685e-06,
+      "loss": 0.9209,
+      "step": 1712
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.08826754385965e-06,
+      "loss": 1.246,
+      "step": 1713
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.087719298245615e-06,
+      "loss": 0.8246,
+      "step": 1714
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.087171052631579e-06,
+      "loss": 1.0159,
+      "step": 1715
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.086622807017544e-06,
+      "loss": 1.6907,
+      "step": 1716
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.08607456140351e-06,
+      "loss": 1.3537,
+      "step": 1717
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.085526315789474e-06,
+      "loss": 0.7756,
+      "step": 1718
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.08497807017544e-06,
+      "loss": 0.7551,
+      "step": 1719
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.084429824561404e-06,
+      "loss": 0.9113,
+      "step": 1720
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.08388157894737e-06,
+      "loss": 1.4894,
+      "step": 1721
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.083333333333333e-06,
+      "loss": 0.8382,
+      "step": 1722
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.082785087719299e-06,
+      "loss": 1.5678,
+      "step": 1723
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.082236842105265e-06,
+      "loss": 1.2844,
+      "step": 1724
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.081688596491229e-06,
+      "loss": 0.9915,
+      "step": 1725
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.081140350877193e-06,
+      "loss": 1.2333,
+      "step": 1726
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.080592105263158e-06,
+      "loss": 1.9785,
+      "step": 1727
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.080043859649124e-06,
+      "loss": 0.9625,
+      "step": 1728
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.079495614035088e-06,
+      "loss": 0.8673,
+      "step": 1729
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.078947368421054e-06,
+      "loss": 1.0767,
+      "step": 1730
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.078399122807018e-06,
+      "loss": 0.7989,
+      "step": 1731
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.077850877192983e-06,
+      "loss": 1.3388,
+      "step": 1732
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.077302631578947e-06,
+      "loss": 0.8909,
+      "step": 1733
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.076754385964913e-06,
+      "loss": 1.4696,
+      "step": 1734
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.076206140350879e-06,
+      "loss": 1.3515,
+      "step": 1735
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.075657894736843e-06,
+      "loss": 1.3674,
+      "step": 1736
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.075109649122808e-06,
+      "loss": 1.5616,
+      "step": 1737
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.074561403508772e-06,
+      "loss": 0.649,
+      "step": 1738
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.074013157894738e-06,
+      "loss": 1.249,
+      "step": 1739
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.073464912280704e-06,
+      "loss": 0.9698,
+      "step": 1740
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.072916666666668e-06,
+      "loss": 0.6137,
+      "step": 1741
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.072368421052632e-06,
+      "loss": 0.7939,
+      "step": 1742
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.071820175438597e-06,
+      "loss": 1.3192,
+      "step": 1743
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.071271929824563e-06,
+      "loss": 1.235,
+      "step": 1744
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.070723684210527e-06,
+      "loss": 0.9064,
+      "step": 1745
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.070175438596493e-06,
+      "loss": 0.9106,
+      "step": 1746
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.069627192982456e-06,
+      "loss": 1.4984,
+      "step": 1747
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.069078947368422e-06,
+      "loss": 0.8018,
+      "step": 1748
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.068530701754386e-06,
+      "loss": 0.9441,
+      "step": 1749
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.067982456140352e-06,
+      "loss": 1.4659,
+      "step": 1750
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.067434210526317e-06,
+      "loss": 0.8842,
+      "step": 1751
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.066885964912281e-06,
+      "loss": 1.81,
+      "step": 1752
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.066337719298245e-06,
+      "loss": 1.1052,
+      "step": 1753
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.065789473684211e-06,
+      "loss": 1.4664,
+      "step": 1754
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.065241228070177e-06,
+      "loss": 0.7072,
+      "step": 1755
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.06469298245614e-06,
+      "loss": 1.1795,
+      "step": 1756
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.064144736842106e-06,
+      "loss": 0.9044,
+      "step": 1757
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.06359649122807e-06,
+      "loss": 0.8085,
+      "step": 1758
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.063048245614036e-06,
+      "loss": 0.7999,
+      "step": 1759
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.0625e-06,
+      "loss": 1.2189,
+      "step": 1760
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.061951754385966e-06,
+      "loss": 1.0734,
+      "step": 1761
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.061403508771931e-06,
+      "loss": 1.0199,
+      "step": 1762
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.060855263157895e-06,
+      "loss": 1.0708,
+      "step": 1763
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.06030701754386e-06,
+      "loss": 0.8939,
+      "step": 1764
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.059758771929825e-06,
+      "loss": 1.0363,
+      "step": 1765
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.05921052631579e-06,
+      "loss": 0.8395,
+      "step": 1766
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.058662280701755e-06,
+      "loss": 0.9102,
+      "step": 1767
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.05811403508772e-06,
+      "loss": 1.393,
+      "step": 1768
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.057565789473684e-06,
+      "loss": 1.6285,
+      "step": 1769
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.05701754385965e-06,
+      "loss": 0.9501,
+      "step": 1770
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.056469298245616e-06,
+      "loss": 1.35,
+      "step": 1771
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.05592105263158e-06,
+      "loss": 1.2363,
+      "step": 1772
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.055372807017545e-06,
+      "loss": 1.6142,
+      "step": 1773
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.05482456140351e-06,
+      "loss": 1.525,
+      "step": 1774
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.054276315789475e-06,
+      "loss": 0.7046,
+      "step": 1775
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.053728070175439e-06,
+      "loss": 1.1281,
+      "step": 1776
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.053179824561405e-06,
+      "loss": 1.0983,
+      "step": 1777
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.05263157894737e-06,
+      "loss": 1.0016,
+      "step": 1778
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.052083333333334e-06,
+      "loss": 1.0649,
+      "step": 1779
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.051535087719298e-06,
+      "loss": 1.1636,
+      "step": 1780
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.050986842105264e-06,
+      "loss": 1.6836,
+      "step": 1781
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.05043859649123e-06,
+      "loss": 1.1432,
+      "step": 1782
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.049890350877194e-06,
+      "loss": 1.2513,
+      "step": 1783
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.04934210526316e-06,
+      "loss": 1.2846,
+      "step": 1784
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.048793859649123e-06,
+      "loss": 0.8352,
+      "step": 1785
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.048245614035089e-06,
+      "loss": 1.3564,
+      "step": 1786
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.047697368421053e-06,
+      "loss": 1.7311,
+      "step": 1787
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.047149122807019e-06,
+      "loss": 1.3057,
+      "step": 1788
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.046600877192984e-06,
+      "loss": 1.2539,
+      "step": 1789
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.046052631578948e-06,
+      "loss": 1.276,
+      "step": 1790
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.045504385964912e-06,
+      "loss": 1.0369,
+      "step": 1791
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.044956140350878e-06,
+      "loss": 1.1978,
+      "step": 1792
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.044407894736844e-06,
+      "loss": 1.4601,
+      "step": 1793
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.043859649122807e-06,
+      "loss": 1.4993,
+      "step": 1794
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.043311403508771e-06,
+      "loss": 0.772,
+      "step": 1795
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.042763157894737e-06,
+      "loss": 1.1639,
+      "step": 1796
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.042214912280703e-06,
+      "loss": 1.2757,
+      "step": 1797
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.041666666666667e-06,
+      "loss": 1.0998,
+      "step": 1798
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.041118421052632e-06,
+      "loss": 1.1394,
+      "step": 1799
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.040570175438598e-06,
+      "loss": 0.7694,
+      "step": 1800
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.040021929824562e-06,
+      "loss": 1.2019,
+      "step": 1801
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.039473684210526e-06,
+      "loss": 1.1108,
+      "step": 1802
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.038925438596492e-06,
+      "loss": 0.5557,
+      "step": 1803
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.038377192982457e-06,
+      "loss": 1.2899,
+      "step": 1804
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.037828947368421e-06,
+      "loss": 0.8236,
+      "step": 1805
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.037280701754385e-06,
+      "loss": 1.1356,
+      "step": 1806
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.036732456140351e-06,
+      "loss": 1.0692,
+      "step": 1807
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.036184210526317e-06,
+      "loss": 0.7961,
+      "step": 1808
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.035635964912282e-06,
+      "loss": 0.8024,
+      "step": 1809
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.035087719298246e-06,
+      "loss": 1.0171,
+      "step": 1810
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.034539473684212e-06,
+      "loss": 0.9045,
+      "step": 1811
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.033991228070176e-06,
+      "loss": 1.0075,
+      "step": 1812
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.033442982456142e-06,
+      "loss": 0.8509,
+      "step": 1813
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.032894736842106e-06,
+      "loss": 1.5342,
+      "step": 1814
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.032346491228071e-06,
+      "loss": 1.4354,
+      "step": 1815
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.031798245614037e-06,
+      "loss": 0.9932,
+      "step": 1816
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.031250000000001e-06,
+      "loss": 0.9368,
+      "step": 1817
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.030701754385965e-06,
+      "loss": 1.1057,
+      "step": 1818
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.03015350877193e-06,
+      "loss": 0.6215,
+      "step": 1819
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.029605263157896e-06,
+      "loss": 1.3892,
+      "step": 1820
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.02905701754386e-06,
+      "loss": 1.1172,
+      "step": 1821
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.028508771929824e-06,
+      "loss": 1.1875,
+      "step": 1822
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.02796052631579e-06,
+      "loss": 0.5062,
+      "step": 1823
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.027412280701756e-06,
+      "loss": 1.3689,
+      "step": 1824
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.02686403508772e-06,
+      "loss": 1.1639,
+      "step": 1825
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.026315789473685e-06,
+      "loss": 0.9245,
+      "step": 1826
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.025767543859651e-06,
+      "loss": 0.8712,
+      "step": 1827
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.025219298245615e-06,
+      "loss": 1.1836,
+      "step": 1828
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.024671052631579e-06,
+      "loss": 1.3288,
+      "step": 1829
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.5119950789419725,
+      "eval_loss": 1.0587667226791382,
+      "eval_roc_auc": 0.8511601943962093,
+      "eval_runtime": 178.8455,
+      "eval_samples_per_second": 27.269,
+      "eval_steps_per_second": 3.411,
+      "step": 1829
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.024122807017545e-06,
+      "loss": 0.8575,
+      "step": 1830
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.02357456140351e-06,
+      "loss": 0.9134,
+      "step": 1831
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.023026315789474e-06,
+      "loss": 1.092,
+      "step": 1832
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.022478070175438e-06,
+      "loss": 1.2473,
+      "step": 1833
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.021929824561404e-06,
+      "loss": 1.0659,
+      "step": 1834
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.02138157894737e-06,
+      "loss": 0.8612,
+      "step": 1835
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.020833333333334e-06,
+      "loss": 0.8058,
+      "step": 1836
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.0202850877193e-06,
+      "loss": 0.821,
+      "step": 1837
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.019736842105263e-06,
+      "loss": 1.5414,
+      "step": 1838
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.019188596491229e-06,
+      "loss": 0.8528,
+      "step": 1839
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.018640350877193e-06,
+      "loss": 0.7917,
+      "step": 1840
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.018092105263158e-06,
+      "loss": 1.8224,
+      "step": 1841
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.017543859649124e-06,
+      "loss": 0.8615,
+      "step": 1842
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.016995614035088e-06,
+      "loss": 0.8595,
+      "step": 1843
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.016447368421054e-06,
+      "loss": 0.9776,
+      "step": 1844
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.015899122807018e-06,
+      "loss": 1.0276,
+      "step": 1845
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.015350877192983e-06,
+      "loss": 1.5239,
+      "step": 1846
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.014802631578949e-06,
+      "loss": 1.2284,
+      "step": 1847
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.014254385964913e-06,
+      "loss": 0.87,
+      "step": 1848
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.013706140350877e-06,
+      "loss": 0.83,
+      "step": 1849
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.013157894736843e-06,
+      "loss": 1.4625,
+      "step": 1850
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.012609649122808e-06,
+      "loss": 1.1521,
+      "step": 1851
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.012061403508772e-06,
+      "loss": 0.7745,
+      "step": 1852
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.011513157894738e-06,
+      "loss": 1.4122,
+      "step": 1853
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.010964912280704e-06,
+      "loss": 1.6638,
+      "step": 1854
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.010416666666668e-06,
+      "loss": 1.211,
+      "step": 1855
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.009868421052632e-06,
+      "loss": 1.0564,
+      "step": 1856
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.009320175438597e-06,
+      "loss": 1.1144,
+      "step": 1857
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.008771929824563e-06,
+      "loss": 1.3321,
+      "step": 1858
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.008223684210527e-06,
+      "loss": 1.7961,
+      "step": 1859
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.007675438596491e-06,
+      "loss": 0.9753,
+      "step": 1860
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.007127192982457e-06,
+      "loss": 1.3416,
+      "step": 1861
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.006578947368422e-06,
+      "loss": 1.0358,
+      "step": 1862
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.006030701754386e-06,
+      "loss": 0.824,
+      "step": 1863
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.005482456140352e-06,
+      "loss": 1.0699,
+      "step": 1864
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.004934210526316e-06,
+      "loss": 0.8896,
+      "step": 1865
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.004385964912282e-06,
+      "loss": 0.7502,
+      "step": 1866
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.003837719298246e-06,
+      "loss": 0.7732,
+      "step": 1867
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.003289473684211e-06,
+      "loss": 1.2282,
+      "step": 1868
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.002741228070177e-06,
+      "loss": 0.5087,
+      "step": 1869
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.002192982456141e-06,
+      "loss": 1.2026,
+      "step": 1870
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.001644736842105e-06,
+      "loss": 1.0353,
+      "step": 1871
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.00109649122807e-06,
+      "loss": 1.3795,
+      "step": 1872
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.000548245614036e-06,
+      "loss": 1.0591,
+      "step": 1873
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9e-06,
+      "loss": 1.2153,
+      "step": 1874
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.999451754385966e-06,
+      "loss": 1.4231,
+      "step": 1875
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.99890350877193e-06,
+      "loss": 0.7053,
+      "step": 1876
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.998355263157896e-06,
+      "loss": 1.133,
+      "step": 1877
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.99780701754386e-06,
+      "loss": 1.5058,
+      "step": 1878
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.997258771929825e-06,
+      "loss": 0.9361,
+      "step": 1879
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.996710526315791e-06,
+      "loss": 0.5991,
+      "step": 1880
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.996162280701755e-06,
+      "loss": 0.7526,
+      "step": 1881
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.99561403508772e-06,
+      "loss": 1.1814,
+      "step": 1882
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.995065789473684e-06,
+      "loss": 0.8114,
+      "step": 1883
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.99451754385965e-06,
+      "loss": 1.3025,
+      "step": 1884
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.993969298245616e-06,
+      "loss": 0.5458,
+      "step": 1885
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.99342105263158e-06,
+      "loss": 1.3894,
+      "step": 1886
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.992872807017544e-06,
+      "loss": 0.8035,
+      "step": 1887
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.99232456140351e-06,
+      "loss": 1.1094,
+      "step": 1888
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.991776315789475e-06,
+      "loss": 0.8263,
+      "step": 1889
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.991228070175439e-06,
+      "loss": 1.1074,
+      "step": 1890
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.990679824561405e-06,
+      "loss": 1.0586,
+      "step": 1891
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.990131578947369e-06,
+      "loss": 1.0636,
+      "step": 1892
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.989583333333334e-06,
+      "loss": 1.2361,
+      "step": 1893
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.989035087719298e-06,
+      "loss": 1.1111,
+      "step": 1894
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.988486842105264e-06,
+      "loss": 0.8889,
+      "step": 1895
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.98793859649123e-06,
+      "loss": 0.6025,
+      "step": 1896
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.987390350877194e-06,
+      "loss": 1.4875,
+      "step": 1897
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.986842105263158e-06,
+      "loss": 0.7322,
+      "step": 1898
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.986293859649123e-06,
+      "loss": 0.7394,
+      "step": 1899
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.985745614035089e-06,
+      "loss": 1.313,
+      "step": 1900
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.985197368421053e-06,
+      "loss": 1.6499,
+      "step": 1901
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.984649122807019e-06,
+      "loss": 1.1271,
+      "step": 1902
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.984100877192983e-06,
+      "loss": 1.7754,
+      "step": 1903
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.983552631578948e-06,
+      "loss": 0.6995,
+      "step": 1904
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.983004385964912e-06,
+      "loss": 1.2296,
+      "step": 1905
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.982456140350878e-06,
+      "loss": 1.4573,
+      "step": 1906
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.981907894736844e-06,
+      "loss": 1.4902,
+      "step": 1907
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.981359649122808e-06,
+      "loss": 0.9554,
+      "step": 1908
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.980811403508772e-06,
+      "loss": 1.0531,
+      "step": 1909
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.980263157894737e-06,
+      "loss": 1.0823,
+      "step": 1910
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.979714912280703e-06,
+      "loss": 1.0513,
+      "step": 1911
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.979166666666667e-06,
+      "loss": 1.2884,
+      "step": 1912
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.978618421052633e-06,
+      "loss": 1.7998,
+      "step": 1913
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.978070175438597e-06,
+      "loss": 0.9997,
+      "step": 1914
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.977521929824562e-06,
+      "loss": 0.7041,
+      "step": 1915
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.976973684210526e-06,
+      "loss": 0.8685,
+      "step": 1916
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.976425438596492e-06,
+      "loss": 1.0615,
+      "step": 1917
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.975877192982458e-06,
+      "loss": 0.7616,
+      "step": 1918
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.975328947368422e-06,
+      "loss": 1.2075,
+      "step": 1919
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.974780701754387e-06,
+      "loss": 0.8129,
+      "step": 1920
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.974232456140351e-06,
+      "loss": 0.7222,
+      "step": 1921
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.973684210526317e-06,
+      "loss": 1.1344,
+      "step": 1922
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.973135964912283e-06,
+      "loss": 0.9691,
+      "step": 1923
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.972587719298247e-06,
+      "loss": 1.1781,
+      "step": 1924
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.97203947368421e-06,
+      "loss": 1.0749,
+      "step": 1925
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.971491228070176e-06,
+      "loss": 1.0963,
+      "step": 1926
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.970942982456142e-06,
+      "loss": 1.0852,
+      "step": 1927
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.970394736842106e-06,
+      "loss": 0.5146,
+      "step": 1928
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.969846491228072e-06,
+      "loss": 0.9596,
+      "step": 1929
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.969298245614035e-06,
+      "loss": 0.9227,
+      "step": 1930
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.968750000000001e-06,
+      "loss": 1.6009,
+      "step": 1931
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.968201754385965e-06,
+      "loss": 1.2921,
+      "step": 1932
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.96765350877193e-06,
+      "loss": 1.3047,
+      "step": 1933
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.967105263157896e-06,
+      "loss": 1.5258,
+      "step": 1934
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.96655701754386e-06,
+      "loss": 1.3496,
+      "step": 1935
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.966008771929824e-06,
+      "loss": 1.138,
+      "step": 1936
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.96546052631579e-06,
+      "loss": 1.102,
+      "step": 1937
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.964912280701756e-06,
+      "loss": 0.9769,
+      "step": 1938
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.96436403508772e-06,
+      "loss": 1.2168,
+      "step": 1939
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.963815789473685e-06,
+      "loss": 0.9074,
+      "step": 1940
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.96326754385965e-06,
+      "loss": 1.26,
+      "step": 1941
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.962719298245615e-06,
+      "loss": 0.4925,
+      "step": 1942
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.962171052631579e-06,
+      "loss": 0.7341,
+      "step": 1943
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.961622807017545e-06,
+      "loss": 0.8174,
+      "step": 1944
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.96107456140351e-06,
+      "loss": 1.0426,
+      "step": 1945
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.960526315789474e-06,
+      "loss": 0.8409,
+      "step": 1946
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.959978070175438e-06,
+      "loss": 0.4562,
+      "step": 1947
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.959429824561404e-06,
+      "loss": 1.2465,
+      "step": 1948
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.95888157894737e-06,
+      "loss": 0.9296,
+      "step": 1949
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.958333333333334e-06,
+      "loss": 1.2033,
+      "step": 1950
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.957785087719298e-06,
+      "loss": 0.9815,
+      "step": 1951
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.957236842105263e-06,
+      "loss": 1.2367,
+      "step": 1952
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.956688596491229e-06,
+      "loss": 0.9049,
+      "step": 1953
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.956140350877193e-06,
+      "loss": 0.7561,
+      "step": 1954
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.955592105263159e-06,
+      "loss": 1.1886,
+      "step": 1955
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.955043859649124e-06,
+      "loss": 1.175,
+      "step": 1956
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.954495614035088e-06,
+      "loss": 1.2065,
+      "step": 1957
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.953947368421054e-06,
+      "loss": 1.2096,
+      "step": 1958
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.953399122807018e-06,
+      "loss": 1.4031,
+      "step": 1959
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.952850877192984e-06,
+      "loss": 1.0539,
+      "step": 1960
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.95230263157895e-06,
+      "loss": 1.0429,
+      "step": 1961
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.951754385964913e-06,
+      "loss": 1.2191,
+      "step": 1962
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.951206140350877e-06,
+      "loss": 1.0756,
+      "step": 1963
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.950657894736843e-06,
+      "loss": 1.1756,
+      "step": 1964
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.950109649122809e-06,
+      "loss": 1.4092,
+      "step": 1965
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.949561403508773e-06,
+      "loss": 0.9697,
+      "step": 1966
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.949013157894738e-06,
+      "loss": 0.6351,
+      "step": 1967
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.948464912280702e-06,
+      "loss": 1.8731,
+      "step": 1968
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.947916666666668e-06,
+      "loss": 0.9962,
+      "step": 1969
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.947368421052632e-06,
+      "loss": 1.452,
+      "step": 1970
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.946820175438598e-06,
+      "loss": 1.3015,
+      "step": 1971
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.946271929824563e-06,
+      "loss": 1.4387,
+      "step": 1972
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.945723684210527e-06,
+      "loss": 0.8608,
+      "step": 1973
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.945175438596491e-06,
+      "loss": 1.1376,
+      "step": 1974
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.944627192982457e-06,
+      "loss": 0.8839,
+      "step": 1975
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.944078947368423e-06,
+      "loss": 0.887,
+      "step": 1976
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.943530701754386e-06,
+      "loss": 0.8307,
+      "step": 1977
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.94298245614035e-06,
+      "loss": 1.2046,
+      "step": 1978
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.942434210526316e-06,
+      "loss": 0.6381,
+      "step": 1979
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.941885964912282e-06,
+      "loss": 1.381,
+      "step": 1980
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.941337719298246e-06,
+      "loss": 0.6128,
+      "step": 1981
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.940789473684211e-06,
+      "loss": 0.8931,
+      "step": 1982
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.940241228070177e-06,
+      "loss": 1.0019,
+      "step": 1983
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.939692982456141e-06,
+      "loss": 1.5579,
+      "step": 1984
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.939144736842105e-06,
+      "loss": 0.7379,
+      "step": 1985
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.93859649122807e-06,
+      "loss": 1.1349,
+      "step": 1986
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.938048245614036e-06,
+      "loss": 1.5454,
+      "step": 1987
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.9375e-06,
+      "loss": 0.5741,
+      "step": 1988
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.936951754385964e-06,
+      "loss": 0.9967,
+      "step": 1989
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.93640350877193e-06,
+      "loss": 1.2454,
+      "step": 1990
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.935855263157896e-06,
+      "loss": 1.0011,
+      "step": 1991
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.93530701754386e-06,
+      "loss": 1.5273,
+      "step": 1992
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.934758771929825e-06,
+      "loss": 0.5082,
+      "step": 1993
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.93421052631579e-06,
+      "loss": 1.0927,
+      "step": 1994
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.933662280701755e-06,
+      "loss": 1.1702,
+      "step": 1995
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.93311403508772e-06,
+      "loss": 0.4545,
+      "step": 1996
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.932565789473685e-06,
+      "loss": 1.675,
+      "step": 1997
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.93201754385965e-06,
+      "loss": 1.0462,
+      "step": 1998
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.931469298245616e-06,
+      "loss": 0.6184,
+      "step": 1999
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.93092105263158e-06,
+      "loss": 0.8372,
+      "step": 2000
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.930372807017544e-06,
+      "loss": 0.7332,
+      "step": 2001
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.92982456140351e-06,
+      "loss": 1.4752,
+      "step": 2002
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.929276315789475e-06,
+      "loss": 0.9738,
+      "step": 2003
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.92872807017544e-06,
+      "loss": 0.8071,
+      "step": 2004
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.928179824561403e-06,
+      "loss": 1.1429,
+      "step": 2005
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.927631578947369e-06,
+      "loss": 1.1236,
+      "step": 2006
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.927083333333335e-06,
+      "loss": 1.1867,
+      "step": 2007
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.926535087719299e-06,
+      "loss": 1.2171,
+      "step": 2008
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.925986842105264e-06,
+      "loss": 1.1127,
+      "step": 2009
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.92543859649123e-06,
+      "loss": 0.8227,
+      "step": 2010
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.924890350877194e-06,
+      "loss": 0.7978,
+      "step": 2011
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.924342105263158e-06,
+      "loss": 1.4253,
+      "step": 2012
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.923793859649124e-06,
+      "loss": 1.1509,
+      "step": 2013
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.92324561403509e-06,
+      "loss": 1.0951,
+      "step": 2014
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.922697368421053e-06,
+      "loss": 1.093,
+      "step": 2015
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.922149122807017e-06,
+      "loss": 1.0576,
+      "step": 2016
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.921600877192983e-06,
+      "loss": 0.8122,
+      "step": 2017
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.921052631578949e-06,
+      "loss": 1.1972,
+      "step": 2018
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.920504385964913e-06,
+      "loss": 0.56,
+      "step": 2019
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.919956140350878e-06,
+      "loss": 0.7273,
+      "step": 2020
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.919407894736842e-06,
+      "loss": 1.7406,
+      "step": 2021
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.918859649122808e-06,
+      "loss": 0.7184,
+      "step": 2022
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.918311403508772e-06,
+      "loss": 0.5468,
+      "step": 2023
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.917763157894737e-06,
+      "loss": 1.0071,
+      "step": 2024
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.917214912280703e-06,
+      "loss": 0.7632,
+      "step": 2025
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.916666666666667e-06,
+      "loss": 1.7089,
+      "step": 2026
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.916118421052631e-06,
+      "loss": 1.7319,
+      "step": 2027
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.915570175438597e-06,
+      "loss": 0.7886,
+      "step": 2028
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.915021929824562e-06,
+      "loss": 1.2629,
+      "step": 2029
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.914473684210528e-06,
+      "loss": 1.0315,
+      "step": 2030
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.913925438596492e-06,
+      "loss": 1.0908,
+      "step": 2031
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.913377192982456e-06,
+      "loss": 1.4261,
+      "step": 2032
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.912828947368422e-06,
+      "loss": 1.0538,
+      "step": 2033
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.912280701754387e-06,
+      "loss": 1.9198,
+      "step": 2034
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.911732456140351e-06,
+      "loss": 0.7512,
+      "step": 2035
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.911184210526317e-06,
+      "loss": 0.6696,
+      "step": 2036
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.910635964912281e-06,
+      "loss": 0.8258,
+      "step": 2037
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.910087719298247e-06,
+      "loss": 1.3667,
+      "step": 2038
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.90953947368421e-06,
+      "loss": 1.2355,
+      "step": 2039
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.908991228070176e-06,
+      "loss": 0.7781,
+      "step": 2040
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.908442982456142e-06,
+      "loss": 1.003,
+      "step": 2041
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.907894736842106e-06,
+      "loss": 1.7645,
+      "step": 2042
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.90734649122807e-06,
+      "loss": 1.0891,
+      "step": 2043
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.906798245614036e-06,
+      "loss": 1.4278,
+      "step": 2044
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.906250000000001e-06,
+      "loss": 1.2579,
+      "step": 2045
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.905701754385965e-06,
+      "loss": 0.9229,
+      "step": 2046
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.905153508771931e-06,
+      "loss": 0.7665,
+      "step": 2047
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.904605263157895e-06,
+      "loss": 1.487,
+      "step": 2048
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.90405701754386e-06,
+      "loss": 0.8356,
+      "step": 2049
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.903508771929825e-06,
+      "loss": 1.1462,
+      "step": 2050
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.90296052631579e-06,
+      "loss": 0.9049,
+      "step": 2051
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.902412280701756e-06,
+      "loss": 1.1498,
+      "step": 2052
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.90186403508772e-06,
+      "loss": 0.5963,
+      "step": 2053
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.901315789473684e-06,
+      "loss": 1.189,
+      "step": 2054
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.90076754385965e-06,
+      "loss": 0.7342,
+      "step": 2055
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.900219298245615e-06,
+      "loss": 1.2389,
+      "step": 2056
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.89967105263158e-06,
+      "loss": 0.8058,
+      "step": 2057
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.899122807017545e-06,
+      "loss": 1.3554,
+      "step": 2058
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.898574561403509e-06,
+      "loss": 1.3335,
+      "step": 2059
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.898026315789475e-06,
+      "loss": 1.5089,
+      "step": 2060
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.897478070175439e-06,
+      "loss": 0.7603,
+      "step": 2061
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.896929824561404e-06,
+      "loss": 0.5765,
+      "step": 2062
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.89638157894737e-06,
+      "loss": 0.8149,
+      "step": 2063
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.895833333333334e-06,
+      "loss": 0.7444,
+      "step": 2064
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.895285087719298e-06,
+      "loss": 1.2883,
+      "step": 2065
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.894736842105264e-06,
+      "loss": 0.6093,
+      "step": 2066
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.89418859649123e-06,
+      "loss": 1.3954,
+      "step": 2067
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.893640350877195e-06,
+      "loss": 0.7832,
+      "step": 2068
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.893092105263159e-06,
+      "loss": 1.4425,
+      "step": 2069
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.892543859649123e-06,
+      "loss": 0.6804,
+      "step": 2070
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.891995614035088e-06,
+      "loss": 1.0513,
+      "step": 2071
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.891447368421054e-06,
+      "loss": 1.3095,
+      "step": 2072
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.890899122807018e-06,
+      "loss": 0.6047,
+      "step": 2073
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.890350877192984e-06,
+      "loss": 1.373,
+      "step": 2074
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.889802631578948e-06,
+      "loss": 1.2036,
+      "step": 2075
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.889254385964913e-06,
+      "loss": 1.0412,
+      "step": 2076
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.888706140350877e-06,
+      "loss": 1.0863,
+      "step": 2077
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.888157894736843e-06,
+      "loss": 0.6102,
+      "step": 2078
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.887609649122809e-06,
+      "loss": 0.4264,
+      "step": 2079
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.887061403508773e-06,
+      "loss": 0.5066,
+      "step": 2080
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.886513157894737e-06,
+      "loss": 1.2189,
+      "step": 2081
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.885964912280702e-06,
+      "loss": 0.8331,
+      "step": 2082
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.885416666666668e-06,
+      "loss": 0.8968,
+      "step": 2083
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.884868421052632e-06,
+      "loss": 1.3628,
+      "step": 2084
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.884320175438598e-06,
+      "loss": 0.8385,
+      "step": 2085
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.883771929824562e-06,
+      "loss": 1.054,
+      "step": 2086
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.883223684210527e-06,
+      "loss": 1.4575,
+      "step": 2087
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.882675438596491e-06,
+      "loss": 0.9653,
+      "step": 2088
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.882127192982457e-06,
+      "loss": 1.0348,
+      "step": 2089
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.881578947368423e-06,
+      "loss": 1.0786,
+      "step": 2090
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.881030701754387e-06,
+      "loss": 0.8083,
+      "step": 2091
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.88048245614035e-06,
+      "loss": 1.3412,
+      "step": 2092
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.879934210526316e-06,
+      "loss": 0.8135,
+      "step": 2093
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.879385964912282e-06,
+      "loss": 0.6207,
+      "step": 2094
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.878837719298246e-06,
+      "loss": 1.4088,
+      "step": 2095
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.878289473684212e-06,
+      "loss": 0.9275,
+      "step": 2096
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.877741228070176e-06,
+      "loss": 0.766,
+      "step": 2097
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.877192982456141e-06,
+      "loss": 1.3101,
+      "step": 2098
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.876644736842105e-06,
+      "loss": 0.9659,
+      "step": 2099
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.876096491228071e-06,
+      "loss": 1.3765,
+      "step": 2100
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.875548245614037e-06,
+      "loss": 1.2906,
+      "step": 2101
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.875e-06,
+      "loss": 1.3038,
+      "step": 2102
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.874451754385965e-06,
+      "loss": 1.1433,
+      "step": 2103
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.87390350877193e-06,
+      "loss": 1.1599,
+      "step": 2104
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.873355263157896e-06,
+      "loss": 1.4355,
+      "step": 2105
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.872807017543862e-06,
+      "loss": 1.4659,
+      "step": 2106
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.872258771929826e-06,
+      "loss": 1.0222,
+      "step": 2107
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.87171052631579e-06,
+      "loss": 1.5913,
+      "step": 2108
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.871162280701755e-06,
+      "loss": 0.5426,
+      "step": 2109
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.870614035087721e-06,
+      "loss": 1.3981,
+      "step": 2110
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.870065789473685e-06,
+      "loss": 0.6864,
+      "step": 2111
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.86951754385965e-06,
+      "loss": 0.7439,
+      "step": 2112
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.868969298245614e-06,
+      "loss": 1.5382,
+      "step": 2113
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.86842105263158e-06,
+      "loss": 0.6886,
+      "step": 2114
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.867872807017544e-06,
+      "loss": 0.7564,
+      "step": 2115
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.86732456140351e-06,
+      "loss": 1.1015,
+      "step": 2116
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.866776315789475e-06,
+      "loss": 1.1744,
+      "step": 2117
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.86622807017544e-06,
+      "loss": 1.0702,
+      "step": 2118
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.865679824561403e-06,
+      "loss": 1.5514,
+      "step": 2119
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.865131578947369e-06,
+      "loss": 2.0869,
+      "step": 2120
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.864583333333335e-06,
+      "loss": 1.6948,
+      "step": 2121
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.864035087719299e-06,
+      "loss": 1.2919,
+      "step": 2122
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.863486842105263e-06,
+      "loss": 0.805,
+      "step": 2123
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.862938596491228e-06,
+      "loss": 0.7303,
+      "step": 2124
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.862390350877194e-06,
+      "loss": 1.0344,
+      "step": 2125
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.861842105263158e-06,
+      "loss": 1.0225,
+      "step": 2126
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.861293859649124e-06,
+      "loss": 1.2157,
+      "step": 2127
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.86074561403509e-06,
+      "loss": 1.181,
+      "step": 2128
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.860197368421053e-06,
+      "loss": 1.1232,
+      "step": 2129
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.859649122807017e-06,
+      "loss": 1.6224,
+      "step": 2130
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.859100877192983e-06,
+      "loss": 1.5047,
+      "step": 2131
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.858552631578949e-06,
+      "loss": 0.9921,
+      "step": 2132
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.858004385964913e-06,
+      "loss": 1.3438,
+      "step": 2133
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.857456140350877e-06,
+      "loss": 1.4372,
+      "step": 2134
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.856907894736842e-06,
+      "loss": 0.8583,
+      "step": 2135
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.856359649122808e-06,
+      "loss": 1.2802,
+      "step": 2136
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.855811403508772e-06,
+      "loss": 1.3571,
+      "step": 2137
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.855263157894738e-06,
+      "loss": 0.9203,
+      "step": 2138
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.854714912280703e-06,
+      "loss": 1.3051,
+      "step": 2139
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.854166666666667e-06,
+      "loss": 1.5454,
+      "step": 2140
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.853618421052633e-06,
+      "loss": 1.1933,
+      "step": 2141
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.853070175438597e-06,
+      "loss": 1.0182,
+      "step": 2142
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.852521929824563e-06,
+      "loss": 0.7749,
+      "step": 2143
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.851973684210528e-06,
+      "loss": 1.0548,
+      "step": 2144
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.851425438596492e-06,
+      "loss": 1.6949,
+      "step": 2145
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.850877192982456e-06,
+      "loss": 1.2086,
+      "step": 2146
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.850328947368422e-06,
+      "loss": 1.1462,
+      "step": 2147
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.849780701754388e-06,
+      "loss": 1.3481,
+      "step": 2148
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.849232456140352e-06,
+      "loss": 1.5043,
+      "step": 2149
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.848684210526316e-06,
+      "loss": 0.6835,
+      "step": 2150
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.848135964912281e-06,
+      "loss": 1.3504,
+      "step": 2151
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.847587719298247e-06,
+      "loss": 0.4576,
+      "step": 2152
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.847039473684211e-06,
+      "loss": 1.0188,
+      "step": 2153
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.846491228070177e-06,
+      "loss": 1.8624,
+      "step": 2154
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.845942982456142e-06,
+      "loss": 0.7835,
+      "step": 2155
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.845394736842106e-06,
+      "loss": 0.8692,
+      "step": 2156
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.84484649122807e-06,
+      "loss": 1.2391,
+      "step": 2157
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.844298245614036e-06,
+      "loss": 1.3683,
+      "step": 2158
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.843750000000002e-06,
+      "loss": 0.8928,
+      "step": 2159
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.843201754385965e-06,
+      "loss": 1.0992,
+      "step": 2160
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.84265350877193e-06,
+      "loss": 0.6607,
+      "step": 2161
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.842105263157895e-06,
+      "loss": 1.1327,
+      "step": 2162
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.84155701754386e-06,
+      "loss": 0.8152,
+      "step": 2163
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.841008771929825e-06,
+      "loss": 0.7193,
+      "step": 2164
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.84046052631579e-06,
+      "loss": 0.8593,
+      "step": 2165
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.839912280701754e-06,
+      "loss": 0.8416,
+      "step": 2166
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.83936403508772e-06,
+      "loss": 1.3159,
+      "step": 2167
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.838815789473684e-06,
+      "loss": 0.7371,
+      "step": 2168
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.83826754385965e-06,
+      "loss": 1.6432,
+      "step": 2169
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.837719298245615e-06,
+      "loss": 0.8019,
+      "step": 2170
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.83717105263158e-06,
+      "loss": 1.3652,
+      "step": 2171
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.836622807017543e-06,
+      "loss": 0.9849,
+      "step": 2172
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.836074561403509e-06,
+      "loss": 1.8683,
+      "step": 2173
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.835526315789475e-06,
+      "loss": 0.7841,
+      "step": 2174
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.834978070175439e-06,
+      "loss": 1.0864,
+      "step": 2175
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.834429824561404e-06,
+      "loss": 0.6537,
+      "step": 2176
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.833881578947368e-06,
+      "loss": 1.9503,
+      "step": 2177
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.833333333333334e-06,
+      "loss": 1.5391,
+      "step": 2178
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.8327850877193e-06,
+      "loss": 1.7156,
+      "step": 2179
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.832236842105264e-06,
+      "loss": 1.1271,
+      "step": 2180
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.83168859649123e-06,
+      "loss": 0.8828,
+      "step": 2181
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.831140350877195e-06,
+      "loss": 0.6207,
+      "step": 2182
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.830592105263159e-06,
+      "loss": 0.7832,
+      "step": 2183
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.830043859649123e-06,
+      "loss": 1.0614,
+      "step": 2184
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.829495614035089e-06,
+      "loss": 1.2123,
+      "step": 2185
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.828947368421054e-06,
+      "loss": 1.4333,
+      "step": 2186
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.828399122807018e-06,
+      "loss": 0.5663,
+      "step": 2187
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.827850877192982e-06,
+      "loss": 1.8352,
+      "step": 2188
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.827302631578948e-06,
+      "loss": 1.0377,
+      "step": 2189
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.826754385964914e-06,
+      "loss": 0.8341,
+      "step": 2190
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.826206140350878e-06,
+      "loss": 1.0219,
+      "step": 2191
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.825657894736843e-06,
+      "loss": 1.2093,
+      "step": 2192
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.825109649122807e-06,
+      "loss": 1.1872,
+      "step": 2193
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.824561403508773e-06,
+      "loss": 1.2393,
+      "step": 2194
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.824013157894737e-06,
+      "loss": 0.7309,
+      "step": 2195
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.823464912280703e-06,
+      "loss": 0.8132,
+      "step": 2196
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.822916666666668e-06,
+      "loss": 1.1028,
+      "step": 2197
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.822368421052632e-06,
+      "loss": 1.0147,
+      "step": 2198
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.821820175438596e-06,
+      "loss": 0.6442,
+      "step": 2199
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.821271929824562e-06,
+      "loss": 1.0869,
+      "step": 2200
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.820723684210528e-06,
+      "loss": 1.5011,
+      "step": 2201
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.820175438596492e-06,
+      "loss": 1.1665,
+      "step": 2202
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.819627192982457e-06,
+      "loss": 0.7187,
+      "step": 2203
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.819078947368421e-06,
+      "loss": 0.7637,
+      "step": 2204
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.818530701754387e-06,
+      "loss": 0.9655,
+      "step": 2205
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.81798245614035e-06,
+      "loss": 1.0191,
+      "step": 2206
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.817434210526316e-06,
+      "loss": 1.15,
+      "step": 2207
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.816885964912282e-06,
+      "loss": 1.4601,
+      "step": 2208
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.816337719298246e-06,
+      "loss": 1.8073,
+      "step": 2209
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.81578947368421e-06,
+      "loss": 1.0025,
+      "step": 2210
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.815241228070176e-06,
+      "loss": 1.169,
+      "step": 2211
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.814692982456141e-06,
+      "loss": 0.4754,
+      "step": 2212
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.814144736842105e-06,
+      "loss": 1.1286,
+      "step": 2213
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.813596491228071e-06,
+      "loss": 0.7131,
+      "step": 2214
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.813048245614035e-06,
+      "loss": 1.2156,
+      "step": 2215
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.8125e-06,
+      "loss": 1.0396,
+      "step": 2216
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.811951754385966e-06,
+      "loss": 0.8552,
+      "step": 2217
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.81140350877193e-06,
+      "loss": 1.1191,
+      "step": 2218
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.810855263157896e-06,
+      "loss": 1.7059,
+      "step": 2219
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.81030701754386e-06,
+      "loss": 1.0332,
+      "step": 2220
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.809758771929826e-06,
+      "loss": 0.6885,
+      "step": 2221
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.80921052631579e-06,
+      "loss": 0.6238,
+      "step": 2222
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.808662280701755e-06,
+      "loss": 1.3478,
+      "step": 2223
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.808114035087721e-06,
+      "loss": 0.8186,
+      "step": 2224
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.807565789473685e-06,
+      "loss": 1.0523,
+      "step": 2225
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.807017543859649e-06,
+      "loss": 0.953,
+      "step": 2226
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.806469298245615e-06,
+      "loss": 0.868,
+      "step": 2227
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.80592105263158e-06,
+      "loss": 1.0579,
+      "step": 2228
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.805372807017544e-06,
+      "loss": 1.4845,
+      "step": 2229
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.80482456140351e-06,
+      "loss": 1.6151,
+      "step": 2230
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.804276315789474e-06,
+      "loss": 1.4718,
+      "step": 2231
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.80372807017544e-06,
+      "loss": 1.0452,
+      "step": 2232
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.803179824561404e-06,
+      "loss": 1.0152,
+      "step": 2233
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.80263157894737e-06,
+      "loss": 0.9302,
+      "step": 2234
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.802083333333335e-06,
+      "loss": 1.3393,
+      "step": 2235
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.801535087719299e-06,
+      "loss": 0.7492,
+      "step": 2236
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.800986842105263e-06,
+      "loss": 1.4841,
+      "step": 2237
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.800438596491229e-06,
+      "loss": 0.912,
+      "step": 2238
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.799890350877194e-06,
+      "loss": 1.0051,
+      "step": 2239
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.799342105263158e-06,
+      "loss": 0.8562,
+      "step": 2240
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.798793859649124e-06,
+      "loss": 0.9693,
+      "step": 2241
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.798245614035088e-06,
+      "loss": 0.7915,
+      "step": 2242
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.797697368421054e-06,
+      "loss": 1.2936,
+      "step": 2243
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.797149122807018e-06,
+      "loss": 0.7209,
+      "step": 2244
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.796600877192983e-06,
+      "loss": 1.0605,
+      "step": 2245
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.796052631578949e-06,
+      "loss": 0.7826,
+      "step": 2246
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.795504385964913e-06,
+      "loss": 0.8966,
+      "step": 2247
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.794956140350877e-06,
+      "loss": 0.9644,
+      "step": 2248
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.794407894736843e-06,
+      "loss": 1.0133,
+      "step": 2249
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.793859649122808e-06,
+      "loss": 0.9081,
+      "step": 2250
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.793311403508772e-06,
+      "loss": 1.3095,
+      "step": 2251
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.792763157894738e-06,
+      "loss": 0.615,
+      "step": 2252
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.792214912280702e-06,
+      "loss": 0.668,
+      "step": 2253
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.791666666666667e-06,
+      "loss": 1.6275,
+      "step": 2254
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.791118421052633e-06,
+      "loss": 1.3157,
+      "step": 2255
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.790570175438597e-06,
+      "loss": 0.9744,
+      "step": 2256
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.790021929824563e-06,
+      "loss": 1.537,
+      "step": 2257
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.789473684210527e-06,
+      "loss": 1.2152,
+      "step": 2258
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.788925438596492e-06,
+      "loss": 1.2973,
+      "step": 2259
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.788377192982456e-06,
+      "loss": 1.4694,
+      "step": 2260
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.787828947368422e-06,
+      "loss": 1.402,
+      "step": 2261
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.787280701754388e-06,
+      "loss": 1.2067,
+      "step": 2262
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.786732456140352e-06,
+      "loss": 0.6394,
+      "step": 2263
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.786184210526316e-06,
+      "loss": 1.1876,
+      "step": 2264
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.785635964912281e-06,
+      "loss": 0.7181,
+      "step": 2265
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.785087719298247e-06,
+      "loss": 1.1967,
+      "step": 2266
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.784539473684211e-06,
+      "loss": 1.5213,
+      "step": 2267
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.783991228070177e-06,
+      "loss": 1.4762,
+      "step": 2268
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.78344298245614e-06,
+      "loss": 0.7203,
+      "step": 2269
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.782894736842106e-06,
+      "loss": 1.2101,
+      "step": 2270
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.78234649122807e-06,
+      "loss": 0.8576,
+      "step": 2271
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.781798245614036e-06,
+      "loss": 1.3881,
+      "step": 2272
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.781250000000002e-06,
+      "loss": 2.0207,
+      "step": 2273
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.780701754385966e-06,
+      "loss": 1.1904,
+      "step": 2274
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.78015350877193e-06,
+      "loss": 0.9967,
+      "step": 2275
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.779605263157895e-06,
+      "loss": 1.1434,
+      "step": 2276
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.779057017543861e-06,
+      "loss": 0.6694,
+      "step": 2277
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.778508771929825e-06,
+      "loss": 0.6325,
+      "step": 2278
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.777960526315789e-06,
+      "loss": 1.2732,
+      "step": 2279
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.777412280701755e-06,
+      "loss": 0.8212,
+      "step": 2280
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.77686403508772e-06,
+      "loss": 1.0076,
+      "step": 2281
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.776315789473684e-06,
+      "loss": 1.6278,
+      "step": 2282
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.77576754385965e-06,
+      "loss": 0.5483,
+      "step": 2283
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.775219298245616e-06,
+      "loss": 0.9049,
+      "step": 2284
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.77467105263158e-06,
+      "loss": 1.022,
+      "step": 2285
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.774122807017544e-06,
+      "loss": 0.9751,
+      "step": 2286
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.77357456140351e-06,
+      "loss": 0.9144,
+      "step": 2287
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.773026315789475e-06,
+      "loss": 1.0593,
+      "step": 2288
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.77247807017544e-06,
+      "loss": 0.9235,
+      "step": 2289
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.771929824561405e-06,
+      "loss": 0.7921,
+      "step": 2290
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.771381578947369e-06,
+      "loss": 0.9761,
+      "step": 2291
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.770833333333334e-06,
+      "loss": 0.612,
+      "step": 2292
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.7702850877193e-06,
+      "loss": 1.3783,
+      "step": 2293
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.769736842105264e-06,
+      "loss": 1.1577,
+      "step": 2294
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.76918859649123e-06,
+      "loss": 0.5778,
+      "step": 2295
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.768640350877194e-06,
+      "loss": 0.7301,
+      "step": 2296
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.76809210526316e-06,
+      "loss": 0.7926,
+      "step": 2297
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.767543859649123e-06,
+      "loss": 0.796,
+      "step": 2298
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.766995614035089e-06,
+      "loss": 1.6625,
+      "step": 2299
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.766447368421054e-06,
+      "loss": 1.4634,
+      "step": 2300
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.765899122807018e-06,
+      "loss": 1.1196,
+      "step": 2301
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.765350877192982e-06,
+      "loss": 1.6599,
+      "step": 2302
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.764802631578948e-06,
+      "loss": 0.7723,
+      "step": 2303
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.764254385964914e-06,
+      "loss": 1.7235,
+      "step": 2304
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.763706140350878e-06,
+      "loss": 1.3126,
+      "step": 2305
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.763157894736842e-06,
+      "loss": 1.0388,
+      "step": 2306
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.762609649122807e-06,
+      "loss": 0.8098,
+      "step": 2307
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.762061403508773e-06,
+      "loss": 1.0984,
+      "step": 2308
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.761513157894737e-06,
+      "loss": 0.5902,
+      "step": 2309
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.760964912280703e-06,
+      "loss": 1.4062,
+      "step": 2310
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.760416666666668e-06,
+      "loss": 0.5122,
+      "step": 2311
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.759868421052632e-06,
+      "loss": 0.7328,
+      "step": 2312
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.759320175438596e-06,
+      "loss": 0.5903,
+      "step": 2313
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.758771929824562e-06,
+      "loss": 1.634,
+      "step": 2314
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.758223684210528e-06,
+      "loss": 1.4906,
+      "step": 2315
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.757675438596492e-06,
+      "loss": 0.6407,
+      "step": 2316
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.757127192982456e-06,
+      "loss": 1.0508,
+      "step": 2317
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.756578947368421e-06,
+      "loss": 0.9011,
+      "step": 2318
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.756030701754387e-06,
+      "loss": 1.7955,
+      "step": 2319
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.755482456140351e-06,
+      "loss": 1.2462,
+      "step": 2320
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.754934210526317e-06,
+      "loss": 1.6784,
+      "step": 2321
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.75438596491228e-06,
+      "loss": 1.0694,
+      "step": 2322
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.753837719298246e-06,
+      "loss": 0.8665,
+      "step": 2323
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.75328947368421e-06,
+      "loss": 1.2542,
+      "step": 2324
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.752741228070176e-06,
+      "loss": 0.8343,
+      "step": 2325
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.752192982456142e-06,
+      "loss": 1.4739,
+      "step": 2326
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.751644736842107e-06,
+      "loss": 1.833,
+      "step": 2327
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.751096491228071e-06,
+      "loss": 0.9429,
+      "step": 2328
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.750548245614035e-06,
+      "loss": 1.236,
+      "step": 2329
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.750000000000001e-06,
+      "loss": 1.217,
+      "step": 2330
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.749451754385967e-06,
+      "loss": 0.982,
+      "step": 2331
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.74890350877193e-06,
+      "loss": 0.695,
+      "step": 2332
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.748355263157895e-06,
+      "loss": 0.9173,
+      "step": 2333
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.74780701754386e-06,
+      "loss": 0.5047,
+      "step": 2334
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.747258771929826e-06,
+      "loss": 1.2705,
+      "step": 2335
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.74671052631579e-06,
+      "loss": 1.671,
+      "step": 2336
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.746162280701756e-06,
+      "loss": 1.3048,
+      "step": 2337
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.745614035087721e-06,
+      "loss": 0.8303,
+      "step": 2338
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.745065789473685e-06,
+      "loss": 1.0829,
+      "step": 2339
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.74451754385965e-06,
+      "loss": 0.5975,
+      "step": 2340
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.743969298245615e-06,
+      "loss": 0.9361,
+      "step": 2341
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.74342105263158e-06,
+      "loss": 0.5098,
+      "step": 2342
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.742872807017544e-06,
+      "loss": 1.4122,
+      "step": 2343
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.742324561403508e-06,
+      "loss": 0.5074,
+      "step": 2344
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.741776315789474e-06,
+      "loss": 0.6802,
+      "step": 2345
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.74122807017544e-06,
+      "loss": 1.4361,
+      "step": 2346
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.740679824561404e-06,
+      "loss": 0.8011,
+      "step": 2347
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.74013157894737e-06,
+      "loss": 1.7727,
+      "step": 2348
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.739583333333333e-06,
+      "loss": 0.9079,
+      "step": 2349
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.739035087719299e-06,
+      "loss": 1.4073,
+      "step": 2350
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.738486842105263e-06,
+      "loss": 1.1221,
+      "step": 2351
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.737938596491229e-06,
+      "loss": 1.6141,
+      "step": 2352
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.737390350877194e-06,
+      "loss": 1.4204,
+      "step": 2353
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.736842105263158e-06,
+      "loss": 0.8825,
+      "step": 2354
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.736293859649122e-06,
+      "loss": 1.0358,
+      "step": 2355
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.735745614035088e-06,
+      "loss": 1.6279,
+      "step": 2356
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.735197368421054e-06,
+      "loss": 0.6156,
+      "step": 2357
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.734649122807018e-06,
+      "loss": 0.787,
+      "step": 2358
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.734100877192983e-06,
+      "loss": 1.031,
+      "step": 2359
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.733552631578947e-06,
+      "loss": 1.0573,
+      "step": 2360
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.733004385964913e-06,
+      "loss": 0.9447,
+      "step": 2361
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.732456140350877e-06,
+      "loss": 1.3945,
+      "step": 2362
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.731907894736843e-06,
+      "loss": 1.1708,
+      "step": 2363
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.731359649122808e-06,
+      "loss": 1.4092,
+      "step": 2364
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.730811403508772e-06,
+      "loss": 1.3312,
+      "step": 2365
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.730263157894738e-06,
+      "loss": 0.5627,
+      "step": 2366
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.729714912280702e-06,
+      "loss": 0.8034,
+      "step": 2367
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.729166666666668e-06,
+      "loss": 1.1641,
+      "step": 2368
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.728618421052633e-06,
+      "loss": 1.2321,
+      "step": 2369
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.728070175438597e-06,
+      "loss": 1.4036,
+      "step": 2370
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.727521929824561e-06,
+      "loss": 1.5501,
+      "step": 2371
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.726973684210527e-06,
+      "loss": 1.6087,
+      "step": 2372
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.726425438596493e-06,
+      "loss": 1.0902,
+      "step": 2373
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.725877192982457e-06,
+      "loss": 0.756,
+      "step": 2374
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.725328947368422e-06,
+      "loss": 1.17,
+      "step": 2375
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.724780701754386e-06,
+      "loss": 0.6166,
+      "step": 2376
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.724232456140352e-06,
+      "loss": 1.4097,
+      "step": 2377
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.723684210526316e-06,
+      "loss": 1.423,
+      "step": 2378
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.723135964912282e-06,
+      "loss": 0.9706,
+      "step": 2379
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.722587719298247e-06,
+      "loss": 0.7238,
+      "step": 2380
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.722039473684211e-06,
+      "loss": 0.8076,
+      "step": 2381
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.721491228070175e-06,
+      "loss": 0.9654,
+      "step": 2382
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.720942982456141e-06,
+      "loss": 0.9364,
+      "step": 2383
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.720394736842107e-06,
+      "loss": 0.7154,
+      "step": 2384
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.71984649122807e-06,
+      "loss": 1.2078,
+      "step": 2385
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.719298245614036e-06,
+      "loss": 0.9005,
+      "step": 2386
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.71875e-06,
+      "loss": 1.2777,
+      "step": 2387
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.718201754385966e-06,
+      "loss": 0.5436,
+      "step": 2388
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.71765350877193e-06,
+      "loss": 1.0376,
+      "step": 2389
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.717105263157895e-06,
+      "loss": 0.6643,
+      "step": 2390
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.716557017543861e-06,
+      "loss": 0.6102,
+      "step": 2391
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.716008771929825e-06,
+      "loss": 1.9505,
+      "step": 2392
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.715460526315789e-06,
+      "loss": 1.2831,
+      "step": 2393
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.714912280701755e-06,
+      "loss": 1.103,
+      "step": 2394
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.71436403508772e-06,
+      "loss": 0.6435,
+      "step": 2395
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.713815789473684e-06,
+      "loss": 1.2797,
+      "step": 2396
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.71326754385965e-06,
+      "loss": 1.5235,
+      "step": 2397
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.712719298245614e-06,
+      "loss": 0.7526,
+      "step": 2398
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.71217105263158e-06,
+      "loss": 1.5248,
+      "step": 2399
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.711622807017545e-06,
+      "loss": 1.4429,
+      "step": 2400
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.71107456140351e-06,
+      "loss": 1.0327,
+      "step": 2401
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.710526315789475e-06,
+      "loss": 0.7581,
+      "step": 2402
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.709978070175439e-06,
+      "loss": 1.012,
+      "step": 2403
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.709429824561405e-06,
+      "loss": 1.0931,
+      "step": 2404
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.708881578947369e-06,
+      "loss": 0.8683,
+      "step": 2405
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.708333333333334e-06,
+      "loss": 1.7864,
+      "step": 2406
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.7077850877193e-06,
+      "loss": 0.9075,
+      "step": 2407
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.707236842105264e-06,
+      "loss": 1.3964,
+      "step": 2408
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.706688596491228e-06,
+      "loss": 1.3905,
+      "step": 2409
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.706140350877194e-06,
+      "loss": 1.0616,
+      "step": 2410
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.70559210526316e-06,
+      "loss": 0.9063,
+      "step": 2411
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.705043859649123e-06,
+      "loss": 0.7567,
+      "step": 2412
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.704495614035089e-06,
+      "loss": 1.5739,
+      "step": 2413
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.703947368421053e-06,
+      "loss": 1.584,
+      "step": 2414
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.703399122807019e-06,
+      "loss": 1.1592,
+      "step": 2415
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.702850877192983e-06,
+      "loss": 1.0887,
+      "step": 2416
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.702302631578948e-06,
+      "loss": 0.8086,
+      "step": 2417
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.701754385964914e-06,
+      "loss": 0.6635,
+      "step": 2418
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.701206140350878e-06,
+      "loss": 0.8417,
+      "step": 2419
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.700657894736842e-06,
+      "loss": 1.3664,
+      "step": 2420
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.700109649122808e-06,
+      "loss": 1.0546,
+      "step": 2421
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.699561403508773e-06,
+      "loss": 1.1522,
+      "step": 2422
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.699013157894737e-06,
+      "loss": 0.9512,
+      "step": 2423
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.698464912280703e-06,
+      "loss": 0.7852,
+      "step": 2424
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.697916666666667e-06,
+      "loss": 1.5247,
+      "step": 2425
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.697368421052633e-06,
+      "loss": 1.293,
+      "step": 2426
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.696820175438597e-06,
+      "loss": 0.9794,
+      "step": 2427
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.696271929824562e-06,
+      "loss": 1.1551,
+      "step": 2428
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.695723684210528e-06,
+      "loss": 0.9605,
+      "step": 2429
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.695175438596492e-06,
+      "loss": 1.2138,
+      "step": 2430
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.694627192982456e-06,
+      "loss": 1.6973,
+      "step": 2431
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.694078947368422e-06,
+      "loss": 0.7818,
+      "step": 2432
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.693530701754387e-06,
+      "loss": 1.0369,
+      "step": 2433
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.692982456140351e-06,
+      "loss": 0.8788,
+      "step": 2434
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.692434210526315e-06,
+      "loss": 0.8534,
+      "step": 2435
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.69188596491228e-06,
+      "loss": 1.0234,
+      "step": 2436
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.691337719298246e-06,
+      "loss": 1.1755,
+      "step": 2437
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.690789473684212e-06,
+      "loss": 0.9945,
+      "step": 2438
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.690241228070176e-06,
+      "loss": 0.7256,
+      "step": 2439
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.689692982456142e-06,
+      "loss": 1.3341,
+      "step": 2440
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.689144736842106e-06,
+      "loss": 1.3188,
+      "step": 2441
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.688596491228071e-06,
+      "loss": 0.5691,
+      "step": 2442
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.688048245614035e-06,
+      "loss": 0.9453,
+      "step": 2443
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.687500000000001e-06,
+      "loss": 0.5871,
+      "step": 2444
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.686951754385967e-06,
+      "loss": 0.9732,
+      "step": 2445
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.68640350877193e-06,
+      "loss": 1.6206,
+      "step": 2446
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.685855263157895e-06,
+      "loss": 1.2797,
+      "step": 2447
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.68530701754386e-06,
+      "loss": 1.1464,
+      "step": 2448
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.684758771929826e-06,
+      "loss": 1.4451,
+      "step": 2449
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.68421052631579e-06,
+      "loss": 1.2122,
+      "step": 2450
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.683662280701754e-06,
+      "loss": 1.1304,
+      "step": 2451
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.68311403508772e-06,
+      "loss": 1.1324,
+      "step": 2452
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.682565789473685e-06,
+      "loss": 0.3551,
+      "step": 2453
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.68201754385965e-06,
+      "loss": 1.4718,
+      "step": 2454
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.681469298245615e-06,
+      "loss": 0.7991,
+      "step": 2455
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.68092105263158e-06,
+      "loss": 0.9131,
+      "step": 2456
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.680372807017545e-06,
+      "loss": 0.8577,
+      "step": 2457
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.679824561403509e-06,
+      "loss": 0.9019,
+      "step": 2458
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.679276315789474e-06,
+      "loss": 1.1274,
+      "step": 2459
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.67872807017544e-06,
+      "loss": 1.0949,
+      "step": 2460
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.678179824561404e-06,
+      "loss": 1.1459,
+      "step": 2461
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.677631578947368e-06,
+      "loss": 0.8477,
+      "step": 2462
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.677083333333334e-06,
+      "loss": 0.9972,
+      "step": 2463
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.6765350877193e-06,
+      "loss": 1.0,
+      "step": 2464
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.675986842105263e-06,
+      "loss": 0.4861,
+      "step": 2465
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.675438596491229e-06,
+      "loss": 1.3054,
+      "step": 2466
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.674890350877195e-06,
+      "loss": 1.0912,
+      "step": 2467
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.674342105263159e-06,
+      "loss": 0.8021,
+      "step": 2468
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.673793859649123e-06,
+      "loss": 1.3372,
+      "step": 2469
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.673245614035088e-06,
+      "loss": 0.7934,
+      "step": 2470
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.672697368421054e-06,
+      "loss": 1.3729,
+      "step": 2471
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.672149122807018e-06,
+      "loss": 1.0037,
+      "step": 2472
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.671600877192984e-06,
+      "loss": 0.7676,
+      "step": 2473
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.671052631578948e-06,
+      "loss": 1.0727,
+      "step": 2474
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.670504385964913e-06,
+      "loss": 1.279,
+      "step": 2475
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.669956140350879e-06,
+      "loss": 1.362,
+      "step": 2476
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.669407894736843e-06,
+      "loss": 0.9893,
+      "step": 2477
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.668859649122807e-06,
+      "loss": 1.2598,
+      "step": 2478
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.668311403508773e-06,
+      "loss": 0.9804,
+      "step": 2479
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.667763157894738e-06,
+      "loss": 1.9221,
+      "step": 2480
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.667214912280702e-06,
+      "loss": 0.7719,
+      "step": 2481
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.666666666666668e-06,
+      "loss": 0.7248,
+      "step": 2482
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.666118421052633e-06,
+      "loss": 2.0262,
+      "step": 2483
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.665570175438597e-06,
+      "loss": 1.5674,
+      "step": 2484
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.665021929824561e-06,
+      "loss": 1.4142,
+      "step": 2485
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.664473684210527e-06,
+      "loss": 0.77,
+      "step": 2486
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.663925438596493e-06,
+      "loss": 1.3477,
+      "step": 2487
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.663377192982457e-06,
+      "loss": 1.1055,
+      "step": 2488
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.66282894736842e-06,
+      "loss": 0.7386,
+      "step": 2489
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.662280701754386e-06,
+      "loss": 1.6405,
+      "step": 2490
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.661732456140352e-06,
+      "loss": 1.0431,
+      "step": 2491
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.661184210526316e-06,
+      "loss": 1.2988,
+      "step": 2492
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.660635964912282e-06,
+      "loss": 0.785,
+      "step": 2493
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.660087719298247e-06,
+      "loss": 1.2312,
+      "step": 2494
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.659539473684211e-06,
+      "loss": 1.203,
+      "step": 2495
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.658991228070175e-06,
+      "loss": 1.0686,
+      "step": 2496
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.658442982456141e-06,
+      "loss": 0.5534,
+      "step": 2497
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.657894736842107e-06,
+      "loss": 0.7953,
+      "step": 2498
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.65734649122807e-06,
+      "loss": 1.1889,
+      "step": 2499
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.656798245614035e-06,
+      "loss": 1.2893,
+      "step": 2500
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.65625e-06,
+      "loss": 1.1408,
+      "step": 2501
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.655701754385966e-06,
+      "loss": 1.0073,
+      "step": 2502
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.65515350877193e-06,
+      "loss": 0.7317,
+      "step": 2503
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.654605263157896e-06,
+      "loss": 0.7488,
+      "step": 2504
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.65405701754386e-06,
+      "loss": 1.5401,
+      "step": 2505
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.653508771929825e-06,
+      "loss": 1.0831,
+      "step": 2506
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.65296052631579e-06,
+      "loss": 1.518,
+      "step": 2507
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.652412280701755e-06,
+      "loss": 1.2627,
+      "step": 2508
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.65186403508772e-06,
+      "loss": 0.8464,
+      "step": 2509
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.651315789473685e-06,
+      "loss": 0.8881,
+      "step": 2510
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.65076754385965e-06,
+      "loss": 1.2098,
+      "step": 2511
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.650219298245614e-06,
+      "loss": 1.0086,
+      "step": 2512
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.64967105263158e-06,
+      "loss": 0.8699,
+      "step": 2513
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.649122807017546e-06,
+      "loss": 0.685,
+      "step": 2514
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.64857456140351e-06,
+      "loss": 0.7565,
+      "step": 2515
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.648026315789474e-06,
+      "loss": 1.6055,
+      "step": 2516
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.64747807017544e-06,
+      "loss": 1.3116,
+      "step": 2517
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.646929824561405e-06,
+      "loss": 1.1876,
+      "step": 2518
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.646381578947369e-06,
+      "loss": 1.2244,
+      "step": 2519
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.645833333333335e-06,
+      "loss": 0.6519,
+      "step": 2520
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.645285087719299e-06,
+      "loss": 0.7279,
+      "step": 2521
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.644736842105264e-06,
+      "loss": 1.7733,
+      "step": 2522
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.644188596491228e-06,
+      "loss": 0.7442,
+      "step": 2523
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.643640350877194e-06,
+      "loss": 1.293,
+      "step": 2524
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.64309210526316e-06,
+      "loss": 0.9602,
+      "step": 2525
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.642543859649124e-06,
+      "loss": 1.0666,
+      "step": 2526
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.641995614035087e-06,
+      "loss": 0.6232,
+      "step": 2527
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.641447368421053e-06,
+      "loss": 1.0403,
+      "step": 2528
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.640899122807019e-06,
+      "loss": 0.7627,
+      "step": 2529
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.640350877192983e-06,
+      "loss": 0.5932,
+      "step": 2530
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.639802631578948e-06,
+      "loss": 0.6297,
+      "step": 2531
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.639254385964912e-06,
+      "loss": 0.8634,
+      "step": 2532
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.638706140350878e-06,
+      "loss": 0.9554,
+      "step": 2533
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.638157894736842e-06,
+      "loss": 0.7991,
+      "step": 2534
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.637609649122808e-06,
+      "loss": 0.8718,
+      "step": 2535
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.637061403508773e-06,
+      "loss": 1.0236,
+      "step": 2536
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.636513157894737e-06,
+      "loss": 0.7154,
+      "step": 2537
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.635964912280701e-06,
+      "loss": 1.0142,
+      "step": 2538
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.635416666666667e-06,
+      "loss": 0.9436,
+      "step": 2539
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.634868421052633e-06,
+      "loss": 0.903,
+      "step": 2540
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.634320175438597e-06,
+      "loss": 1.3038,
+      "step": 2541
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.633771929824562e-06,
+      "loss": 0.5194,
+      "step": 2542
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.633223684210526e-06,
+      "loss": 0.8123,
+      "step": 2543
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.632675438596492e-06,
+      "loss": 0.9678,
+      "step": 2544
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.632127192982456e-06,
+      "loss": 1.1897,
+      "step": 2545
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.631578947368422e-06,
+      "loss": 1.1578,
+      "step": 2546
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.631030701754387e-06,
+      "loss": 1.0735,
+      "step": 2547
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.630482456140351e-06,
+      "loss": 1.0784,
+      "step": 2548
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.629934210526317e-06,
+      "loss": 0.6241,
+      "step": 2549
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.629385964912281e-06,
+      "loss": 0.8269,
+      "step": 2550
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.628837719298247e-06,
+      "loss": 1.278,
+      "step": 2551
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.628289473684212e-06,
+      "loss": 0.9214,
+      "step": 2552
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.627741228070176e-06,
+      "loss": 1.4938,
+      "step": 2553
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.62719298245614e-06,
+      "loss": 1.495,
+      "step": 2554
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.626644736842106e-06,
+      "loss": 1.2415,
+      "step": 2555
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.626096491228072e-06,
+      "loss": 0.9105,
+      "step": 2556
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.625548245614036e-06,
+      "loss": 0.9015,
+      "step": 2557
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.625000000000001e-06,
+      "loss": 1.448,
+      "step": 2558
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.624451754385965e-06,
+      "loss": 0.6772,
+      "step": 2559
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.623903508771931e-06,
+      "loss": 0.9787,
+      "step": 2560
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.623355263157895e-06,
+      "loss": 1.0139,
+      "step": 2561
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.62280701754386e-06,
+      "loss": 1.1904,
+      "step": 2562
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.622258771929826e-06,
+      "loss": 1.07,
+      "step": 2563
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.62171052631579e-06,
+      "loss": 0.7628,
+      "step": 2564
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.621162280701754e-06,
+      "loss": 1.1297,
+      "step": 2565
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.62061403508772e-06,
+      "loss": 0.7766,
+      "step": 2566
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.620065789473686e-06,
+      "loss": 0.6872,
+      "step": 2567
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.61951754385965e-06,
+      "loss": 0.692,
+      "step": 2568
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.618969298245615e-06,
+      "loss": 1.1912,
+      "step": 2569
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.61842105263158e-06,
+      "loss": 0.6386,
+      "step": 2570
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.617872807017545e-06,
+      "loss": 0.7716,
+      "step": 2571
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.617324561403509e-06,
+      "loss": 0.7977,
+      "step": 2572
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.616776315789474e-06,
+      "loss": 0.9657,
+      "step": 2573
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.61622807017544e-06,
+      "loss": 0.6152,
+      "step": 2574
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.615679824561404e-06,
+      "loss": 0.9623,
+      "step": 2575
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.615131578947368e-06,
+      "loss": 0.9732,
+      "step": 2576
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.614583333333334e-06,
+      "loss": 0.4053,
+      "step": 2577
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.6140350877193e-06,
+      "loss": 1.0157,
+      "step": 2578
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.613486842105263e-06,
+      "loss": 0.765,
+      "step": 2579
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.612938596491229e-06,
+      "loss": 0.9378,
+      "step": 2580
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.612390350877193e-06,
+      "loss": 1.2451,
+      "step": 2581
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.611842105263159e-06,
+      "loss": 0.7553,
+      "step": 2582
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.611293859649123e-06,
+      "loss": 1.1795,
+      "step": 2583
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.610745614035088e-06,
+      "loss": 1.2494,
+      "step": 2584
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.610197368421054e-06,
+      "loss": 1.2707,
+      "step": 2585
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.609649122807018e-06,
+      "loss": 1.0192,
+      "step": 2586
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.609100877192984e-06,
+      "loss": 0.8604,
+      "step": 2587
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.608552631578948e-06,
+      "loss": 1.1148,
+      "step": 2588
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.608004385964913e-06,
+      "loss": 0.6384,
+      "step": 2589
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.607456140350879e-06,
+      "loss": 0.8253,
+      "step": 2590
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.606907894736843e-06,
+      "loss": 1.4461,
+      "step": 2591
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.606359649122807e-06,
+      "loss": 0.912,
+      "step": 2592
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.605811403508773e-06,
+      "loss": 0.6762,
+      "step": 2593
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.605263157894738e-06,
+      "loss": 1.0084,
+      "step": 2594
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.604714912280702e-06,
+      "loss": 0.688,
+      "step": 2595
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.604166666666668e-06,
+      "loss": 1.5301,
+      "step": 2596
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.603618421052632e-06,
+      "loss": 0.812,
+      "step": 2597
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.603070175438598e-06,
+      "loss": 0.6698,
+      "step": 2598
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.602521929824562e-06,
+      "loss": 0.704,
+      "step": 2599
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.601973684210527e-06,
+      "loss": 0.6402,
+      "step": 2600
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.601425438596493e-06,
+      "loss": 0.7088,
+      "step": 2601
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.600877192982457e-06,
+      "loss": 1.1162,
+      "step": 2602
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.600328947368421e-06,
+      "loss": 0.7993,
+      "step": 2603
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.599780701754387e-06,
+      "loss": 1.1792,
+      "step": 2604
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.599232456140352e-06,
+      "loss": 0.7952,
+      "step": 2605
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.598684210526316e-06,
+      "loss": 1.5711,
+      "step": 2606
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.59813596491228e-06,
+      "loss": 0.8318,
+      "step": 2607
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.597587719298246e-06,
+      "loss": 0.8776,
+      "step": 2608
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.597039473684212e-06,
+      "loss": 0.8287,
+      "step": 2609
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.596491228070176e-06,
+      "loss": 0.7129,
+      "step": 2610
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.595942982456141e-06,
+      "loss": 1.7141,
+      "step": 2611
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.595394736842107e-06,
+      "loss": 0.5681,
+      "step": 2612
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.594846491228071e-06,
+      "loss": 1.2015,
+      "step": 2613
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.594298245614035e-06,
+      "loss": 1.3074,
+      "step": 2614
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.59375e-06,
+      "loss": 0.809,
+      "step": 2615
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.593201754385966e-06,
+      "loss": 0.7991,
+      "step": 2616
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.59265350877193e-06,
+      "loss": 1.072,
+      "step": 2617
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.592105263157894e-06,
+      "loss": 1.5692,
+      "step": 2618
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.59155701754386e-06,
+      "loss": 0.9832,
+      "step": 2619
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.591008771929825e-06,
+      "loss": 0.5494,
+      "step": 2620
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.59046052631579e-06,
+      "loss": 1.0509,
+      "step": 2621
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.589912280701755e-06,
+      "loss": 1.5645,
+      "step": 2622
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.58936403508772e-06,
+      "loss": 1.3604,
+      "step": 2623
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.588815789473685e-06,
+      "loss": 1.3557,
+      "step": 2624
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.58826754385965e-06,
+      "loss": 0.8298,
+      "step": 2625
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.587719298245614e-06,
+      "loss": 1.4125,
+      "step": 2626
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.58717105263158e-06,
+      "loss": 1.6971,
+      "step": 2627
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.586622807017546e-06,
+      "loss": 0.828,
+      "step": 2628
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.58607456140351e-06,
+      "loss": 0.8813,
+      "step": 2629
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.585526315789474e-06,
+      "loss": 0.9984,
+      "step": 2630
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.58497807017544e-06,
+      "loss": 1.2976,
+      "step": 2631
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.584429824561405e-06,
+      "loss": 1.3384,
+      "step": 2632
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.583881578947369e-06,
+      "loss": 1.1589,
+      "step": 2633
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.583333333333333e-06,
+      "loss": 0.7424,
+      "step": 2634
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.582785087719299e-06,
+      "loss": 1.1569,
+      "step": 2635
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.582236842105264e-06,
+      "loss": 0.9107,
+      "step": 2636
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.581688596491228e-06,
+      "loss": 1.2548,
+      "step": 2637
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.581140350877194e-06,
+      "loss": 1.3101,
+      "step": 2638
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.58059210526316e-06,
+      "loss": 1.2273,
+      "step": 2639
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.580043859649124e-06,
+      "loss": 1.0551,
+      "step": 2640
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.579495614035088e-06,
+      "loss": 1.593,
+      "step": 2641
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.578947368421053e-06,
+      "loss": 0.8474,
+      "step": 2642
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.578399122807019e-06,
+      "loss": 1.5032,
+      "step": 2643
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.577850877192983e-06,
+      "loss": 0.8472,
+      "step": 2644
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.577302631578947e-06,
+      "loss": 1.0478,
+      "step": 2645
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.576754385964913e-06,
+      "loss": 0.9796,
+      "step": 2646
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.576206140350878e-06,
+      "loss": 0.6158,
+      "step": 2647
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.575657894736842e-06,
+      "loss": 1.3221,
+      "step": 2648
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.575109649122808e-06,
+      "loss": 0.934,
+      "step": 2649
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.574561403508772e-06,
+      "loss": 1.1056,
+      "step": 2650
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.574013157894738e-06,
+      "loss": 0.9521,
+      "step": 2651
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.573464912280702e-06,
+      "loss": 0.8053,
+      "step": 2652
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.572916666666667e-06,
+      "loss": 0.7522,
+      "step": 2653
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.572368421052633e-06,
+      "loss": 1.2707,
+      "step": 2654
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.571820175438597e-06,
+      "loss": 0.9296,
+      "step": 2655
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.571271929824561e-06,
+      "loss": 1.0898,
+      "step": 2656
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.570723684210527e-06,
+      "loss": 1.2028,
+      "step": 2657
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.570175438596492e-06,
+      "loss": 1.2873,
+      "step": 2658
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.569627192982458e-06,
+      "loss": 0.8535,
+      "step": 2659
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.569078947368422e-06,
+      "loss": 1.1848,
+      "step": 2660
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.568530701754386e-06,
+      "loss": 0.7058,
+      "step": 2661
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.567982456140352e-06,
+      "loss": 1.4335,
+      "step": 2662
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.567434210526317e-06,
+      "loss": 0.966,
+      "step": 2663
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.566885964912281e-06,
+      "loss": 1.9359,
+      "step": 2664
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.566337719298247e-06,
+      "loss": 0.7109,
+      "step": 2665
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.565789473684213e-06,
+      "loss": 1.5319,
+      "step": 2666
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.565241228070176e-06,
+      "loss": 0.9673,
+      "step": 2667
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.56469298245614e-06,
+      "loss": 0.5167,
+      "step": 2668
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.564144736842106e-06,
+      "loss": 0.8208,
+      "step": 2669
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.563596491228072e-06,
+      "loss": 0.7581,
+      "step": 2670
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.563048245614036e-06,
+      "loss": 0.585,
+      "step": 2671
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.5625e-06,
+      "loss": 0.8087,
+      "step": 2672
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.561951754385965e-06,
+      "loss": 0.9599,
+      "step": 2673
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.561403508771931e-06,
+      "loss": 1.2166,
+      "step": 2674
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.560855263157895e-06,
+      "loss": 0.6136,
+      "step": 2675
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.56030701754386e-06,
+      "loss": 0.7688,
+      "step": 2676
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.559758771929825e-06,
+      "loss": 0.8117,
+      "step": 2677
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.55921052631579e-06,
+      "loss": 1.5499,
+      "step": 2678
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.558662280701754e-06,
+      "loss": 1.376,
+      "step": 2679
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.55811403508772e-06,
+      "loss": 1.1618,
+      "step": 2680
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.557565789473686e-06,
+      "loss": 0.8003,
+      "step": 2681
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.55701754385965e-06,
+      "loss": 1.026,
+      "step": 2682
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.556469298245614e-06,
+      "loss": 0.6467,
+      "step": 2683
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.55592105263158e-06,
+      "loss": 1.0472,
+      "step": 2684
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.555372807017545e-06,
+      "loss": 1.2178,
+      "step": 2685
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.554824561403509e-06,
+      "loss": 0.8197,
+      "step": 2686
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.554276315789475e-06,
+      "loss": 0.8743,
+      "step": 2687
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.553728070175439e-06,
+      "loss": 0.9417,
+      "step": 2688
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.553179824561404e-06,
+      "loss": 1.1538,
+      "step": 2689
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.552631578947368e-06,
+      "loss": 1.7618,
+      "step": 2690
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.552083333333334e-06,
+      "loss": 1.2338,
+      "step": 2691
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.5515350877193e-06,
+      "loss": 1.7109,
+      "step": 2692
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.550986842105264e-06,
+      "loss": 0.6639,
+      "step": 2693
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.550438596491228e-06,
+      "loss": 1.1053,
+      "step": 2694
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.549890350877193e-06,
+      "loss": 0.8807,
+      "step": 2695
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.549342105263159e-06,
+      "loss": 1.2035,
+      "step": 2696
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.548793859649125e-06,
+      "loss": 0.8907,
+      "step": 2697
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.548245614035089e-06,
+      "loss": 1.54,
+      "step": 2698
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.547697368421053e-06,
+      "loss": 0.7532,
+      "step": 2699
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.547149122807018e-06,
+      "loss": 0.5412,
+      "step": 2700
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.546600877192984e-06,
+      "loss": 1.3182,
+      "step": 2701
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.546052631578948e-06,
+      "loss": 0.6719,
+      "step": 2702
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.545504385964914e-06,
+      "loss": 0.8429,
+      "step": 2703
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.544956140350878e-06,
+      "loss": 1.0082,
+      "step": 2704
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.544407894736843e-06,
+      "loss": 1.4095,
+      "step": 2705
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.543859649122807e-06,
+      "loss": 1.0034,
+      "step": 2706
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.543311403508773e-06,
+      "loss": 0.6799,
+      "step": 2707
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.542763157894739e-06,
+      "loss": 1.0634,
+      "step": 2708
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.542214912280703e-06,
+      "loss": 1.455,
+      "step": 2709
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.541666666666666e-06,
+      "loss": 1.1895,
+      "step": 2710
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.541118421052632e-06,
+      "loss": 1.138,
+      "step": 2711
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.540570175438598e-06,
+      "loss": 0.8899,
+      "step": 2712
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.540021929824562e-06,
+      "loss": 1.092,
+      "step": 2713
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.539473684210527e-06,
+      "loss": 1.2902,
+      "step": 2714
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.538925438596491e-06,
+      "loss": 1.0003,
+      "step": 2715
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.538377192982457e-06,
+      "loss": 1.0851,
+      "step": 2716
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.537828947368421e-06,
+      "loss": 0.6867,
+      "step": 2717
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.537280701754387e-06,
+      "loss": 1.1816,
+      "step": 2718
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.536732456140352e-06,
+      "loss": 0.9278,
+      "step": 2719
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.536184210526316e-06,
+      "loss": 1.279,
+      "step": 2720
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.53563596491228e-06,
+      "loss": 1.4725,
+      "step": 2721
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.535087719298246e-06,
+      "loss": 0.7448,
+      "step": 2722
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.534539473684212e-06,
+      "loss": 1.3295,
+      "step": 2723
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.533991228070176e-06,
+      "loss": 0.7098,
+      "step": 2724
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.533442982456141e-06,
+      "loss": 0.7662,
+      "step": 2725
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.532894736842105e-06,
+      "loss": 1.1564,
+      "step": 2726
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.532346491228071e-06,
+      "loss": 0.9512,
+      "step": 2727
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.531798245614035e-06,
+      "loss": 1.0063,
+      "step": 2728
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.53125e-06,
+      "loss": 0.8957,
+      "step": 2729
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.530701754385966e-06,
+      "loss": 1.0349,
+      "step": 2730
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.53015350877193e-06,
+      "loss": 0.9631,
+      "step": 2731
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.529605263157894e-06,
+      "loss": 0.836,
+      "step": 2732
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.52905701754386e-06,
+      "loss": 1.0677,
+      "step": 2733
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.528508771929826e-06,
+      "loss": 1.0787,
+      "step": 2734
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.527960526315791e-06,
+      "loss": 1.1064,
+      "step": 2735
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.527412280701755e-06,
+      "loss": 0.5368,
+      "step": 2736
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.52686403508772e-06,
+      "loss": 0.81,
+      "step": 2737
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.526315789473685e-06,
+      "loss": 1.0308,
+      "step": 2738
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.52576754385965e-06,
+      "loss": 1.6287,
+      "step": 2739
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.525219298245615e-06,
+      "loss": 0.6435,
+      "step": 2740
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.52467105263158e-06,
+      "loss": 0.9423,
+      "step": 2741
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.524122807017544e-06,
+      "loss": 0.9522,
+      "step": 2742
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.52357456140351e-06,
+      "loss": 1.5984,
+      "step": 2743
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.523026315789474e-06,
+      "loss": 0.9689,
+      "step": 2744
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.52247807017544e-06,
+      "loss": 1.3884,
+      "step": 2745
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.521929824561405e-06,
+      "loss": 1.287,
+      "step": 2746
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.52138157894737e-06,
+      "loss": 0.6544,
+      "step": 2747
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.520833333333333e-06,
+      "loss": 0.6348,
+      "step": 2748
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.520285087719299e-06,
+      "loss": 1.3219,
+      "step": 2749
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.519736842105265e-06,
+      "loss": 1.4808,
+      "step": 2750
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.519188596491229e-06,
+      "loss": 1.0646,
+      "step": 2751
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.518640350877194e-06,
+      "loss": 1.1296,
+      "step": 2752
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.518092105263158e-06,
+      "loss": 0.862,
+      "step": 2753
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.517543859649124e-06,
+      "loss": 1.5967,
+      "step": 2754
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.516995614035088e-06,
+      "loss": 0.6203,
+      "step": 2755
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.516447368421053e-06,
+      "loss": 1.0184,
+      "step": 2756
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.51589912280702e-06,
+      "loss": 1.2614,
+      "step": 2757
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.515350877192983e-06,
+      "loss": 1.1554,
+      "step": 2758
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.514802631578947e-06,
+      "loss": 1.1095,
+      "step": 2759
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.514254385964913e-06,
+      "loss": 0.9307,
+      "step": 2760
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.513706140350878e-06,
+      "loss": 0.6955,
+      "step": 2761
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.513157894736842e-06,
+      "loss": 1.9382,
+      "step": 2762
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.512609649122806e-06,
+      "loss": 0.7253,
+      "step": 2763
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.512061403508772e-06,
+      "loss": 1.2029,
+      "step": 2764
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.511513157894738e-06,
+      "loss": 0.4237,
+      "step": 2765
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.510964912280702e-06,
+      "loss": 0.984,
+      "step": 2766
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.510416666666667e-06,
+      "loss": 1.0211,
+      "step": 2767
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.509868421052633e-06,
+      "loss": 1.1402,
+      "step": 2768
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.509320175438597e-06,
+      "loss": 1.4729,
+      "step": 2769
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.508771929824563e-06,
+      "loss": 1.0711,
+      "step": 2770
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.508223684210527e-06,
+      "loss": 0.8659,
+      "step": 2771
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.507675438596492e-06,
+      "loss": 1.0428,
+      "step": 2772
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.507127192982458e-06,
+      "loss": 1.6351,
+      "step": 2773
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.506578947368422e-06,
+      "loss": 0.5823,
+      "step": 2774
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.506030701754386e-06,
+      "loss": 0.7343,
+      "step": 2775
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.505482456140352e-06,
+      "loss": 1.1199,
+      "step": 2776
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.504934210526317e-06,
+      "loss": 0.7589,
+      "step": 2777
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.504385964912281e-06,
+      "loss": 0.7143,
+      "step": 2778
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.503837719298247e-06,
+      "loss": 0.5168,
+      "step": 2779
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.503289473684211e-06,
+      "loss": 1.2653,
+      "step": 2780
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.502741228070177e-06,
+      "loss": 1.4149,
+      "step": 2781
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.50219298245614e-06,
+      "loss": 0.58,
+      "step": 2782
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.501644736842106e-06,
+      "loss": 1.345,
+      "step": 2783
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.501096491228072e-06,
+      "loss": 0.7998,
+      "step": 2784
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.500548245614036e-06,
+      "loss": 1.1183,
+      "step": 2785
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.5e-06,
+      "loss": 0.7036,
+      "step": 2786
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.499451754385966e-06,
+      "loss": 0.7658,
+      "step": 2787
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.498903508771931e-06,
+      "loss": 0.634,
+      "step": 2788
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.498355263157895e-06,
+      "loss": 0.6269,
+      "step": 2789
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.49780701754386e-06,
+      "loss": 1.3985,
+      "step": 2790
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.497258771929825e-06,
+      "loss": 1.0048,
+      "step": 2791
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.49671052631579e-06,
+      "loss": 1.1959,
+      "step": 2792
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.496162280701755e-06,
+      "loss": 0.6479,
+      "step": 2793
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.49561403508772e-06,
+      "loss": 0.9069,
+      "step": 2794
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.495065789473686e-06,
+      "loss": 0.7723,
+      "step": 2795
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.49451754385965e-06,
+      "loss": 0.6556,
+      "step": 2796
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.493969298245614e-06,
+      "loss": 0.9084,
+      "step": 2797
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.49342105263158e-06,
+      "loss": 1.0061,
+      "step": 2798
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.492872807017545e-06,
+      "loss": 0.8136,
+      "step": 2799
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.49232456140351e-06,
+      "loss": 1.0887,
+      "step": 2800
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.491776315789473e-06,
+      "loss": 0.8308,
+      "step": 2801
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.491228070175439e-06,
+      "loss": 0.9567,
+      "step": 2802
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.490679824561404e-06,
+      "loss": 1.2067,
+      "step": 2803
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.490131578947368e-06,
+      "loss": 1.0687,
+      "step": 2804
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.489583333333334e-06,
+      "loss": 0.8727,
+      "step": 2805
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.489035087719298e-06,
+      "loss": 0.7378,
+      "step": 2806
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.488486842105264e-06,
+      "loss": 0.5213,
+      "step": 2807
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.48793859649123e-06,
+      "loss": 0.7721,
+      "step": 2808
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.487390350877193e-06,
+      "loss": 1.3524,
+      "step": 2809
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.486842105263159e-06,
+      "loss": 0.9982,
+      "step": 2810
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.486293859649125e-06,
+      "loss": 1.3934,
+      "step": 2811
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.485745614035089e-06,
+      "loss": 1.05,
+      "step": 2812
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.485197368421053e-06,
+      "loss": 1.4873,
+      "step": 2813
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.484649122807018e-06,
+      "loss": 1.3732,
+      "step": 2814
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.484100877192984e-06,
+      "loss": 1.0424,
+      "step": 2815
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.483552631578948e-06,
+      "loss": 1.1779,
+      "step": 2816
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.483004385964912e-06,
+      "loss": 1.183,
+      "step": 2817
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.482456140350878e-06,
+      "loss": 0.8304,
+      "step": 2818
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.481907894736843e-06,
+      "loss": 1.122,
+      "step": 2819
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.481359649122807e-06,
+      "loss": 0.8047,
+      "step": 2820
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.480811403508773e-06,
+      "loss": 0.6489,
+      "step": 2821
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.480263157894739e-06,
+      "loss": 1.5399,
+      "step": 2822
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.479714912280703e-06,
+      "loss": 1.9311,
+      "step": 2823
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.479166666666667e-06,
+      "loss": 1.5299,
+      "step": 2824
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.478618421052632e-06,
+      "loss": 1.3395,
+      "step": 2825
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.478070175438598e-06,
+      "loss": 1.4994,
+      "step": 2826
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.477521929824562e-06,
+      "loss": 1.0378,
+      "step": 2827
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.476973684210526e-06,
+      "loss": 1.4684,
+      "step": 2828
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.476425438596492e-06,
+      "loss": 0.6593,
+      "step": 2829
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.475877192982457e-06,
+      "loss": 1.0003,
+      "step": 2830
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.475328947368421e-06,
+      "loss": 0.9585,
+      "step": 2831
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.474780701754387e-06,
+      "loss": 0.8674,
+      "step": 2832
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.474232456140351e-06,
+      "loss": 0.7772,
+      "step": 2833
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.473684210526317e-06,
+      "loss": 0.9057,
+      "step": 2834
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.47313596491228e-06,
+      "loss": 0.8103,
+      "step": 2835
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.472587719298246e-06,
+      "loss": 0.8906,
+      "step": 2836
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.472039473684212e-06,
+      "loss": 1.3209,
+      "step": 2837
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.471491228070176e-06,
+      "loss": 1.0034,
+      "step": 2838
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.47094298245614e-06,
+      "loss": 0.7882,
+      "step": 2839
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.470394736842106e-06,
+      "loss": 1.1146,
+      "step": 2840
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.469846491228071e-06,
+      "loss": 1.327,
+      "step": 2841
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.469298245614035e-06,
+      "loss": 1.2306,
+      "step": 2842
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.468750000000001e-06,
+      "loss": 1.5321,
+      "step": 2843
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.468201754385965e-06,
+      "loss": 0.5537,
+      "step": 2844
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.46765350877193e-06,
+      "loss": 0.9383,
+      "step": 2845
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.467105263157896e-06,
+      "loss": 1.5305,
+      "step": 2846
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.46655701754386e-06,
+      "loss": 1.1195,
+      "step": 2847
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.466008771929826e-06,
+      "loss": 1.3863,
+      "step": 2848
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.46546052631579e-06,
+      "loss": 0.8216,
+      "step": 2849
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.464912280701755e-06,
+      "loss": 0.8139,
+      "step": 2850
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.46436403508772e-06,
+      "loss": 1.1968,
+      "step": 2851
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.463815789473685e-06,
+      "loss": 0.5921,
+      "step": 2852
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.46326754385965e-06,
+      "loss": 0.7702,
+      "step": 2853
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.462719298245615e-06,
+      "loss": 0.9857,
+      "step": 2854
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.462171052631579e-06,
+      "loss": 1.2316,
+      "step": 2855
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.461622807017544e-06,
+      "loss": 1.9266,
+      "step": 2856
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.46107456140351e-06,
+      "loss": 0.7789,
+      "step": 2857
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.460526315789474e-06,
+      "loss": 0.7342,
+      "step": 2858
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.45997807017544e-06,
+      "loss": 0.7146,
+      "step": 2859
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.459429824561404e-06,
+      "loss": 1.0076,
+      "step": 2860
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.45888157894737e-06,
+      "loss": 0.9955,
+      "step": 2861
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.458333333333333e-06,
+      "loss": 1.4595,
+      "step": 2862
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.457785087719299e-06,
+      "loss": 0.9638,
+      "step": 2863
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.457236842105265e-06,
+      "loss": 0.9329,
+      "step": 2864
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.456688596491229e-06,
+      "loss": 1.2202,
+      "step": 2865
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.456140350877193e-06,
+      "loss": 0.7276,
+      "step": 2866
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.455592105263158e-06,
+      "loss": 0.6312,
+      "step": 2867
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.455043859649124e-06,
+      "loss": 1.4722,
+      "step": 2868
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.454495614035088e-06,
+      "loss": 0.7864,
+      "step": 2869
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.453947368421054e-06,
+      "loss": 1.3304,
+      "step": 2870
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.453399122807018e-06,
+      "loss": 0.8828,
+      "step": 2871
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.452850877192983e-06,
+      "loss": 0.7106,
+      "step": 2872
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.452302631578947e-06,
+      "loss": 1.3294,
+      "step": 2873
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.451754385964913e-06,
+      "loss": 0.8576,
+      "step": 2874
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.451206140350879e-06,
+      "loss": 0.5385,
+      "step": 2875
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.450657894736843e-06,
+      "loss": 0.8295,
+      "step": 2876
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.450109649122807e-06,
+      "loss": 0.6619,
+      "step": 2877
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.449561403508772e-06,
+      "loss": 1.0477,
+      "step": 2878
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.449013157894738e-06,
+      "loss": 0.8324,
+      "step": 2879
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.448464912280702e-06,
+      "loss": 1.3819,
+      "step": 2880
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.447916666666668e-06,
+      "loss": 0.8143,
+      "step": 2881
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.447368421052632e-06,
+      "loss": 1.0725,
+      "step": 2882
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.446820175438597e-06,
+      "loss": 0.7943,
+      "step": 2883
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.446271929824563e-06,
+      "loss": 0.8939,
+      "step": 2884
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.445723684210527e-06,
+      "loss": 0.6095,
+      "step": 2885
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.445175438596493e-06,
+      "loss": 1.0247,
+      "step": 2886
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.444627192982457e-06,
+      "loss": 0.9175,
+      "step": 2887
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.444078947368422e-06,
+      "loss": 1.6773,
+      "step": 2888
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.443530701754386e-06,
+      "loss": 1.2018,
+      "step": 2889
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.442982456140352e-06,
+      "loss": 0.9235,
+      "step": 2890
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.442434210526318e-06,
+      "loss": 1.367,
+      "step": 2891
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.441885964912282e-06,
+      "loss": 0.8919,
+      "step": 2892
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.441337719298245e-06,
+      "loss": 1.8344,
+      "step": 2893
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.440789473684211e-06,
+      "loss": 1.1526,
+      "step": 2894
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.440241228070177e-06,
+      "loss": 1.0155,
+      "step": 2895
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.43969298245614e-06,
+      "loss": 1.0417,
+      "step": 2896
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.439144736842106e-06,
+      "loss": 0.9051,
+      "step": 2897
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.43859649122807e-06,
+      "loss": 1.2941,
+      "step": 2898
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.438048245614036e-06,
+      "loss": 0.9917,
+      "step": 2899
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.4375e-06,
+      "loss": 0.5388,
+      "step": 2900
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.436951754385966e-06,
+      "loss": 1.0861,
+      "step": 2901
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.436403508771931e-06,
+      "loss": 1.1533,
+      "step": 2902
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.435855263157895e-06,
+      "loss": 0.4849,
+      "step": 2903
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.43530701754386e-06,
+      "loss": 0.5366,
+      "step": 2904
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.434758771929825e-06,
+      "loss": 0.8381,
+      "step": 2905
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.43421052631579e-06,
+      "loss": 0.8345,
+      "step": 2906
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.433662280701755e-06,
+      "loss": 0.7279,
+      "step": 2907
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.43311403508772e-06,
+      "loss": 0.8088,
+      "step": 2908
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.432565789473684e-06,
+      "loss": 0.7857,
+      "step": 2909
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.43201754385965e-06,
+      "loss": 1.138,
+      "step": 2910
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.431469298245614e-06,
+      "loss": 1.1587,
+      "step": 2911
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.43092105263158e-06,
+      "loss": 1.4714,
+      "step": 2912
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.430372807017545e-06,
+      "loss": 0.7551,
+      "step": 2913
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.42982456140351e-06,
+      "loss": 1.0544,
+      "step": 2914
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.429276315789473e-06,
+      "loss": 1.3004,
+      "step": 2915
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.428728070175439e-06,
+      "loss": 1.305,
+      "step": 2916
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.428179824561405e-06,
+      "loss": 1.6701,
+      "step": 2917
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.42763157894737e-06,
+      "loss": 0.9932,
+      "step": 2918
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.427083333333334e-06,
+      "loss": 1.2618,
+      "step": 2919
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.426535087719298e-06,
+      "loss": 0.9359,
+      "step": 2920
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.425986842105264e-06,
+      "loss": 0.9511,
+      "step": 2921
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.42543859649123e-06,
+      "loss": 0.7523,
+      "step": 2922
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.424890350877194e-06,
+      "loss": 0.9838,
+      "step": 2923
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.42434210526316e-06,
+      "loss": 1.6823,
+      "step": 2924
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.423793859649123e-06,
+      "loss": 1.149,
+      "step": 2925
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.423245614035089e-06,
+      "loss": 0.8092,
+      "step": 2926
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.422697368421053e-06,
+      "loss": 0.5145,
+      "step": 2927
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.422149122807019e-06,
+      "loss": 0.9736,
+      "step": 2928
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.421600877192984e-06,
+      "loss": 1.3525,
+      "step": 2929
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.421052631578948e-06,
+      "loss": 1.1122,
+      "step": 2930
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.420504385964912e-06,
+      "loss": 0.7955,
+      "step": 2931
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.419956140350878e-06,
+      "loss": 1.5103,
+      "step": 2932
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.419407894736844e-06,
+      "loss": 1.3499,
+      "step": 2933
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.418859649122808e-06,
+      "loss": 1.109,
+      "step": 2934
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.418311403508772e-06,
+      "loss": 0.8176,
+      "step": 2935
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.417763157894737e-06,
+      "loss": 1.121,
+      "step": 2936
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.417214912280703e-06,
+      "loss": 0.8134,
+      "step": 2937
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.416666666666667e-06,
+      "loss": 0.6402,
+      "step": 2938
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.416118421052633e-06,
+      "loss": 1.7074,
+      "step": 2939
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.415570175438598e-06,
+      "loss": 1.287,
+      "step": 2940
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.415021929824562e-06,
+      "loss": 0.995,
+      "step": 2941
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.414473684210526e-06,
+      "loss": 1.6427,
+      "step": 2942
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.413925438596492e-06,
+      "loss": 0.6077,
+      "step": 2943
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.413377192982457e-06,
+      "loss": 1.017,
+      "step": 2944
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.412828947368421e-06,
+      "loss": 1.0273,
+      "step": 2945
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.412280701754385e-06,
+      "loss": 0.7822,
+      "step": 2946
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.411732456140351e-06,
+      "loss": 1.2545,
+      "step": 2947
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.411184210526317e-06,
+      "loss": 0.6181,
+      "step": 2948
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.41063596491228e-06,
+      "loss": 1.3046,
+      "step": 2949
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.410087719298246e-06,
+      "loss": 1.382,
+      "step": 2950
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.409539473684212e-06,
+      "loss": 1.7152,
+      "step": 2951
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.408991228070176e-06,
+      "loss": 1.2394,
+      "step": 2952
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.40844298245614e-06,
+      "loss": 1.1755,
+      "step": 2953
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.407894736842106e-06,
+      "loss": 1.3718,
+      "step": 2954
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.407346491228071e-06,
+      "loss": 1.1649,
+      "step": 2955
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.406798245614037e-06,
+      "loss": 1.365,
+      "step": 2956
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.406250000000001e-06,
+      "loss": 1.3217,
+      "step": 2957
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.405701754385965e-06,
+      "loss": 0.5725,
+      "step": 2958
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.40515350877193e-06,
+      "loss": 1.2375,
+      "step": 2959
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.404605263157896e-06,
+      "loss": 1.168,
+      "step": 2960
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.40405701754386e-06,
+      "loss": 0.7513,
+      "step": 2961
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.403508771929824e-06,
+      "loss": 1.6619,
+      "step": 2962
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.40296052631579e-06,
+      "loss": 1.0423,
+      "step": 2963
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.402412280701756e-06,
+      "loss": 1.3586,
+      "step": 2964
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.40186403508772e-06,
+      "loss": 0.8153,
+      "step": 2965
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.401315789473685e-06,
+      "loss": 0.9125,
+      "step": 2966
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.400767543859651e-06,
+      "loss": 1.0041,
+      "step": 2967
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.400219298245615e-06,
+      "loss": 1.0244,
+      "step": 2968
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.399671052631579e-06,
+      "loss": 0.7778,
+      "step": 2969
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.399122807017545e-06,
+      "loss": 1.1799,
+      "step": 2970
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.39857456140351e-06,
+      "loss": 1.3501,
+      "step": 2971
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.398026315789474e-06,
+      "loss": 1.218,
+      "step": 2972
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.397478070175438e-06,
+      "loss": 1.2072,
+      "step": 2973
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.396929824561404e-06,
+      "loss": 1.0833,
+      "step": 2974
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.39638157894737e-06,
+      "loss": 1.18,
+      "step": 2975
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.395833333333334e-06,
+      "loss": 0.9486,
+      "step": 2976
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.3952850877193e-06,
+      "loss": 1.0519,
+      "step": 2977
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.394736842105263e-06,
+      "loss": 0.6111,
+      "step": 2978
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.394188596491229e-06,
+      "loss": 0.8546,
+      "step": 2979
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.393640350877193e-06,
+      "loss": 0.8231,
+      "step": 2980
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.393092105263159e-06,
+      "loss": 1.2088,
+      "step": 2981
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.392543859649124e-06,
+      "loss": 0.6161,
+      "step": 2982
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.391995614035088e-06,
+      "loss": 1.2738,
+      "step": 2983
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.391447368421052e-06,
+      "loss": 1.3202,
+      "step": 2984
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.390899122807018e-06,
+      "loss": 0.5734,
+      "step": 2985
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.390350877192983e-06,
+      "loss": 0.8927,
+      "step": 2986
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.389802631578947e-06,
+      "loss": 0.6626,
+      "step": 2987
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.389254385964913e-06,
+      "loss": 0.9655,
+      "step": 2988
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.388706140350877e-06,
+      "loss": 1.7246,
+      "step": 2989
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.388157894736843e-06,
+      "loss": 1.3026,
+      "step": 2990
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.387609649122807e-06,
+      "loss": 1.055,
+      "step": 2991
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.387061403508772e-06,
+      "loss": 0.9783,
+      "step": 2992
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.386513157894738e-06,
+      "loss": 1.0884,
+      "step": 2993
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.385964912280704e-06,
+      "loss": 0.7378,
+      "step": 2994
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.385416666666668e-06,
+      "loss": 0.6509,
+      "step": 2995
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.384868421052632e-06,
+      "loss": 0.9751,
+      "step": 2996
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.384320175438597e-06,
+      "loss": 0.6274,
+      "step": 2997
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.383771929824563e-06,
+      "loss": 1.2068,
+      "step": 2998
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.383223684210527e-06,
+      "loss": 0.6922,
+      "step": 2999
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.382675438596491e-06,
+      "loss": 0.873,
+      "step": 3000
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.382127192982457e-06,
+      "loss": 1.0664,
+      "step": 3001
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.381578947368422e-06,
+      "loss": 1.0427,
+      "step": 3002
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.381030701754386e-06,
+      "loss": 0.7553,
+      "step": 3003
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.380482456140352e-06,
+      "loss": 1.5709,
+      "step": 3004
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.379934210526316e-06,
+      "loss": 0.9516,
+      "step": 3005
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.379385964912282e-06,
+      "loss": 0.7072,
+      "step": 3006
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.378837719298246e-06,
+      "loss": 1.0449,
+      "step": 3007
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.378289473684211e-06,
+      "loss": 1.1634,
+      "step": 3008
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.377741228070177e-06,
+      "loss": 1.0209,
+      "step": 3009
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.377192982456141e-06,
+      "loss": 0.7184,
+      "step": 3010
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.376644736842105e-06,
+      "loss": 0.7919,
+      "step": 3011
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.37609649122807e-06,
+      "loss": 1.2223,
+      "step": 3012
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.375548245614036e-06,
+      "loss": 1.1703,
+      "step": 3013
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.375e-06,
+      "loss": 1.1254,
+      "step": 3014
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.374451754385966e-06,
+      "loss": 1.2622,
+      "step": 3015
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.37390350877193e-06,
+      "loss": 0.7683,
+      "step": 3016
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.373355263157896e-06,
+      "loss": 0.7067,
+      "step": 3017
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.37280701754386e-06,
+      "loss": 0.9652,
+      "step": 3018
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.372258771929825e-06,
+      "loss": 0.7607,
+      "step": 3019
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.371710526315791e-06,
+      "loss": 1.2855,
+      "step": 3020
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.371162280701755e-06,
+      "loss": 0.9861,
+      "step": 3021
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.370614035087719e-06,
+      "loss": 1.3621,
+      "step": 3022
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.370065789473685e-06,
+      "loss": 0.6355,
+      "step": 3023
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.36951754385965e-06,
+      "loss": 0.7882,
+      "step": 3024
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.368969298245614e-06,
+      "loss": 0.8133,
+      "step": 3025
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.36842105263158e-06,
+      "loss": 1.1051,
+      "step": 3026
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.367872807017544e-06,
+      "loss": 0.7962,
+      "step": 3027
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.36732456140351e-06,
+      "loss": 0.923,
+      "step": 3028
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.366776315789475e-06,
+      "loss": 0.4774,
+      "step": 3029
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.36622807017544e-06,
+      "loss": 1.2936,
+      "step": 3030
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.365679824561405e-06,
+      "loss": 0.8185,
+      "step": 3031
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.365131578947369e-06,
+      "loss": 0.8824,
+      "step": 3032
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.364583333333334e-06,
+      "loss": 1.3825,
+      "step": 3033
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.364035087719298e-06,
+      "loss": 1.3144,
+      "step": 3034
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.363486842105264e-06,
+      "loss": 1.0774,
+      "step": 3035
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.36293859649123e-06,
+      "loss": 1.2098,
+      "step": 3036
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.362390350877194e-06,
+      "loss": 0.7466,
+      "step": 3037
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.361842105263158e-06,
+      "loss": 1.282,
+      "step": 3038
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.361293859649123e-06,
+      "loss": 1.028,
+      "step": 3039
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.360745614035089e-06,
+      "loss": 1.6738,
+      "step": 3040
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.360197368421053e-06,
+      "loss": 1.2915,
+      "step": 3041
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.359649122807019e-06,
+      "loss": 1.0968,
+      "step": 3042
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.359100877192983e-06,
+      "loss": 0.687,
+      "step": 3043
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.358552631578948e-06,
+      "loss": 0.7747,
+      "step": 3044
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.358004385964912e-06,
+      "loss": 1.1732,
+      "step": 3045
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.357456140350878e-06,
+      "loss": 1.2085,
+      "step": 3046
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.356907894736844e-06,
+      "loss": 1.3247,
+      "step": 3047
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.356359649122808e-06,
+      "loss": 0.8578,
+      "step": 3048
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.355811403508772e-06,
+      "loss": 0.5922,
+      "step": 3049
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.355263157894737e-06,
+      "loss": 0.8768,
+      "step": 3050
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.354714912280703e-06,
+      "loss": 1.0553,
+      "step": 3051
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.354166666666667e-06,
+      "loss": 0.961,
+      "step": 3052
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.353618421052633e-06,
+      "loss": 0.811,
+      "step": 3053
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.353070175438597e-06,
+      "loss": 1.4822,
+      "step": 3054
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.352521929824562e-06,
+      "loss": 0.9705,
+      "step": 3055
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.351973684210526e-06,
+      "loss": 1.1758,
+      "step": 3056
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.351425438596492e-06,
+      "loss": 0.4026,
+      "step": 3057
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.350877192982458e-06,
+      "loss": 0.9627,
+      "step": 3058
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.350328947368422e-06,
+      "loss": 1.0797,
+      "step": 3059
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.349780701754386e-06,
+      "loss": 0.7852,
+      "step": 3060
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.349232456140351e-06,
+      "loss": 0.9118,
+      "step": 3061
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.348684210526317e-06,
+      "loss": 1.0896,
+      "step": 3062
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.348135964912281e-06,
+      "loss": 0.3703,
+      "step": 3063
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.347587719298247e-06,
+      "loss": 0.7462,
+      "step": 3064
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.34703947368421e-06,
+      "loss": 0.7393,
+      "step": 3065
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.346491228070176e-06,
+      "loss": 0.8629,
+      "step": 3066
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.345942982456142e-06,
+      "loss": 0.8932,
+      "step": 3067
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.345394736842106e-06,
+      "loss": 0.7643,
+      "step": 3068
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.344846491228072e-06,
+      "loss": 1.736,
+      "step": 3069
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.344298245614036e-06,
+      "loss": 0.7428,
+      "step": 3070
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.343750000000001e-06,
+      "loss": 0.6108,
+      "step": 3071
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.343201754385965e-06,
+      "loss": 1.3784,
+      "step": 3072
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.342653508771931e-06,
+      "loss": 1.1986,
+      "step": 3073
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.342105263157897e-06,
+      "loss": 1.1713,
+      "step": 3074
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.34155701754386e-06,
+      "loss": 1.0169,
+      "step": 3075
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.341008771929824e-06,
+      "loss": 0.7906,
+      "step": 3076
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.34046052631579e-06,
+      "loss": 1.0197,
+      "step": 3077
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.339912280701756e-06,
+      "loss": 1.1237,
+      "step": 3078
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.33936403508772e-06,
+      "loss": 1.0592,
+      "step": 3079
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.338815789473685e-06,
+      "loss": 0.9096,
+      "step": 3080
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.33826754385965e-06,
+      "loss": 0.8981,
+      "step": 3081
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.337719298245615e-06,
+      "loss": 0.8086,
+      "step": 3082
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.337171052631579e-06,
+      "loss": 1.7188,
+      "step": 3083
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.336622807017545e-06,
+      "loss": 0.9907,
+      "step": 3084
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.33607456140351e-06,
+      "loss": 1.2808,
+      "step": 3085
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.335526315789474e-06,
+      "loss": 1.1744,
+      "step": 3086
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.334978070175438e-06,
+      "loss": 1.2843,
+      "step": 3087
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.334429824561404e-06,
+      "loss": 0.7088,
+      "step": 3088
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.33388157894737e-06,
+      "loss": 1.3684,
+      "step": 3089
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 0.8791,
+      "step": 3090
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.332785087719298e-06,
+      "loss": 1.1215,
+      "step": 3091
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.332236842105263e-06,
+      "loss": 0.6991,
+      "step": 3092
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.331688596491229e-06,
+      "loss": 1.4992,
+      "step": 3093
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.331140350877193e-06,
+      "loss": 1.0868,
+      "step": 3094
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.330592105263159e-06,
+      "loss": 1.0591,
+      "step": 3095
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.330043859649124e-06,
+      "loss": 1.1836,
+      "step": 3096
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.329495614035088e-06,
+      "loss": 0.618,
+      "step": 3097
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.328947368421052e-06,
+      "loss": 1.1031,
+      "step": 3098
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.328399122807018e-06,
+      "loss": 0.8862,
+      "step": 3099
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.327850877192984e-06,
+      "loss": 0.5809,
+      "step": 3100
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.327302631578948e-06,
+      "loss": 1.1003,
+      "step": 3101
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.326754385964913e-06,
+      "loss": 1.0888,
+      "step": 3102
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.326206140350877e-06,
+      "loss": 0.9642,
+      "step": 3103
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.325657894736843e-06,
+      "loss": 1.4013,
+      "step": 3104
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.325109649122809e-06,
+      "loss": 1.1154,
+      "step": 3105
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.324561403508773e-06,
+      "loss": 1.7882,
+      "step": 3106
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.324013157894738e-06,
+      "loss": 1.071,
+      "step": 3107
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.323464912280702e-06,
+      "loss": 1.0892,
+      "step": 3108
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.322916666666668e-06,
+      "loss": 1.2602,
+      "step": 3109
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.322368421052632e-06,
+      "loss": 1.1208,
+      "step": 3110
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.321820175438598e-06,
+      "loss": 1.1171,
+      "step": 3111
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.321271929824563e-06,
+      "loss": 0.3186,
+      "step": 3112
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.320723684210527e-06,
+      "loss": 0.9366,
+      "step": 3113
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.320175438596491e-06,
+      "loss": 1.5412,
+      "step": 3114
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.319627192982457e-06,
+      "loss": 0.9262,
+      "step": 3115
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.319078947368423e-06,
+      "loss": 1.0669,
+      "step": 3116
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.318530701754387e-06,
+      "loss": 1.0599,
+      "step": 3117
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.31798245614035e-06,
+      "loss": 1.3886,
+      "step": 3118
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.317434210526316e-06,
+      "loss": 1.2216,
+      "step": 3119
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.316885964912282e-06,
+      "loss": 1.2303,
+      "step": 3120
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.316337719298246e-06,
+      "loss": 1.068,
+      "step": 3121
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.315789473684212e-06,
+      "loss": 0.6382,
+      "step": 3122
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.315241228070177e-06,
+      "loss": 1.4184,
+      "step": 3123
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.314692982456141e-06,
+      "loss": 0.8974,
+      "step": 3124
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.314144736842105e-06,
+      "loss": 0.8452,
+      "step": 3125
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.31359649122807e-06,
+      "loss": 1.276,
+      "step": 3126
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.313048245614036e-06,
+      "loss": 0.9736,
+      "step": 3127
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.3125e-06,
+      "loss": 1.4103,
+      "step": 3128
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.311951754385964e-06,
+      "loss": 0.9205,
+      "step": 3129
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.31140350877193e-06,
+      "loss": 0.6654,
+      "step": 3130
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.310855263157896e-06,
+      "loss": 1.1271,
+      "step": 3131
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.31030701754386e-06,
+      "loss": 0.7878,
+      "step": 3132
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.309758771929825e-06,
+      "loss": 1.1705,
+      "step": 3133
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.30921052631579e-06,
+      "loss": 1.2133,
+      "step": 3134
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.308662280701755e-06,
+      "loss": 0.8346,
+      "step": 3135
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.308114035087719e-06,
+      "loss": 1.2087,
+      "step": 3136
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.307565789473685e-06,
+      "loss": 1.5163,
+      "step": 3137
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.30701754385965e-06,
+      "loss": 0.7557,
+      "step": 3138
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.306469298245614e-06,
+      "loss": 0.4979,
+      "step": 3139
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.30592105263158e-06,
+      "loss": 1.1828,
+      "step": 3140
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.305372807017544e-06,
+      "loss": 1.0492,
+      "step": 3141
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.30482456140351e-06,
+      "loss": 0.7724,
+      "step": 3142
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.304276315789475e-06,
+      "loss": 1.2191,
+      "step": 3143
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.30372807017544e-06,
+      "loss": 1.6779,
+      "step": 3144
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.303179824561403e-06,
+      "loss": 0.7024,
+      "step": 3145
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.302631578947369e-06,
+      "loss": 0.9287,
+      "step": 3146
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.302083333333335e-06,
+      "loss": 0.9625,
+      "step": 3147
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.301535087719299e-06,
+      "loss": 0.6816,
+      "step": 3148
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.300986842105264e-06,
+      "loss": 0.8402,
+      "step": 3149
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.30043859649123e-06,
+      "loss": 0.6393,
+      "step": 3150
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.299890350877194e-06,
+      "loss": 1.0716,
+      "step": 3151
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.299342105263158e-06,
+      "loss": 0.609,
+      "step": 3152
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.298793859649124e-06,
+      "loss": 0.965,
+      "step": 3153
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.29824561403509e-06,
+      "loss": 1.0337,
+      "step": 3154
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.297697368421053e-06,
+      "loss": 1.3178,
+      "step": 3155
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.297149122807017e-06,
+      "loss": 1.0408,
+      "step": 3156
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.296600877192983e-06,
+      "loss": 0.3202,
+      "step": 3157
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.296052631578949e-06,
+      "loss": 0.7336,
+      "step": 3158
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.295504385964913e-06,
+      "loss": 0.6114,
+      "step": 3159
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.294956140350878e-06,
+      "loss": 1.224,
+      "step": 3160
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.294407894736842e-06,
+      "loss": 0.9144,
+      "step": 3161
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.293859649122808e-06,
+      "loss": 0.9996,
+      "step": 3162
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.293311403508772e-06,
+      "loss": 0.7944,
+      "step": 3163
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.292763157894738e-06,
+      "loss": 1.1097,
+      "step": 3164
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.292214912280703e-06,
+      "loss": 1.2142,
+      "step": 3165
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.291666666666667e-06,
+      "loss": 0.9477,
+      "step": 3166
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.291118421052631e-06,
+      "loss": 1.453,
+      "step": 3167
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.290570175438597e-06,
+      "loss": 1.0543,
+      "step": 3168
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.290021929824563e-06,
+      "loss": 0.5032,
+      "step": 3169
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.289473684210526e-06,
+      "loss": 0.5911,
+      "step": 3170
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.288925438596492e-06,
+      "loss": 1.0326,
+      "step": 3171
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.288377192982456e-06,
+      "loss": 0.9394,
+      "step": 3172
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.287828947368422e-06,
+      "loss": 0.9642,
+      "step": 3173
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.287280701754386e-06,
+      "loss": 1.2918,
+      "step": 3174
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.286732456140351e-06,
+      "loss": 0.8222,
+      "step": 3175
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.286184210526317e-06,
+      "loss": 0.6963,
+      "step": 3176
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.285635964912281e-06,
+      "loss": 0.6131,
+      "step": 3177
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.285087719298247e-06,
+      "loss": 1.6902,
+      "step": 3178
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.28453947368421e-06,
+      "loss": 0.3653,
+      "step": 3179
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.283991228070176e-06,
+      "loss": 0.7815,
+      "step": 3180
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.283442982456142e-06,
+      "loss": 1.0512,
+      "step": 3181
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.282894736842106e-06,
+      "loss": 1.3314,
+      "step": 3182
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.28234649122807e-06,
+      "loss": 0.6755,
+      "step": 3183
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.281798245614036e-06,
+      "loss": 1.1354,
+      "step": 3184
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.281250000000001e-06,
+      "loss": 1.8024,
+      "step": 3185
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.280701754385965e-06,
+      "loss": 0.5187,
+      "step": 3186
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.280153508771931e-06,
+      "loss": 1.0032,
+      "step": 3187
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.279605263157895e-06,
+      "loss": 1.0715,
+      "step": 3188
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.27905701754386e-06,
+      "loss": 1.1196,
+      "step": 3189
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.278508771929825e-06,
+      "loss": 0.5756,
+      "step": 3190
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.27796052631579e-06,
+      "loss": 0.7296,
+      "step": 3191
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.277412280701756e-06,
+      "loss": 0.9316,
+      "step": 3192
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.27686403508772e-06,
+      "loss": 1.562,
+      "step": 3193
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.276315789473684e-06,
+      "loss": 1.2982,
+      "step": 3194
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.27576754385965e-06,
+      "loss": 1.395,
+      "step": 3195
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.275219298245615e-06,
+      "loss": 0.5285,
+      "step": 3196
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.27467105263158e-06,
+      "loss": 1.4477,
+      "step": 3197
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.274122807017545e-06,
+      "loss": 0.7004,
+      "step": 3198
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.273574561403509e-06,
+      "loss": 1.2221,
+      "step": 3199
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.273026315789475e-06,
+      "loss": 1.4695,
+      "step": 3200
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.272478070175439e-06,
+      "loss": 0.5974,
+      "step": 3201
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.271929824561404e-06,
+      "loss": 1.4941,
+      "step": 3202
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.27138157894737e-06,
+      "loss": 0.9277,
+      "step": 3203
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.270833333333334e-06,
+      "loss": 0.628,
+      "step": 3204
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.270285087719298e-06,
+      "loss": 0.893,
+      "step": 3205
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.269736842105264e-06,
+      "loss": 0.7275,
+      "step": 3206
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.26918859649123e-06,
+      "loss": 1.3254,
+      "step": 3207
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.268640350877193e-06,
+      "loss": 0.9049,
+      "step": 3208
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.268092105263159e-06,
+      "loss": 0.9921,
+      "step": 3209
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.267543859649123e-06,
+      "loss": 2.1111,
+      "step": 3210
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.266995614035089e-06,
+      "loss": 1.0748,
+      "step": 3211
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.266447368421053e-06,
+      "loss": 1.6469,
+      "step": 3212
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.265899122807018e-06,
+      "loss": 1.0801,
+      "step": 3213
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.265350877192984e-06,
+      "loss": 0.5915,
+      "step": 3214
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.264802631578948e-06,
+      "loss": 0.7673,
+      "step": 3215
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.264254385964913e-06,
+      "loss": 0.6155,
+      "step": 3216
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.263706140350877e-06,
+      "loss": 1.3893,
+      "step": 3217
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.263157894736843e-06,
+      "loss": 0.584,
+      "step": 3218
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.262609649122809e-06,
+      "loss": 1.7281,
+      "step": 3219
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.262061403508773e-06,
+      "loss": 0.8053,
+      "step": 3220
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.261513157894737e-06,
+      "loss": 0.8164,
+      "step": 3221
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.260964912280702e-06,
+      "loss": 1.1448,
+      "step": 3222
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.260416666666668e-06,
+      "loss": 0.8381,
+      "step": 3223
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.259868421052632e-06,
+      "loss": 0.8131,
+      "step": 3224
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.259320175438598e-06,
+      "loss": 0.6842,
+      "step": 3225
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.258771929824562e-06,
+      "loss": 0.4928,
+      "step": 3226
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.258223684210527e-06,
+      "loss": 0.6541,
+      "step": 3227
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.257675438596491e-06,
+      "loss": 0.8784,
+      "step": 3228
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.257127192982457e-06,
+      "loss": 1.1052,
+      "step": 3229
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.256578947368423e-06,
+      "loss": 0.6676,
+      "step": 3230
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.256030701754387e-06,
+      "loss": 1.4604,
+      "step": 3231
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.25548245614035e-06,
+      "loss": 1.2938,
+      "step": 3232
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.254934210526316e-06,
+      "loss": 0.9694,
+      "step": 3233
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.254385964912282e-06,
+      "loss": 0.8851,
+      "step": 3234
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.253837719298246e-06,
+      "loss": 0.9079,
+      "step": 3235
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.253289473684212e-06,
+      "loss": 0.6215,
+      "step": 3236
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.252741228070176e-06,
+      "loss": 1.0562,
+      "step": 3237
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.252192982456141e-06,
+      "loss": 1.2798,
+      "step": 3238
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.251644736842105e-06,
+      "loss": 1.3177,
+      "step": 3239
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.251096491228071e-06,
+      "loss": 0.6736,
+      "step": 3240
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.250548245614037e-06,
+      "loss": 0.7455,
+      "step": 3241
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.25e-06,
+      "loss": 1.0854,
+      "step": 3242
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.249451754385965e-06,
+      "loss": 1.1956,
+      "step": 3243
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.24890350877193e-06,
+      "loss": 0.665,
+      "step": 3244
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.248355263157896e-06,
+      "loss": 1.1671,
+      "step": 3245
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.24780701754386e-06,
+      "loss": 0.7661,
+      "step": 3246
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.247258771929824e-06,
+      "loss": 0.797,
+      "step": 3247
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.24671052631579e-06,
+      "loss": 1.1143,
+      "step": 3248
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.246162280701755e-06,
+      "loss": 1.082,
+      "step": 3249
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.24561403508772e-06,
+      "loss": 1.2669,
+      "step": 3250
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.245065789473685e-06,
+      "loss": 0.6725,
+      "step": 3251
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.24451754385965e-06,
+      "loss": 0.7755,
+      "step": 3252
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.243969298245615e-06,
+      "loss": 0.87,
+      "step": 3253
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.24342105263158e-06,
+      "loss": 1.6082,
+      "step": 3254
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.242872807017544e-06,
+      "loss": 0.9837,
+      "step": 3255
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.24232456140351e-06,
+      "loss": 0.9585,
+      "step": 3256
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.241776315789476e-06,
+      "loss": 1.3748,
+      "step": 3257
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.24122807017544e-06,
+      "loss": 1.1347,
+      "step": 3258
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.240679824561403e-06,
+      "loss": 0.798,
+      "step": 3259
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.24013157894737e-06,
+      "loss": 1.0078,
+      "step": 3260
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.239583333333335e-06,
+      "loss": 1.0424,
+      "step": 3261
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.239035087719299e-06,
+      "loss": 0.6152,
+      "step": 3262
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.238486842105263e-06,
+      "loss": 0.689,
+      "step": 3263
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.237938596491228e-06,
+      "loss": 0.5364,
+      "step": 3264
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.237390350877194e-06,
+      "loss": 0.7782,
+      "step": 3265
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.236842105263158e-06,
+      "loss": 0.7771,
+      "step": 3266
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.236293859649124e-06,
+      "loss": 0.9413,
+      "step": 3267
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.23574561403509e-06,
+      "loss": 0.8427,
+      "step": 3268
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.235197368421053e-06,
+      "loss": 0.7157,
+      "step": 3269
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.234649122807017e-06,
+      "loss": 1.3286,
+      "step": 3270
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.234100877192983e-06,
+      "loss": 0.5166,
+      "step": 3271
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.233552631578949e-06,
+      "loss": 0.8643,
+      "step": 3272
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.233004385964913e-06,
+      "loss": 1.1509,
+      "step": 3273
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.232456140350877e-06,
+      "loss": 1.5058,
+      "step": 3274
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.231907894736842e-06,
+      "loss": 1.1084,
+      "step": 3275
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.231359649122808e-06,
+      "loss": 1.1544,
+      "step": 3276
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.230811403508772e-06,
+      "loss": 1.1185,
+      "step": 3277
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.230263157894738e-06,
+      "loss": 0.9526,
+      "step": 3278
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.229714912280703e-06,
+      "loss": 1.2767,
+      "step": 3279
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.229166666666667e-06,
+      "loss": 0.4963,
+      "step": 3280
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.228618421052631e-06,
+      "loss": 1.0979,
+      "step": 3281
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.228070175438597e-06,
+      "loss": 0.5587,
+      "step": 3282
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.227521929824563e-06,
+      "loss": 1.4994,
+      "step": 3283
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.226973684210527e-06,
+      "loss": 1.3745,
+      "step": 3284
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.22642543859649e-06,
+      "loss": 0.337,
+      "step": 3285
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.225877192982456e-06,
+      "loss": 0.5803,
+      "step": 3286
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.225328947368422e-06,
+      "loss": 0.9652,
+      "step": 3287
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.224780701754388e-06,
+      "loss": 0.7895,
+      "step": 3288
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.224232456140352e-06,
+      "loss": 1.2192,
+      "step": 3289
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.223684210526316e-06,
+      "loss": 0.9154,
+      "step": 3290
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.223135964912281e-06,
+      "loss": 0.9477,
+      "step": 3291
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.222587719298247e-06,
+      "loss": 0.528,
+      "step": 3292
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.222039473684211e-06,
+      "loss": 1.3605,
+      "step": 3293
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.221491228070177e-06,
+      "loss": 0.5892,
+      "step": 3294
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.220942982456142e-06,
+      "loss": 0.7152,
+      "step": 3295
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.220394736842106e-06,
+      "loss": 1.6383,
+      "step": 3296
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.21984649122807e-06,
+      "loss": 0.5425,
+      "step": 3297
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.219298245614036e-06,
+      "loss": 0.6536,
+      "step": 3298
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.218750000000002e-06,
+      "loss": 0.4798,
+      "step": 3299
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.218201754385966e-06,
+      "loss": 0.9963,
+      "step": 3300
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.21765350877193e-06,
+      "loss": 0.6891,
+      "step": 3301
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.217105263157895e-06,
+      "loss": 0.5859,
+      "step": 3302
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.216557017543861e-06,
+      "loss": 0.7442,
+      "step": 3303
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.216008771929825e-06,
+      "loss": 1.2107,
+      "step": 3304
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.21546052631579e-06,
+      "loss": 1.1693,
+      "step": 3305
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.214912280701754e-06,
+      "loss": 0.454,
+      "step": 3306
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.21436403508772e-06,
+      "loss": 1.4326,
+      "step": 3307
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.213815789473684e-06,
+      "loss": 1.0274,
+      "step": 3308
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.21326754385965e-06,
+      "loss": 0.8432,
+      "step": 3309
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.212719298245615e-06,
+      "loss": 0.6858,
+      "step": 3310
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.21217105263158e-06,
+      "loss": 1.2034,
+      "step": 3311
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.211622807017543e-06,
+      "loss": 1.0827,
+      "step": 3312
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.211074561403509e-06,
+      "loss": 0.7478,
+      "step": 3313
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.210526315789475e-06,
+      "loss": 0.6136,
+      "step": 3314
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.209978070175439e-06,
+      "loss": 0.6369,
+      "step": 3315
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.209429824561404e-06,
+      "loss": 0.7615,
+      "step": 3316
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.208881578947368e-06,
+      "loss": 1.0307,
+      "step": 3317
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.208333333333334e-06,
+      "loss": 0.9373,
+      "step": 3318
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.207785087719298e-06,
+      "loss": 1.0434,
+      "step": 3319
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.207236842105264e-06,
+      "loss": 1.3433,
+      "step": 3320
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.20668859649123e-06,
+      "loss": 1.0634,
+      "step": 3321
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.206140350877193e-06,
+      "loss": 1.3583,
+      "step": 3322
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.205592105263157e-06,
+      "loss": 1.1369,
+      "step": 3323
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.205043859649123e-06,
+      "loss": 1.113,
+      "step": 3324
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.204495614035089e-06,
+      "loss": 0.6595,
+      "step": 3325
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.203947368421054e-06,
+      "loss": 1.073,
+      "step": 3326
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.203399122807018e-06,
+      "loss": 0.9761,
+      "step": 3327
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.202850877192982e-06,
+      "loss": 0.3472,
+      "step": 3328
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.202302631578948e-06,
+      "loss": 1.0296,
+      "step": 3329
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.201754385964914e-06,
+      "loss": 0.981,
+      "step": 3330
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.201206140350878e-06,
+      "loss": 1.1449,
+      "step": 3331
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.200657894736843e-06,
+      "loss": 1.175,
+      "step": 3332
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.200109649122807e-06,
+      "loss": 0.719,
+      "step": 3333
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.199561403508773e-06,
+      "loss": 0.9613,
+      "step": 3334
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.199013157894737e-06,
+      "loss": 0.4765,
+      "step": 3335
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.198464912280703e-06,
+      "loss": 1.6766,
+      "step": 3336
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.197916666666668e-06,
+      "loss": 1.5156,
+      "step": 3337
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.197368421052632e-06,
+      "loss": 0.3302,
+      "step": 3338
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.196820175438596e-06,
+      "loss": 1.1599,
+      "step": 3339
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.196271929824562e-06,
+      "loss": 1.3955,
+      "step": 3340
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.195723684210528e-06,
+      "loss": 1.0324,
+      "step": 3341
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.195175438596492e-06,
+      "loss": 0.7966,
+      "step": 3342
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.194627192982457e-06,
+      "loss": 0.5096,
+      "step": 3343
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.194078947368421e-06,
+      "loss": 0.554,
+      "step": 3344
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.193530701754387e-06,
+      "loss": 1.2473,
+      "step": 3345
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.192982456140351e-06,
+      "loss": 0.4373,
+      "step": 3346
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.192434210526317e-06,
+      "loss": 0.4479,
+      "step": 3347
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.191885964912282e-06,
+      "loss": 1.1173,
+      "step": 3348
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.191337719298246e-06,
+      "loss": 1.1976,
+      "step": 3349
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.19078947368421e-06,
+      "loss": 1.3492,
+      "step": 3350
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.190241228070176e-06,
+      "loss": 0.9023,
+      "step": 3351
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.189692982456142e-06,
+      "loss": 0.5504,
+      "step": 3352
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.189144736842105e-06,
+      "loss": 0.9,
+      "step": 3353
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.188596491228071e-06,
+      "loss": 0.9233,
+      "step": 3354
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.188048245614035e-06,
+      "loss": 0.9463,
+      "step": 3355
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.1875e-06,
+      "loss": 0.4785,
+      "step": 3356
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.186951754385965e-06,
+      "loss": 1.1654,
+      "step": 3357
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.18640350877193e-06,
+      "loss": 0.7538,
+      "step": 3358
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.185855263157896e-06,
+      "loss": 1.3812,
+      "step": 3359
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.18530701754386e-06,
+      "loss": 1.1759,
+      "step": 3360
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.184758771929826e-06,
+      "loss": 0.9761,
+      "step": 3361
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.18421052631579e-06,
+      "loss": 1.2287,
+      "step": 3362
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.183662280701755e-06,
+      "loss": 1.1218,
+      "step": 3363
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.183114035087721e-06,
+      "loss": 1.2816,
+      "step": 3364
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.182565789473685e-06,
+      "loss": 0.7759,
+      "step": 3365
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.182017543859649e-06,
+      "loss": 0.7248,
+      "step": 3366
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.181469298245615e-06,
+      "loss": 0.6937,
+      "step": 3367
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.18092105263158e-06,
+      "loss": 0.6801,
+      "step": 3368
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.180372807017544e-06,
+      "loss": 1.0227,
+      "step": 3369
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.17982456140351e-06,
+      "loss": 1.4241,
+      "step": 3370
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.179276315789474e-06,
+      "loss": 0.957,
+      "step": 3371
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.17872807017544e-06,
+      "loss": 0.8596,
+      "step": 3372
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.178179824561404e-06,
+      "loss": 0.4021,
+      "step": 3373
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.17763157894737e-06,
+      "loss": 0.8512,
+      "step": 3374
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.177083333333335e-06,
+      "loss": 1.332,
+      "step": 3375
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.176535087719299e-06,
+      "loss": 0.9047,
+      "step": 3376
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.175986842105263e-06,
+      "loss": 0.6748,
+      "step": 3377
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.175438596491229e-06,
+      "loss": 0.857,
+      "step": 3378
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.174890350877194e-06,
+      "loss": 1.0931,
+      "step": 3379
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.174342105263158e-06,
+      "loss": 0.6151,
+      "step": 3380
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.173793859649124e-06,
+      "loss": 1.0436,
+      "step": 3381
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.173245614035088e-06,
+      "loss": 0.8726,
+      "step": 3382
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.172697368421054e-06,
+      "loss": 0.9575,
+      "step": 3383
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.172149122807018e-06,
+      "loss": 0.7213,
+      "step": 3384
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.171600877192983e-06,
+      "loss": 1.2215,
+      "step": 3385
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.171052631578949e-06,
+      "loss": 0.3821,
+      "step": 3386
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.170504385964913e-06,
+      "loss": 1.1619,
+      "step": 3387
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.169956140350877e-06,
+      "loss": 1.0094,
+      "step": 3388
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.169407894736843e-06,
+      "loss": 1.02,
+      "step": 3389
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.168859649122808e-06,
+      "loss": 1.2056,
+      "step": 3390
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.168311403508772e-06,
+      "loss": 1.0508,
+      "step": 3391
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.167763157894738e-06,
+      "loss": 0.5524,
+      "step": 3392
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.167214912280702e-06,
+      "loss": 1.3214,
+      "step": 3393
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.166666666666668e-06,
+      "loss": 0.9433,
+      "step": 3394
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.166118421052632e-06,
+      "loss": 0.3918,
+      "step": 3395
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.165570175438597e-06,
+      "loss": 1.1256,
+      "step": 3396
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.165021929824563e-06,
+      "loss": 1.1784,
+      "step": 3397
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.164473684210527e-06,
+      "loss": 1.2992,
+      "step": 3398
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.163925438596493e-06,
+      "loss": 0.8182,
+      "step": 3399
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.163377192982456e-06,
+      "loss": 0.6517,
+      "step": 3400
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.162828947368422e-06,
+      "loss": 0.9295,
+      "step": 3401
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.162280701754388e-06,
+      "loss": 1.0449,
+      "step": 3402
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.161732456140352e-06,
+      "loss": 1.7872,
+      "step": 3403
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.161184210526316e-06,
+      "loss": 0.8749,
+      "step": 3404
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.160635964912281e-06,
+      "loss": 1.1134,
+      "step": 3405
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.160087719298247e-06,
+      "loss": 1.1474,
+      "step": 3406
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.159539473684211e-06,
+      "loss": 1.0112,
+      "step": 3407
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.158991228070177e-06,
+      "loss": 1.0228,
+      "step": 3408
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.15844298245614e-06,
+      "loss": 0.7107,
+      "step": 3409
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.157894736842106e-06,
+      "loss": 0.9276,
+      "step": 3410
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.15734649122807e-06,
+      "loss": 0.9962,
+      "step": 3411
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.156798245614036e-06,
+      "loss": 1.3517,
+      "step": 3412
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.156250000000002e-06,
+      "loss": 0.6307,
+      "step": 3413
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.155701754385966e-06,
+      "loss": 0.5294,
+      "step": 3414
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.15515350877193e-06,
+      "loss": 0.8361,
+      "step": 3415
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.154605263157895e-06,
+      "loss": 0.8283,
+      "step": 3416
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.154057017543861e-06,
+      "loss": 0.5469,
+      "step": 3417
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.153508771929825e-06,
+      "loss": 1.171,
+      "step": 3418
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.152960526315789e-06,
+      "loss": 0.7513,
+      "step": 3419
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.152412280701755e-06,
+      "loss": 0.621,
+      "step": 3420
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.15186403508772e-06,
+      "loss": 0.9083,
+      "step": 3421
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.151315789473684e-06,
+      "loss": 2.0029,
+      "step": 3422
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.15076754385965e-06,
+      "loss": 1.0942,
+      "step": 3423
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.150219298245616e-06,
+      "loss": 0.5708,
+      "step": 3424
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.14967105263158e-06,
+      "loss": 0.5145,
+      "step": 3425
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.149122807017544e-06,
+      "loss": 0.8829,
+      "step": 3426
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.14857456140351e-06,
+      "loss": 1.3198,
+      "step": 3427
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.148026315789475e-06,
+      "loss": 1.0052,
+      "step": 3428
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.147478070175439e-06,
+      "loss": 1.0011,
+      "step": 3429
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.146929824561403e-06,
+      "loss": 1.2314,
+      "step": 3430
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.146381578947369e-06,
+      "loss": 0.5401,
+      "step": 3431
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.145833333333334e-06,
+      "loss": 0.7609,
+      "step": 3432
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.145285087719298e-06,
+      "loss": 0.8905,
+      "step": 3433
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.144736842105264e-06,
+      "loss": 1.0334,
+      "step": 3434
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.14418859649123e-06,
+      "loss": 0.4497,
+      "step": 3435
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.143640350877194e-06,
+      "loss": 0.5012,
+      "step": 3436
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.14309210526316e-06,
+      "loss": 1.3085,
+      "step": 3437
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.142543859649123e-06,
+      "loss": 0.9039,
+      "step": 3438
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.141995614035089e-06,
+      "loss": 0.9134,
+      "step": 3439
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.141447368421055e-06,
+      "loss": 1.5101,
+      "step": 3440
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.140899122807019e-06,
+      "loss": 1.2942,
+      "step": 3441
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.140350877192983e-06,
+      "loss": 0.7034,
+      "step": 3442
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.139802631578948e-06,
+      "loss": 0.679,
+      "step": 3443
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.139254385964914e-06,
+      "loss": 0.5039,
+      "step": 3444
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.138706140350878e-06,
+      "loss": 0.6257,
+      "step": 3445
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.138157894736842e-06,
+      "loss": 1.1512,
+      "step": 3446
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.137609649122807e-06,
+      "loss": 0.9799,
+      "step": 3447
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.137061403508773e-06,
+      "loss": 0.8081,
+      "step": 3448
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.136513157894737e-06,
+      "loss": 1.2687,
+      "step": 3449
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.135964912280703e-06,
+      "loss": 0.8453,
+      "step": 3450
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.135416666666668e-06,
+      "loss": 1.1888,
+      "step": 3451
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.134868421052632e-06,
+      "loss": 0.5693,
+      "step": 3452
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.134320175438596e-06,
+      "loss": 0.862,
+      "step": 3453
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.133771929824562e-06,
+      "loss": 0.5086,
+      "step": 3454
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.133223684210528e-06,
+      "loss": 0.8134,
+      "step": 3455
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.132675438596492e-06,
+      "loss": 0.8871,
+      "step": 3456
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.132127192982456e-06,
+      "loss": 0.4449,
+      "step": 3457
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.131578947368421e-06,
+      "loss": 0.4616,
+      "step": 3458
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.131030701754387e-06,
+      "loss": 0.7456,
+      "step": 3459
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.130482456140351e-06,
+      "loss": 0.5838,
+      "step": 3460
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.129934210526317e-06,
+      "loss": 0.8229,
+      "step": 3461
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.12938596491228e-06,
+      "loss": 1.1782,
+      "step": 3462
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.128837719298246e-06,
+      "loss": 0.5248,
+      "step": 3463
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.12828947368421e-06,
+      "loss": 0.8072,
+      "step": 3464
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.127741228070176e-06,
+      "loss": 1.1426,
+      "step": 3465
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.127192982456142e-06,
+      "loss": 1.6604,
+      "step": 3466
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.126644736842106e-06,
+      "loss": 0.7523,
+      "step": 3467
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.12609649122807e-06,
+      "loss": 1.54,
+      "step": 3468
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.125548245614035e-06,
+      "loss": 0.7674,
+      "step": 3469
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.125000000000001e-06,
+      "loss": 0.9958,
+      "step": 3470
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.124451754385965e-06,
+      "loss": 1.2119,
+      "step": 3471
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.12390350877193e-06,
+      "loss": 0.3853,
+      "step": 3472
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.123355263157895e-06,
+      "loss": 0.9559,
+      "step": 3473
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.12280701754386e-06,
+      "loss": 1.5693,
+      "step": 3474
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.122258771929826e-06,
+      "loss": 0.6256,
+      "step": 3475
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.12171052631579e-06,
+      "loss": 0.971,
+      "step": 3476
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.121162280701756e-06,
+      "loss": 0.7708,
+      "step": 3477
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.120614035087721e-06,
+      "loss": 0.6678,
+      "step": 3478
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.120065789473685e-06,
+      "loss": 1.0299,
+      "step": 3479
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.11951754385965e-06,
+      "loss": 1.0066,
+      "step": 3480
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.118969298245615e-06,
+      "loss": 1.3143,
+      "step": 3481
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.11842105263158e-06,
+      "loss": 0.4193,
+      "step": 3482
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.117872807017545e-06,
+      "loss": 1.2307,
+      "step": 3483
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.117324561403509e-06,
+      "loss": 0.5709,
+      "step": 3484
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.116776315789474e-06,
+      "loss": 1.6691,
+      "step": 3485
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.11622807017544e-06,
+      "loss": 1.3956,
+      "step": 3486
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.115679824561404e-06,
+      "loss": 0.8511,
+      "step": 3487
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.11513157894737e-06,
+      "loss": 0.6388,
+      "step": 3488
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.114583333333333e-06,
+      "loss": 1.2042,
+      "step": 3489
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.1140350877193e-06,
+      "loss": 0.7386,
+      "step": 3490
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.113486842105263e-06,
+      "loss": 0.8555,
+      "step": 3491
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.112938596491229e-06,
+      "loss": 1.4425,
+      "step": 3492
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.112390350877194e-06,
+      "loss": 1.3039,
+      "step": 3493
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.111842105263158e-06,
+      "loss": 1.4049,
+      "step": 3494
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.111293859649122e-06,
+      "loss": 0.7864,
+      "step": 3495
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.110745614035088e-06,
+      "loss": 1.0746,
+      "step": 3496
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.110197368421054e-06,
+      "loss": 0.6949,
+      "step": 3497
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.109649122807018e-06,
+      "loss": 1.041,
+      "step": 3498
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.109100877192983e-06,
+      "loss": 1.6082,
+      "step": 3499
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.108552631578947e-06,
+      "loss": 0.8358,
+      "step": 3500
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.108004385964913e-06,
+      "loss": 1.0717,
+      "step": 3501
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.107456140350877e-06,
+      "loss": 0.3059,
+      "step": 3502
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.106907894736843e-06,
+      "loss": 0.515,
+      "step": 3503
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.106359649122808e-06,
+      "loss": 0.9496,
+      "step": 3504
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.105811403508772e-06,
+      "loss": 0.7099,
+      "step": 3505
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.105263157894736e-06,
+      "loss": 0.9954,
+      "step": 3506
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.104714912280702e-06,
+      "loss": 1.2083,
+      "step": 3507
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.104166666666668e-06,
+      "loss": 1.2354,
+      "step": 3508
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.103618421052632e-06,
+      "loss": 1.4272,
+      "step": 3509
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.103070175438597e-06,
+      "loss": 0.7502,
+      "step": 3510
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.102521929824561e-06,
+      "loss": 0.4534,
+      "step": 3511
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.101973684210527e-06,
+      "loss": 0.5639,
+      "step": 3512
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.101425438596493e-06,
+      "loss": 0.6342,
+      "step": 3513
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.100877192982457e-06,
+      "loss": 1.5382,
+      "step": 3514
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.100328947368422e-06,
+      "loss": 1.1616,
+      "step": 3515
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.099780701754386e-06,
+      "loss": 1.5433,
+      "step": 3516
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.099232456140352e-06,
+      "loss": 0.6118,
+      "step": 3517
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.098684210526316e-06,
+      "loss": 0.6371,
+      "step": 3518
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.098135964912282e-06,
+      "loss": 1.1223,
+      "step": 3519
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.097587719298247e-06,
+      "loss": 0.5709,
+      "step": 3520
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.097039473684211e-06,
+      "loss": 0.8551,
+      "step": 3521
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.096491228070175e-06,
+      "loss": 1.4091,
+      "step": 3522
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.095942982456141e-06,
+      "loss": 0.5175,
+      "step": 3523
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.095394736842107e-06,
+      "loss": 1.0471,
+      "step": 3524
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.09484649122807e-06,
+      "loss": 1.3868,
+      "step": 3525
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.094298245614036e-06,
+      "loss": 1.1825,
+      "step": 3526
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.09375e-06,
+      "loss": 0.8525,
+      "step": 3527
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.093201754385966e-06,
+      "loss": 1.1022,
+      "step": 3528
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.09265350877193e-06,
+      "loss": 0.406,
+      "step": 3529
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.092105263157896e-06,
+      "loss": 1.6531,
+      "step": 3530
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.091557017543861e-06,
+      "loss": 1.0038,
+      "step": 3531
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.091008771929825e-06,
+      "loss": 1.0314,
+      "step": 3532
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.09046052631579e-06,
+      "loss": 1.6843,
+      "step": 3533
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.089912280701755e-06,
+      "loss": 1.1026,
+      "step": 3534
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.08936403508772e-06,
+      "loss": 0.9728,
+      "step": 3535
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.088815789473684e-06,
+      "loss": 1.0617,
+      "step": 3536
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.08826754385965e-06,
+      "loss": 0.889,
+      "step": 3537
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.087719298245614e-06,
+      "loss": 0.6841,
+      "step": 3538
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.08717105263158e-06,
+      "loss": 0.9386,
+      "step": 3539
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.086622807017544e-06,
+      "loss": 1.1706,
+      "step": 3540
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.08607456140351e-06,
+      "loss": 0.8829,
+      "step": 3541
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.085526315789475e-06,
+      "loss": 0.9665,
+      "step": 3542
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.084978070175439e-06,
+      "loss": 1.0505,
+      "step": 3543
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.084429824561403e-06,
+      "loss": 0.6594,
+      "step": 3544
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.083881578947369e-06,
+      "loss": 0.5206,
+      "step": 3545
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.083333333333334e-06,
+      "loss": 1.419,
+      "step": 3546
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.0827850877193e-06,
+      "loss": 0.4271,
+      "step": 3547
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.082236842105264e-06,
+      "loss": 0.705,
+      "step": 3548
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.081688596491228e-06,
+      "loss": 0.7423,
+      "step": 3549
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.081140350877194e-06,
+      "loss": 0.7545,
+      "step": 3550
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.08059210526316e-06,
+      "loss": 1.3032,
+      "step": 3551
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.080043859649123e-06,
+      "loss": 1.097,
+      "step": 3552
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.079495614035089e-06,
+      "loss": 0.5352,
+      "step": 3553
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.078947368421053e-06,
+      "loss": 1.0658,
+      "step": 3554
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.078399122807019e-06,
+      "loss": 1.031,
+      "step": 3555
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.077850877192983e-06,
+      "loss": 1.5021,
+      "step": 3556
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.077302631578948e-06,
+      "loss": 0.9342,
+      "step": 3557
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.076754385964914e-06,
+      "loss": 0.5498,
+      "step": 3558
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.076206140350878e-06,
+      "loss": 1.4681,
+      "step": 3559
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.075657894736842e-06,
+      "loss": 0.6633,
+      "step": 3560
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.075109649122808e-06,
+      "loss": 0.53,
+      "step": 3561
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.074561403508773e-06,
+      "loss": 0.6721,
+      "step": 3562
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.074013157894737e-06,
+      "loss": 0.7497,
+      "step": 3563
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.073464912280703e-06,
+      "loss": 0.6277,
+      "step": 3564
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.072916666666667e-06,
+      "loss": 1.704,
+      "step": 3565
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.072368421052633e-06,
+      "loss": 1.2509,
+      "step": 3566
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.071820175438597e-06,
+      "loss": 1.4193,
+      "step": 3567
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.071271929824562e-06,
+      "loss": 1.3421,
+      "step": 3568
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.070723684210528e-06,
+      "loss": 1.0543,
+      "step": 3569
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.070175438596492e-06,
+      "loss": 0.7195,
+      "step": 3570
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.069627192982456e-06,
+      "loss": 1.075,
+      "step": 3571
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.069078947368422e-06,
+      "loss": 1.0819,
+      "step": 3572
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.068530701754387e-06,
+      "loss": 1.1257,
+      "step": 3573
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.067982456140351e-06,
+      "loss": 1.386,
+      "step": 3574
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.067434210526315e-06,
+      "loss": 0.6618,
+      "step": 3575
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.066885964912281e-06,
+      "loss": 0.6757,
+      "step": 3576
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.066337719298247e-06,
+      "loss": 0.8413,
+      "step": 3577
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.06578947368421e-06,
+      "loss": 1.3732,
+      "step": 3578
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.065241228070176e-06,
+      "loss": 1.3856,
+      "step": 3579
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.064692982456142e-06,
+      "loss": 0.5272,
+      "step": 3580
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.064144736842106e-06,
+      "loss": 1.1311,
+      "step": 3581
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.06359649122807e-06,
+      "loss": 0.9031,
+      "step": 3582
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.063048245614035e-06,
+      "loss": 0.7639,
+      "step": 3583
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.062500000000001e-06,
+      "loss": 0.9261,
+      "step": 3584
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.061951754385967e-06,
+      "loss": 1.1598,
+      "step": 3585
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.06140350877193e-06,
+      "loss": 1.1546,
+      "step": 3586
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.060855263157895e-06,
+      "loss": 1.0941,
+      "step": 3587
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.06030701754386e-06,
+      "loss": 0.6962,
+      "step": 3588
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.059758771929826e-06,
+      "loss": 0.5725,
+      "step": 3589
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.05921052631579e-06,
+      "loss": 0.6902,
+      "step": 3590
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.058662280701754e-06,
+      "loss": 0.9608,
+      "step": 3591
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.05811403508772e-06,
+      "loss": 1.0506,
+      "step": 3592
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.057565789473685e-06,
+      "loss": 0.3574,
+      "step": 3593
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.05701754385965e-06,
+      "loss": 0.9285,
+      "step": 3594
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.056469298245615e-06,
+      "loss": 0.9727,
+      "step": 3595
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.05592105263158e-06,
+      "loss": 1.2263,
+      "step": 3596
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.055372807017545e-06,
+      "loss": 1.4803,
+      "step": 3597
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.054824561403509e-06,
+      "loss": 0.8747,
+      "step": 3598
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.054276315789474e-06,
+      "loss": 0.7196,
+      "step": 3599
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.05372807017544e-06,
+      "loss": 0.8497,
+      "step": 3600
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.053179824561404e-06,
+      "loss": 1.1532,
+      "step": 3601
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.052631578947368e-06,
+      "loss": 1.0654,
+      "step": 3602
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.052083333333334e-06,
+      "loss": 1.0718,
+      "step": 3603
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.0515350877193e-06,
+      "loss": 1.238,
+      "step": 3604
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.050986842105263e-06,
+      "loss": 1.0254,
+      "step": 3605
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.050438596491229e-06,
+      "loss": 0.8198,
+      "step": 3606
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.049890350877195e-06,
+      "loss": 1.372,
+      "step": 3607
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.049342105263159e-06,
+      "loss": 1.5665,
+      "step": 3608
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.048793859649123e-06,
+      "loss": 0.5629,
+      "step": 3609
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.048245614035088e-06,
+      "loss": 0.9504,
+      "step": 3610
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.047697368421054e-06,
+      "loss": 1.4518,
+      "step": 3611
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.047149122807018e-06,
+      "loss": 0.9431,
+      "step": 3612
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.046600877192982e-06,
+      "loss": 1.0517,
+      "step": 3613
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.046052631578948e-06,
+      "loss": 1.0312,
+      "step": 3614
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.045504385964913e-06,
+      "loss": 1.1901,
+      "step": 3615
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.044956140350877e-06,
+      "loss": 1.9864,
+      "step": 3616
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.044407894736843e-06,
+      "loss": 0.4616,
+      "step": 3617
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.043859649122807e-06,
+      "loss": 0.7872,
+      "step": 3618
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.043311403508773e-06,
+      "loss": 1.3928,
+      "step": 3619
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.042763157894737e-06,
+      "loss": 1.0419,
+      "step": 3620
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.042214912280702e-06,
+      "loss": 1.4878,
+      "step": 3621
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.041666666666668e-06,
+      "loss": 0.8277,
+      "step": 3622
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.041118421052634e-06,
+      "loss": 0.5856,
+      "step": 3623
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.040570175438598e-06,
+      "loss": 1.2822,
+      "step": 3624
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.040021929824562e-06,
+      "loss": 0.7812,
+      "step": 3625
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.039473684210527e-06,
+      "loss": 0.8683,
+      "step": 3626
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.038925438596493e-06,
+      "loss": 0.8529,
+      "step": 3627
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.038377192982457e-06,
+      "loss": 0.6817,
+      "step": 3628
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.03782894736842e-06,
+      "loss": 0.7974,
+      "step": 3629
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.037280701754386e-06,
+      "loss": 1.1423,
+      "step": 3630
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.036732456140352e-06,
+      "loss": 0.4858,
+      "step": 3631
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.036184210526316e-06,
+      "loss": 1.338,
+      "step": 3632
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.035635964912282e-06,
+      "loss": 1.1074,
+      "step": 3633
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.035087719298247e-06,
+      "loss": 0.7151,
+      "step": 3634
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.034539473684211e-06,
+      "loss": 1.2527,
+      "step": 3635
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.033991228070175e-06,
+      "loss": 1.4514,
+      "step": 3636
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.033442982456141e-06,
+      "loss": 0.9059,
+      "step": 3637
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.032894736842107e-06,
+      "loss": 1.1834,
+      "step": 3638
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.03234649122807e-06,
+      "loss": 0.4661,
+      "step": 3639
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.031798245614035e-06,
+      "loss": 0.7354,
+      "step": 3640
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.03125e-06,
+      "loss": 1.3866,
+      "step": 3641
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.030701754385966e-06,
+      "loss": 1.107,
+      "step": 3642
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.03015350877193e-06,
+      "loss": 0.9915,
+      "step": 3643
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.029605263157896e-06,
+      "loss": 1.2292,
+      "step": 3644
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.02905701754386e-06,
+      "loss": 0.4632,
+      "step": 3645
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.028508771929825e-06,
+      "loss": 1.2464,
+      "step": 3646
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.02796052631579e-06,
+      "loss": 0.899,
+      "step": 3647
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.027412280701755e-06,
+      "loss": 0.814,
+      "step": 3648
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.02686403508772e-06,
+      "loss": 0.403,
+      "step": 3649
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.026315789473685e-06,
+      "loss": 0.7297,
+      "step": 3650
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.025767543859649e-06,
+      "loss": 0.7868,
+      "step": 3651
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.025219298245614e-06,
+      "loss": 0.6208,
+      "step": 3652
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.02467105263158e-06,
+      "loss": 0.8342,
+      "step": 3653
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.024122807017544e-06,
+      "loss": 2.0636,
+      "step": 3654
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.02357456140351e-06,
+      "loss": 0.4321,
+      "step": 3655
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.023026315789474e-06,
+      "loss": 0.9277,
+      "step": 3656
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.02247807017544e-06,
+      "loss": 0.7314,
+      "step": 3657
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.021929824561405e-06,
+      "loss": 0.647,
+      "step": 3658
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.6157473856879229,
+      "eval_loss": 0.9592559933662415,
+      "eval_roc_auc": 0.8826880368704012,
+      "eval_runtime": 181.5148,
+      "eval_samples_per_second": 26.868,
+      "eval_steps_per_second": 3.361,
+      "step": 3658
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.021381578947369e-06,
+      "loss": 0.7844,
+      "step": 3659
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.020833333333335e-06,
+      "loss": 0.7517,
+      "step": 3660
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.020285087719299e-06,
+      "loss": 1.4834,
+      "step": 3661
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.019736842105264e-06,
+      "loss": 1.4818,
+      "step": 3662
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.019188596491228e-06,
+      "loss": 0.4292,
+      "step": 3663
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.018640350877194e-06,
+      "loss": 1.0389,
+      "step": 3664
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.01809210526316e-06,
+      "loss": 1.3471,
+      "step": 3665
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.017543859649124e-06,
+      "loss": 0.5101,
+      "step": 3666
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.016995614035088e-06,
+      "loss": 0.4284,
+      "step": 3667
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 8.016447368421053e-06,
+      "loss": 1.4567,
+      "step": 3668
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 8.015899122807019e-06,
+      "loss": 0.6501,
+      "step": 3669
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 8.015350877192983e-06,
+      "loss": 0.7802,
+      "step": 3670
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 8.014802631578949e-06,
+      "loss": 0.7623,
+      "step": 3671
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 8.014254385964913e-06,
+      "loss": 0.8132,
+      "step": 3672
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 8.013706140350878e-06,
+      "loss": 0.6935,
+      "step": 3673
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 8.013157894736842e-06,
+      "loss": 1.0384,
+      "step": 3674
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 8.012609649122808e-06,
+      "loss": 0.6387,
+      "step": 3675
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 8.012061403508773e-06,
+      "loss": 0.9756,
+      "step": 3676
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 8.011513157894737e-06,
+      "loss": 0.9748,
+      "step": 3677
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 8.010964912280701e-06,
+      "loss": 1.133,
+      "step": 3678
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 8.010416666666667e-06,
+      "loss": 0.6335,
+      "step": 3679
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 8.009868421052633e-06,
+      "loss": 0.538,
+      "step": 3680
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 8.009320175438597e-06,
+      "loss": 0.8496,
+      "step": 3681
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 8.008771929824562e-06,
+      "loss": 0.8572,
+      "step": 3682
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 8.008223684210526e-06,
+      "loss": 0.6775,
+      "step": 3683
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 8.007675438596492e-06,
+      "loss": 0.4506,
+      "step": 3684
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 8.007127192982456e-06,
+      "loss": 1.0255,
+      "step": 3685
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 8.006578947368422e-06,
+      "loss": 0.6999,
+      "step": 3686
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 8.006030701754387e-06,
+      "loss": 0.5851,
+      "step": 3687
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 8.005482456140351e-06,
+      "loss": 0.4459,
+      "step": 3688
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 8.004934210526315e-06,
+      "loss": 0.6701,
+      "step": 3689
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 8.004385964912281e-06,
+      "loss": 0.8659,
+      "step": 3690
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 8.003837719298247e-06,
+      "loss": 0.7912,
+      "step": 3691
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 8.00328947368421e-06,
+      "loss": 0.5062,
+      "step": 3692
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 8.002741228070176e-06,
+      "loss": 0.785,
+      "step": 3693
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 8.00219298245614e-06,
+      "loss": 1.4397,
+      "step": 3694
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 8.001644736842106e-06,
+      "loss": 1.1385,
+      "step": 3695
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 8.001096491228072e-06,
+      "loss": 0.6979,
+      "step": 3696
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 8.000548245614036e-06,
+      "loss": 1.2964,
+      "step": 3697
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.6157,
+      "step": 3698
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 7.999451754385965e-06,
+      "loss": 1.4576,
+      "step": 3699
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 7.998903508771931e-06,
+      "loss": 0.8537,
+      "step": 3700
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 7.998355263157895e-06,
+      "loss": 0.6739,
+      "step": 3701
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 7.99780701754386e-06,
+      "loss": 0.7594,
+      "step": 3702
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 7.997258771929826e-06,
+      "loss": 1.0592,
+      "step": 3703
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.99671052631579e-06,
+      "loss": 1.1258,
+      "step": 3704
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.996162280701754e-06,
+      "loss": 0.7781,
+      "step": 3705
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.99561403508772e-06,
+      "loss": 0.7996,
+      "step": 3706
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.995065789473686e-06,
+      "loss": 0.6665,
+      "step": 3707
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.99451754385965e-06,
+      "loss": 1.0764,
+      "step": 3708
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.993969298245615e-06,
+      "loss": 0.7172,
+      "step": 3709
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.99342105263158e-06,
+      "loss": 1.0116,
+      "step": 3710
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.992872807017545e-06,
+      "loss": 0.4264,
+      "step": 3711
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.992324561403509e-06,
+      "loss": 0.5463,
+      "step": 3712
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.991776315789475e-06,
+      "loss": 0.7244,
+      "step": 3713
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.99122807017544e-06,
+      "loss": 1.0897,
+      "step": 3714
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.990679824561404e-06,
+      "loss": 0.8261,
+      "step": 3715
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.990131578947368e-06,
+      "loss": 0.8446,
+      "step": 3716
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.989583333333334e-06,
+      "loss": 0.8149,
+      "step": 3717
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.9890350877193e-06,
+      "loss": 1.3195,
+      "step": 3718
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.988486842105263e-06,
+      "loss": 0.7773,
+      "step": 3719
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.98793859649123e-06,
+      "loss": 0.9569,
+      "step": 3720
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.987390350877193e-06,
+      "loss": 1.053,
+      "step": 3721
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.986842105263159e-06,
+      "loss": 0.7193,
+      "step": 3722
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.986293859649123e-06,
+      "loss": 1.1413,
+      "step": 3723
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.985745614035088e-06,
+      "loss": 1.5122,
+      "step": 3724
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.985197368421054e-06,
+      "loss": 0.8016,
+      "step": 3725
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.984649122807018e-06,
+      "loss": 1.2719,
+      "step": 3726
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.984100877192982e-06,
+      "loss": 0.9337,
+      "step": 3727
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.983552631578948e-06,
+      "loss": 1.0847,
+      "step": 3728
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.983004385964913e-06,
+      "loss": 1.0194,
+      "step": 3729
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.982456140350877e-06,
+      "loss": 1.0736,
+      "step": 3730
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.981907894736843e-06,
+      "loss": 1.5724,
+      "step": 3731
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.981359649122807e-06,
+      "loss": 1.0023,
+      "step": 3732
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.980811403508773e-06,
+      "loss": 1.3455,
+      "step": 3733
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.980263157894738e-06,
+      "loss": 0.6758,
+      "step": 3734
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.979714912280702e-06,
+      "loss": 1.1107,
+      "step": 3735
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.979166666666668e-06,
+      "loss": 0.8815,
+      "step": 3736
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.978618421052632e-06,
+      "loss": 0.8585,
+      "step": 3737
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.978070175438598e-06,
+      "loss": 0.7532,
+      "step": 3738
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.977521929824562e-06,
+      "loss": 1.1574,
+      "step": 3739
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.976973684210527e-06,
+      "loss": 0.694,
+      "step": 3740
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.976425438596493e-06,
+      "loss": 0.9434,
+      "step": 3741
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.975877192982457e-06,
+      "loss": 1.0685,
+      "step": 3742
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.975328947368421e-06,
+      "loss": 0.7496,
+      "step": 3743
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.974780701754387e-06,
+      "loss": 0.9055,
+      "step": 3744
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.974232456140352e-06,
+      "loss": 0.9739,
+      "step": 3745
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.973684210526316e-06,
+      "loss": 0.9681,
+      "step": 3746
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.97313596491228e-06,
+      "loss": 1.1913,
+      "step": 3747
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.972587719298246e-06,
+      "loss": 0.6693,
+      "step": 3748
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.972039473684212e-06,
+      "loss": 0.5799,
+      "step": 3749
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.971491228070176e-06,
+      "loss": 1.1371,
+      "step": 3750
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.970942982456141e-06,
+      "loss": 1.2825,
+      "step": 3751
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.970394736842107e-06,
+      "loss": 1.2603,
+      "step": 3752
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.969846491228071e-06,
+      "loss": 1.3992,
+      "step": 3753
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.969298245614035e-06,
+      "loss": 0.6381,
+      "step": 3754
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.96875e-06,
+      "loss": 0.8167,
+      "step": 3755
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.968201754385966e-06,
+      "loss": 0.8673,
+      "step": 3756
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.96765350877193e-06,
+      "loss": 0.7904,
+      "step": 3757
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.967105263157894e-06,
+      "loss": 1.0697,
+      "step": 3758
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.96655701754386e-06,
+      "loss": 1.024,
+      "step": 3759
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.966008771929826e-06,
+      "loss": 1.1071,
+      "step": 3760
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.96546052631579e-06,
+      "loss": 1.2121,
+      "step": 3761
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.964912280701755e-06,
+      "loss": 0.9444,
+      "step": 3762
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.964364035087721e-06,
+      "loss": 0.8583,
+      "step": 3763
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.963815789473685e-06,
+      "loss": 0.859,
+      "step": 3764
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.963267543859649e-06,
+      "loss": 1.0551,
+      "step": 3765
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.962719298245614e-06,
+      "loss": 0.4741,
+      "step": 3766
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.96217105263158e-06,
+      "loss": 1.501,
+      "step": 3767
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.961622807017544e-06,
+      "loss": 1.173,
+      "step": 3768
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.96107456140351e-06,
+      "loss": 0.6793,
+      "step": 3769
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.960526315789474e-06,
+      "loss": 1.5674,
+      "step": 3770
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.95997807017544e-06,
+      "loss": 0.7153,
+      "step": 3771
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.959429824561405e-06,
+      "loss": 0.9326,
+      "step": 3772
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.958881578947369e-06,
+      "loss": 0.789,
+      "step": 3773
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.958333333333333e-06,
+      "loss": 1.0853,
+      "step": 3774
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.957785087719299e-06,
+      "loss": 0.8774,
+      "step": 3775
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.957236842105264e-06,
+      "loss": 0.6941,
+      "step": 3776
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.956688596491228e-06,
+      "loss": 2.0036,
+      "step": 3777
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.956140350877194e-06,
+      "loss": 0.8147,
+      "step": 3778
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.95559210526316e-06,
+      "loss": 1.127,
+      "step": 3779
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.955043859649124e-06,
+      "loss": 0.8984,
+      "step": 3780
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.954495614035088e-06,
+      "loss": 0.6551,
+      "step": 3781
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.953947368421053e-06,
+      "loss": 1.1802,
+      "step": 3782
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.953399122807019e-06,
+      "loss": 1.1445,
+      "step": 3783
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.952850877192983e-06,
+      "loss": 1.1696,
+      "step": 3784
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.952302631578947e-06,
+      "loss": 1.0705,
+      "step": 3785
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.951754385964913e-06,
+      "loss": 1.0591,
+      "step": 3786
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.951206140350878e-06,
+      "loss": 0.7052,
+      "step": 3787
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.950657894736842e-06,
+      "loss": 1.312,
+      "step": 3788
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.950109649122808e-06,
+      "loss": 0.7846,
+      "step": 3789
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.949561403508772e-06,
+      "loss": 0.7502,
+      "step": 3790
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.949013157894738e-06,
+      "loss": 0.4194,
+      "step": 3791
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.948464912280702e-06,
+      "loss": 1.1687,
+      "step": 3792
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.947916666666667e-06,
+      "loss": 1.1413,
+      "step": 3793
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.947368421052633e-06,
+      "loss": 0.6967,
+      "step": 3794
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.946820175438597e-06,
+      "loss": 0.8647,
+      "step": 3795
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.946271929824561e-06,
+      "loss": 0.6163,
+      "step": 3796
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.945723684210527e-06,
+      "loss": 0.8621,
+      "step": 3797
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.945175438596492e-06,
+      "loss": 0.547,
+      "step": 3798
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.944627192982456e-06,
+      "loss": 0.5943,
+      "step": 3799
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.944078947368422e-06,
+      "loss": 0.7192,
+      "step": 3800
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.943530701754386e-06,
+      "loss": 0.8316,
+      "step": 3801
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.942982456140352e-06,
+      "loss": 0.6291,
+      "step": 3802
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.942434210526316e-06,
+      "loss": 1.2224,
+      "step": 3803
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.941885964912281e-06,
+      "loss": 1.2537,
+      "step": 3804
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.941337719298247e-06,
+      "loss": 1.3699,
+      "step": 3805
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.940789473684213e-06,
+      "loss": 0.814,
+      "step": 3806
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.940241228070177e-06,
+      "loss": 1.3933,
+      "step": 3807
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.93969298245614e-06,
+      "loss": 0.593,
+      "step": 3808
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.939144736842106e-06,
+      "loss": 0.7959,
+      "step": 3809
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.938596491228072e-06,
+      "loss": 1.2251,
+      "step": 3810
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.938048245614036e-06,
+      "loss": 0.5351,
+      "step": 3811
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.9375e-06,
+      "loss": 1.3225,
+      "step": 3812
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.936951754385965e-06,
+      "loss": 0.8192,
+      "step": 3813
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.936403508771931e-06,
+      "loss": 0.8053,
+      "step": 3814
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.935855263157895e-06,
+      "loss": 0.5981,
+      "step": 3815
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.93530701754386e-06,
+      "loss": 1.0753,
+      "step": 3816
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.934758771929825e-06,
+      "loss": 0.8577,
+      "step": 3817
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.93421052631579e-06,
+      "loss": 0.9782,
+      "step": 3818
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.933662280701754e-06,
+      "loss": 0.9824,
+      "step": 3819
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.93311403508772e-06,
+      "loss": 0.7631,
+      "step": 3820
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.932565789473686e-06,
+      "loss": 0.3845,
+      "step": 3821
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.93201754385965e-06,
+      "loss": 1.2596,
+      "step": 3822
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.931469298245614e-06,
+      "loss": 1.1534,
+      "step": 3823
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.93092105263158e-06,
+      "loss": 0.6252,
+      "step": 3824
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.930372807017545e-06,
+      "loss": 1.4537,
+      "step": 3825
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.929824561403509e-06,
+      "loss": 0.5694,
+      "step": 3826
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.929276315789475e-06,
+      "loss": 1.587,
+      "step": 3827
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.928728070175439e-06,
+      "loss": 0.5841,
+      "step": 3828
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.928179824561404e-06,
+      "loss": 0.6721,
+      "step": 3829
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.927631578947368e-06,
+      "loss": 1.052,
+      "step": 3830
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.927083333333334e-06,
+      "loss": 0.7252,
+      "step": 3831
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.9265350877193e-06,
+      "loss": 1.695,
+      "step": 3832
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.925986842105264e-06,
+      "loss": 1.5707,
+      "step": 3833
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.925438596491228e-06,
+      "loss": 1.1624,
+      "step": 3834
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.924890350877193e-06,
+      "loss": 0.8282,
+      "step": 3835
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.924342105263159e-06,
+      "loss": 0.5093,
+      "step": 3836
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.923793859649123e-06,
+      "loss": 0.4345,
+      "step": 3837
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.923245614035089e-06,
+      "loss": 0.6027,
+      "step": 3838
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.922697368421053e-06,
+      "loss": 0.6497,
+      "step": 3839
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.922149122807018e-06,
+      "loss": 0.9154,
+      "step": 3840
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.921600877192982e-06,
+      "loss": 0.4464,
+      "step": 3841
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.921052631578948e-06,
+      "loss": 0.7263,
+      "step": 3842
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.920504385964914e-06,
+      "loss": 0.7314,
+      "step": 3843
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.919956140350878e-06,
+      "loss": 0.3525,
+      "step": 3844
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.919407894736843e-06,
+      "loss": 1.1877,
+      "step": 3845
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.918859649122807e-06,
+      "loss": 0.9466,
+      "step": 3846
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.918311403508773e-06,
+      "loss": 0.9143,
+      "step": 3847
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.917763157894739e-06,
+      "loss": 0.6842,
+      "step": 3848
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.917214912280703e-06,
+      "loss": 0.5318,
+      "step": 3849
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.916666666666667e-06,
+      "loss": 1.1359,
+      "step": 3850
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.916118421052632e-06,
+      "loss": 1.0402,
+      "step": 3851
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.915570175438598e-06,
+      "loss": 0.6187,
+      "step": 3852
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.915021929824562e-06,
+      "loss": 1.4803,
+      "step": 3853
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.914473684210528e-06,
+      "loss": 0.7683,
+      "step": 3854
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.913925438596492e-06,
+      "loss": 1.2289,
+      "step": 3855
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.913377192982457e-06,
+      "loss": 0.6811,
+      "step": 3856
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.912828947368421e-06,
+      "loss": 1.3572,
+      "step": 3857
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.912280701754387e-06,
+      "loss": 0.6546,
+      "step": 3858
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.911732456140353e-06,
+      "loss": 0.5256,
+      "step": 3859
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.911184210526316e-06,
+      "loss": 0.9225,
+      "step": 3860
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.91063596491228e-06,
+      "loss": 1.0074,
+      "step": 3861
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.910087719298246e-06,
+      "loss": 0.5352,
+      "step": 3862
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.909539473684212e-06,
+      "loss": 0.9012,
+      "step": 3863
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.908991228070176e-06,
+      "loss": 0.7382,
+      "step": 3864
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.908442982456141e-06,
+      "loss": 1.3048,
+      "step": 3865
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.907894736842105e-06,
+      "loss": 0.7521,
+      "step": 3866
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.907346491228071e-06,
+      "loss": 0.3772,
+      "step": 3867
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.906798245614035e-06,
+      "loss": 0.8417,
+      "step": 3868
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.90625e-06,
+      "loss": 1.3392,
+      "step": 3869
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.905701754385966e-06,
+      "loss": 0.8754,
+      "step": 3870
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.90515350877193e-06,
+      "loss": 0.8154,
+      "step": 3871
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.904605263157894e-06,
+      "loss": 1.1374,
+      "step": 3872
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.90405701754386e-06,
+      "loss": 1.0138,
+      "step": 3873
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.903508771929826e-06,
+      "loss": 1.2466,
+      "step": 3874
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.90296052631579e-06,
+      "loss": 0.4166,
+      "step": 3875
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.902412280701754e-06,
+      "loss": 0.3442,
+      "step": 3876
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.90186403508772e-06,
+      "loss": 0.3271,
+      "step": 3877
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.901315789473685e-06,
+      "loss": 0.7347,
+      "step": 3878
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.900767543859649e-06,
+      "loss": 1.0277,
+      "step": 3879
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.900219298245615e-06,
+      "loss": 0.6114,
+      "step": 3880
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.89967105263158e-06,
+      "loss": 0.7129,
+      "step": 3881
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.899122807017544e-06,
+      "loss": 1.1142,
+      "step": 3882
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.89857456140351e-06,
+      "loss": 0.5466,
+      "step": 3883
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.898026315789474e-06,
+      "loss": 0.8415,
+      "step": 3884
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.89747807017544e-06,
+      "loss": 1.1405,
+      "step": 3885
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.896929824561405e-06,
+      "loss": 0.7255,
+      "step": 3886
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.89638157894737e-06,
+      "loss": 0.6648,
+      "step": 3887
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.895833333333333e-06,
+      "loss": 1.146,
+      "step": 3888
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.895285087719299e-06,
+      "loss": 0.7404,
+      "step": 3889
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.894736842105265e-06,
+      "loss": 0.7592,
+      "step": 3890
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.894188596491229e-06,
+      "loss": 1.0972,
+      "step": 3891
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.893640350877194e-06,
+      "loss": 1.2079,
+      "step": 3892
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.893092105263158e-06,
+      "loss": 1.0332,
+      "step": 3893
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.892543859649124e-06,
+      "loss": 1.2201,
+      "step": 3894
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.891995614035088e-06,
+      "loss": 0.5615,
+      "step": 3895
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.891447368421054e-06,
+      "loss": 0.9985,
+      "step": 3896
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.89089912280702e-06,
+      "loss": 1.5279,
+      "step": 3897
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.890350877192983e-06,
+      "loss": 0.7858,
+      "step": 3898
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.889802631578947e-06,
+      "loss": 0.7201,
+      "step": 3899
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.889254385964913e-06,
+      "loss": 0.7951,
+      "step": 3900
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.888706140350879e-06,
+      "loss": 0.4769,
+      "step": 3901
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.888157894736843e-06,
+      "loss": 1.064,
+      "step": 3902
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.887609649122806e-06,
+      "loss": 0.8824,
+      "step": 3903
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.887061403508772e-06,
+      "loss": 0.9078,
+      "step": 3904
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.886513157894738e-06,
+      "loss": 1.4867,
+      "step": 3905
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.885964912280702e-06,
+      "loss": 0.9724,
+      "step": 3906
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.885416666666667e-06,
+      "loss": 1.4207,
+      "step": 3907
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.884868421052633e-06,
+      "loss": 1.1661,
+      "step": 3908
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.884320175438597e-06,
+      "loss": 1.3944,
+      "step": 3909
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.883771929824561e-06,
+      "loss": 0.8186,
+      "step": 3910
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.883223684210527e-06,
+      "loss": 1.3056,
+      "step": 3911
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.882675438596492e-06,
+      "loss": 0.5387,
+      "step": 3912
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.882127192982456e-06,
+      "loss": 0.2731,
+      "step": 3913
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.88157894736842e-06,
+      "loss": 1.3269,
+      "step": 3914
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.881030701754386e-06,
+      "loss": 0.5557,
+      "step": 3915
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.880482456140352e-06,
+      "loss": 0.54,
+      "step": 3916
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.879934210526317e-06,
+      "loss": 1.5119,
+      "step": 3917
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.879385964912281e-06,
+      "loss": 0.8163,
+      "step": 3918
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.878837719298247e-06,
+      "loss": 0.4018,
+      "step": 3919
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.878289473684211e-06,
+      "loss": 0.4545,
+      "step": 3920
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.877741228070177e-06,
+      "loss": 1.3607,
+      "step": 3921
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.87719298245614e-06,
+      "loss": 1.2921,
+      "step": 3922
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.876644736842106e-06,
+      "loss": 1.4247,
+      "step": 3923
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.876096491228072e-06,
+      "loss": 1.0454,
+      "step": 3924
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.875548245614036e-06,
+      "loss": 1.0772,
+      "step": 3925
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.875e-06,
+      "loss": 0.8605,
+      "step": 3926
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.874451754385966e-06,
+      "loss": 1.7378,
+      "step": 3927
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.873903508771931e-06,
+      "loss": 0.8102,
+      "step": 3928
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.873355263157895e-06,
+      "loss": 0.9331,
+      "step": 3929
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.87280701754386e-06,
+      "loss": 1.0229,
+      "step": 3930
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.872258771929825e-06,
+      "loss": 0.6937,
+      "step": 3931
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.87171052631579e-06,
+      "loss": 1.0528,
+      "step": 3932
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.871162280701755e-06,
+      "loss": 1.2333,
+      "step": 3933
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.87061403508772e-06,
+      "loss": 0.5935,
+      "step": 3934
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.870065789473686e-06,
+      "loss": 0.7633,
+      "step": 3935
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.86951754385965e-06,
+      "loss": 1.1718,
+      "step": 3936
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.868969298245614e-06,
+      "loss": 0.6793,
+      "step": 3937
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.86842105263158e-06,
+      "loss": 0.9077,
+      "step": 3938
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.867872807017545e-06,
+      "loss": 1.1124,
+      "step": 3939
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.86732456140351e-06,
+      "loss": 0.7627,
+      "step": 3940
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.866776315789473e-06,
+      "loss": 1.0598,
+      "step": 3941
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.866228070175439e-06,
+      "loss": 1.0669,
+      "step": 3942
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.865679824561405e-06,
+      "loss": 1.2496,
+      "step": 3943
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.865131578947369e-06,
+      "loss": 0.7456,
+      "step": 3944
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.864583333333334e-06,
+      "loss": 0.8271,
+      "step": 3945
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.864035087719298e-06,
+      "loss": 1.1574,
+      "step": 3946
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.863486842105264e-06,
+      "loss": 0.428,
+      "step": 3947
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.862938596491228e-06,
+      "loss": 0.8395,
+      "step": 3948
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.862390350877193e-06,
+      "loss": 0.7908,
+      "step": 3949
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.86184210526316e-06,
+      "loss": 0.9425,
+      "step": 3950
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.861293859649123e-06,
+      "loss": 0.867,
+      "step": 3951
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.860745614035087e-06,
+      "loss": 0.3342,
+      "step": 3952
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.860197368421053e-06,
+      "loss": 0.8334,
+      "step": 3953
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.859649122807018e-06,
+      "loss": 0.6222,
+      "step": 3954
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.859100877192984e-06,
+      "loss": 0.7991,
+      "step": 3955
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.858552631578948e-06,
+      "loss": 1.318,
+      "step": 3956
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.858004385964912e-06,
+      "loss": 0.7088,
+      "step": 3957
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.857456140350878e-06,
+      "loss": 0.8308,
+      "step": 3958
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.856907894736843e-06,
+      "loss": 0.845,
+      "step": 3959
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.856359649122807e-06,
+      "loss": 0.9677,
+      "step": 3960
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.855811403508773e-06,
+      "loss": 0.6581,
+      "step": 3961
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.855263157894739e-06,
+      "loss": 0.8261,
+      "step": 3962
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.854714912280703e-06,
+      "loss": 0.8976,
+      "step": 3963
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.854166666666667e-06,
+      "loss": 0.5729,
+      "step": 3964
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.853618421052632e-06,
+      "loss": 0.5421,
+      "step": 3965
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.853070175438598e-06,
+      "loss": 0.7744,
+      "step": 3966
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.852521929824562e-06,
+      "loss": 0.7008,
+      "step": 3967
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.851973684210526e-06,
+      "loss": 1.1247,
+      "step": 3968
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.851425438596492e-06,
+      "loss": 0.843,
+      "step": 3969
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.850877192982457e-06,
+      "loss": 0.9871,
+      "step": 3970
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.850328947368421e-06,
+      "loss": 0.5739,
+      "step": 3971
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.849780701754387e-06,
+      "loss": 0.5974,
+      "step": 3972
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.849232456140351e-06,
+      "loss": 0.6364,
+      "step": 3973
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.848684210526317e-06,
+      "loss": 0.617,
+      "step": 3974
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.84813596491228e-06,
+      "loss": 0.9331,
+      "step": 3975
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.847587719298246e-06,
+      "loss": 0.7408,
+      "step": 3976
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.847039473684212e-06,
+      "loss": 1.4404,
+      "step": 3977
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.846491228070176e-06,
+      "loss": 0.7951,
+      "step": 3978
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.84594298245614e-06,
+      "loss": 0.8933,
+      "step": 3979
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.845394736842106e-06,
+      "loss": 1.1818,
+      "step": 3980
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.844846491228071e-06,
+      "loss": 0.7289,
+      "step": 3981
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.844298245614035e-06,
+      "loss": 1.8762,
+      "step": 3982
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.843750000000001e-06,
+      "loss": 0.903,
+      "step": 3983
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.843201754385965e-06,
+      "loss": 1.1413,
+      "step": 3984
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.84265350877193e-06,
+      "loss": 0.9099,
+      "step": 3985
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.842105263157895e-06,
+      "loss": 1.1309,
+      "step": 3986
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.84155701754386e-06,
+      "loss": 1.918,
+      "step": 3987
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.841008771929826e-06,
+      "loss": 1.3735,
+      "step": 3988
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.84046052631579e-06,
+      "loss": 0.5511,
+      "step": 3989
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.839912280701756e-06,
+      "loss": 0.4107,
+      "step": 3990
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.83936403508772e-06,
+      "loss": 0.8086,
+      "step": 3991
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.838815789473685e-06,
+      "loss": 0.9504,
+      "step": 3992
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.838267543859651e-06,
+      "loss": 1.498,
+      "step": 3993
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.837719298245615e-06,
+      "loss": 1.1097,
+      "step": 3994
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.837171052631579e-06,
+      "loss": 0.7532,
+      "step": 3995
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.836622807017544e-06,
+      "loss": 0.9959,
+      "step": 3996
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.83607456140351e-06,
+      "loss": 0.3866,
+      "step": 3997
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.835526315789474e-06,
+      "loss": 0.7778,
+      "step": 3998
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.83497807017544e-06,
+      "loss": 1.0402,
+      "step": 3999
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.834429824561404e-06,
+      "loss": 0.6462,
+      "step": 4000
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.83388157894737e-06,
+      "loss": 0.7774,
+      "step": 4001
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.833333333333333e-06,
+      "loss": 1.0486,
+      "step": 4002
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.832785087719299e-06,
+      "loss": 0.5071,
+      "step": 4003
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.832236842105265e-06,
+      "loss": 0.4191,
+      "step": 4004
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.831688596491229e-06,
+      "loss": 0.9229,
+      "step": 4005
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.831140350877193e-06,
+      "loss": 1.0612,
+      "step": 4006
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.830592105263158e-06,
+      "loss": 0.8405,
+      "step": 4007
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.830043859649124e-06,
+      "loss": 0.3221,
+      "step": 4008
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.829495614035088e-06,
+      "loss": 0.7904,
+      "step": 4009
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.828947368421054e-06,
+      "loss": 0.6868,
+      "step": 4010
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.828399122807018e-06,
+      "loss": 1.2468,
+      "step": 4011
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.827850877192983e-06,
+      "loss": 0.5809,
+      "step": 4012
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.827302631578947e-06,
+      "loss": 0.9027,
+      "step": 4013
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.826754385964913e-06,
+      "loss": 1.1574,
+      "step": 4014
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.826206140350879e-06,
+      "loss": 0.5907,
+      "step": 4015
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.825657894736843e-06,
+      "loss": 0.757,
+      "step": 4016
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.825109649122807e-06,
+      "loss": 0.7919,
+      "step": 4017
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.824561403508772e-06,
+      "loss": 0.5689,
+      "step": 4018
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.824013157894738e-06,
+      "loss": 0.4766,
+      "step": 4019
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.823464912280702e-06,
+      "loss": 0.8197,
+      "step": 4020
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.822916666666668e-06,
+      "loss": 0.5292,
+      "step": 4021
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.822368421052632e-06,
+      "loss": 0.6688,
+      "step": 4022
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.821820175438597e-06,
+      "loss": 1.1955,
+      "step": 4023
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.821271929824561e-06,
+      "loss": 1.1308,
+      "step": 4024
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.820723684210527e-06,
+      "loss": 0.7294,
+      "step": 4025
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.820175438596493e-06,
+      "loss": 1.0126,
+      "step": 4026
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.819627192982457e-06,
+      "loss": 1.6621,
+      "step": 4027
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.819078947368422e-06,
+      "loss": 0.9536,
+      "step": 4028
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.818530701754386e-06,
+      "loss": 0.8677,
+      "step": 4029
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.817982456140352e-06,
+      "loss": 0.664,
+      "step": 4030
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.817434210526318e-06,
+      "loss": 1.0282,
+      "step": 4031
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.816885964912282e-06,
+      "loss": 1.357,
+      "step": 4032
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.816337719298246e-06,
+      "loss": 0.6185,
+      "step": 4033
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.815789473684211e-06,
+      "loss": 1.4455,
+      "step": 4034
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.815241228070177e-06,
+      "loss": 1.1373,
+      "step": 4035
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.814692982456141e-06,
+      "loss": 0.794,
+      "step": 4036
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.814144736842107e-06,
+      "loss": 0.6868,
+      "step": 4037
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.81359649122807e-06,
+      "loss": 1.3946,
+      "step": 4038
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.813048245614036e-06,
+      "loss": 1.2214,
+      "step": 4039
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.8125e-06,
+      "loss": 1.2897,
+      "step": 4040
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.811951754385966e-06,
+      "loss": 0.5294,
+      "step": 4041
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.811403508771932e-06,
+      "loss": 1.2147,
+      "step": 4042
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.810855263157895e-06,
+      "loss": 0.5751,
+      "step": 4043
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.81030701754386e-06,
+      "loss": 1.2952,
+      "step": 4044
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.809758771929825e-06,
+      "loss": 0.985,
+      "step": 4045
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.80921052631579e-06,
+      "loss": 0.5584,
+      "step": 4046
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.808662280701755e-06,
+      "loss": 0.4484,
+      "step": 4047
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.80811403508772e-06,
+      "loss": 0.8632,
+      "step": 4048
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.807565789473684e-06,
+      "loss": 0.8337,
+      "step": 4049
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.80701754385965e-06,
+      "loss": 1.5052,
+      "step": 4050
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.806469298245614e-06,
+      "loss": 0.8328,
+      "step": 4051
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.80592105263158e-06,
+      "loss": 0.85,
+      "step": 4052
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.805372807017545e-06,
+      "loss": 1.3498,
+      "step": 4053
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.80482456140351e-06,
+      "loss": 1.0664,
+      "step": 4054
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.804276315789473e-06,
+      "loss": 0.5082,
+      "step": 4055
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.803728070175439e-06,
+      "loss": 0.7882,
+      "step": 4056
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.803179824561405e-06,
+      "loss": 1.1735,
+      "step": 4057
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.802631578947369e-06,
+      "loss": 0.5937,
+      "step": 4058
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.802083333333333e-06,
+      "loss": 1.208,
+      "step": 4059
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.801535087719298e-06,
+      "loss": 0.4047,
+      "step": 4060
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.800986842105264e-06,
+      "loss": 0.5406,
+      "step": 4061
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.800438596491228e-06,
+      "loss": 0.6923,
+      "step": 4062
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.799890350877194e-06,
+      "loss": 0.7245,
+      "step": 4063
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.79934210526316e-06,
+      "loss": 0.8063,
+      "step": 4064
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.798793859649123e-06,
+      "loss": 0.3979,
+      "step": 4065
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.798245614035089e-06,
+      "loss": 0.4782,
+      "step": 4066
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.797697368421053e-06,
+      "loss": 1.4077,
+      "step": 4067
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.797149122807019e-06,
+      "loss": 0.398,
+      "step": 4068
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.796600877192984e-06,
+      "loss": 1.0799,
+      "step": 4069
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.796052631578948e-06,
+      "loss": 1.9221,
+      "step": 4070
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.795504385964912e-06,
+      "loss": 0.3694,
+      "step": 4071
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.794956140350878e-06,
+      "loss": 1.1499,
+      "step": 4072
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.794407894736844e-06,
+      "loss": 0.6151,
+      "step": 4073
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.793859649122808e-06,
+      "loss": 1.1384,
+      "step": 4074
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.793311403508772e-06,
+      "loss": 1.1286,
+      "step": 4075
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.792763157894737e-06,
+      "loss": 0.5065,
+      "step": 4076
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.792214912280703e-06,
+      "loss": 0.5954,
+      "step": 4077
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.791666666666667e-06,
+      "loss": 1.0373,
+      "step": 4078
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.791118421052633e-06,
+      "loss": 0.5834,
+      "step": 4079
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.790570175438598e-06,
+      "loss": 0.4745,
+      "step": 4080
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.790021929824562e-06,
+      "loss": 0.7804,
+      "step": 4081
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.789473684210526e-06,
+      "loss": 1.2003,
+      "step": 4082
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.788925438596492e-06,
+      "loss": 1.1024,
+      "step": 4083
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.788377192982458e-06,
+      "loss": 0.7154,
+      "step": 4084
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.787828947368422e-06,
+      "loss": 1.2022,
+      "step": 4085
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.787280701754385e-06,
+      "loss": 0.561,
+      "step": 4086
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.786732456140351e-06,
+      "loss": 0.865,
+      "step": 4087
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.786184210526317e-06,
+      "loss": 1.0059,
+      "step": 4088
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.78563596491228e-06,
+      "loss": 0.9398,
+      "step": 4089
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.785087719298246e-06,
+      "loss": 0.987,
+      "step": 4090
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.784539473684212e-06,
+      "loss": 1.7833,
+      "step": 4091
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.783991228070176e-06,
+      "loss": 1.3313,
+      "step": 4092
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.78344298245614e-06,
+      "loss": 0.8305,
+      "step": 4093
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.782894736842106e-06,
+      "loss": 1.1704,
+      "step": 4094
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.782346491228071e-06,
+      "loss": 0.8613,
+      "step": 4095
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.781798245614035e-06,
+      "loss": 1.2264,
+      "step": 4096
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.78125e-06,
+      "loss": 1.0648,
+      "step": 4097
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.780701754385965e-06,
+      "loss": 0.4191,
+      "step": 4098
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.78015350877193e-06,
+      "loss": 0.8631,
+      "step": 4099
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.779605263157895e-06,
+      "loss": 0.828,
+      "step": 4100
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.77905701754386e-06,
+      "loss": 0.8731,
+      "step": 4101
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.778508771929824e-06,
+      "loss": 1.5338,
+      "step": 4102
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.77796052631579e-06,
+      "loss": 1.5278,
+      "step": 4103
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.777412280701756e-06,
+      "loss": 0.5063,
+      "step": 4104
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.77686403508772e-06,
+      "loss": 0.9023,
+      "step": 4105
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.776315789473685e-06,
+      "loss": 1.5501,
+      "step": 4106
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.775767543859651e-06,
+      "loss": 0.9772,
+      "step": 4107
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.775219298245615e-06,
+      "loss": 0.7387,
+      "step": 4108
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.774671052631579e-06,
+      "loss": 0.8304,
+      "step": 4109
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.774122807017545e-06,
+      "loss": 0.3982,
+      "step": 4110
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.77357456140351e-06,
+      "loss": 0.5234,
+      "step": 4111
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.773026315789474e-06,
+      "loss": 0.9379,
+      "step": 4112
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.772478070175438e-06,
+      "loss": 0.2739,
+      "step": 4113
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.771929824561404e-06,
+      "loss": 0.9469,
+      "step": 4114
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.77138157894737e-06,
+      "loss": 0.6413,
+      "step": 4115
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.770833333333334e-06,
+      "loss": 1.5063,
+      "step": 4116
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.7702850877193e-06,
+      "loss": 0.5664,
+      "step": 4117
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.769736842105263e-06,
+      "loss": 0.3269,
+      "step": 4118
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.769188596491229e-06,
+      "loss": 1.5635,
+      "step": 4119
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.768640350877193e-06,
+      "loss": 0.9538,
+      "step": 4120
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.768092105263159e-06,
+      "loss": 1.1415,
+      "step": 4121
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.767543859649124e-06,
+      "loss": 0.5895,
+      "step": 4122
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.766995614035088e-06,
+      "loss": 0.4107,
+      "step": 4123
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.766447368421052e-06,
+      "loss": 0.6373,
+      "step": 4124
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.765899122807018e-06,
+      "loss": 0.8439,
+      "step": 4125
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.765350877192984e-06,
+      "loss": 0.9245,
+      "step": 4126
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.764802631578948e-06,
+      "loss": 0.5308,
+      "step": 4127
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.764254385964913e-06,
+      "loss": 0.652,
+      "step": 4128
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.763706140350877e-06,
+      "loss": 1.2799,
+      "step": 4129
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.763157894736843e-06,
+      "loss": 0.7613,
+      "step": 4130
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.762609649122807e-06,
+      "loss": 1.1265,
+      "step": 4131
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.762061403508773e-06,
+      "loss": 0.8261,
+      "step": 4132
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.761513157894738e-06,
+      "loss": 0.6408,
+      "step": 4133
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.760964912280702e-06,
+      "loss": 0.9511,
+      "step": 4134
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.760416666666666e-06,
+      "loss": 0.6009,
+      "step": 4135
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.759868421052632e-06,
+      "loss": 1.591,
+      "step": 4136
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.759320175438597e-06,
+      "loss": 1.3348,
+      "step": 4137
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.758771929824561e-06,
+      "loss": 1.2798,
+      "step": 4138
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.758223684210527e-06,
+      "loss": 1.1434,
+      "step": 4139
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.757675438596491e-06,
+      "loss": 1.0963,
+      "step": 4140
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.757127192982457e-06,
+      "loss": 0.9345,
+      "step": 4141
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.756578947368422e-06,
+      "loss": 0.3923,
+      "step": 4142
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.756030701754386e-06,
+      "loss": 0.5987,
+      "step": 4143
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.755482456140352e-06,
+      "loss": 1.279,
+      "step": 4144
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.754934210526316e-06,
+      "loss": 0.2665,
+      "step": 4145
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.754385964912282e-06,
+      "loss": 0.6459,
+      "step": 4146
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.753837719298246e-06,
+      "loss": 1.3335,
+      "step": 4147
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.753289473684211e-06,
+      "loss": 0.8764,
+      "step": 4148
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.752741228070177e-06,
+      "loss": 0.935,
+      "step": 4149
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.752192982456141e-06,
+      "loss": 1.4888,
+      "step": 4150
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.751644736842105e-06,
+      "loss": 0.8592,
+      "step": 4151
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.75109649122807e-06,
+      "loss": 0.9042,
+      "step": 4152
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.750548245614036e-06,
+      "loss": 0.9305,
+      "step": 4153
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.75e-06,
+      "loss": 0.5212,
+      "step": 4154
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.749451754385966e-06,
+      "loss": 1.1361,
+      "step": 4155
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.74890350877193e-06,
+      "loss": 1.1936,
+      "step": 4156
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.748355263157896e-06,
+      "loss": 1.4334,
+      "step": 4157
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.74780701754386e-06,
+      "loss": 1.3436,
+      "step": 4158
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.747258771929825e-06,
+      "loss": 1.0549,
+      "step": 4159
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.746710526315791e-06,
+      "loss": 0.8019,
+      "step": 4160
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.746162280701755e-06,
+      "loss": 1.0353,
+      "step": 4161
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.745614035087719e-06,
+      "loss": 0.6201,
+      "step": 4162
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.745065789473685e-06,
+      "loss": 0.6426,
+      "step": 4163
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.74451754385965e-06,
+      "loss": 1.0659,
+      "step": 4164
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.743969298245614e-06,
+      "loss": 0.9483,
+      "step": 4165
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.74342105263158e-06,
+      "loss": 0.6477,
+      "step": 4166
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.742872807017544e-06,
+      "loss": 1.0172,
+      "step": 4167
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.74232456140351e-06,
+      "loss": 0.7071,
+      "step": 4168
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.741776315789474e-06,
+      "loss": 0.6359,
+      "step": 4169
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.74122807017544e-06,
+      "loss": 0.4972,
+      "step": 4170
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.740679824561405e-06,
+      "loss": 0.7354,
+      "step": 4171
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.740131578947369e-06,
+      "loss": 0.4428,
+      "step": 4172
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.739583333333333e-06,
+      "loss": 1.3033,
+      "step": 4173
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.739035087719299e-06,
+      "loss": 1.0824,
+      "step": 4174
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.738486842105264e-06,
+      "loss": 0.6764,
+      "step": 4175
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.73793859649123e-06,
+      "loss": 0.8093,
+      "step": 4176
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.737390350877194e-06,
+      "loss": 0.8805,
+      "step": 4177
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.736842105263158e-06,
+      "loss": 0.943,
+      "step": 4178
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.736293859649123e-06,
+      "loss": 0.3575,
+      "step": 4179
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.73574561403509e-06,
+      "loss": 1.3702,
+      "step": 4180
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.735197368421053e-06,
+      "loss": 1.2969,
+      "step": 4181
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.734649122807019e-06,
+      "loss": 1.0077,
+      "step": 4182
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.734100877192983e-06,
+      "loss": 1.7885,
+      "step": 4183
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.733552631578948e-06,
+      "loss": 1.1147,
+      "step": 4184
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.733004385964912e-06,
+      "loss": 1.3046,
+      "step": 4185
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.732456140350878e-06,
+      "loss": 1.0391,
+      "step": 4186
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.731907894736844e-06,
+      "loss": 1.0928,
+      "step": 4187
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.731359649122808e-06,
+      "loss": 0.9267,
+      "step": 4188
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.730811403508772e-06,
+      "loss": 1.3018,
+      "step": 4189
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.730263157894737e-06,
+      "loss": 0.8926,
+      "step": 4190
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.729714912280703e-06,
+      "loss": 0.8033,
+      "step": 4191
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.729166666666667e-06,
+      "loss": 1.2233,
+      "step": 4192
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.728618421052633e-06,
+      "loss": 0.4249,
+      "step": 4193
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.728070175438597e-06,
+      "loss": 1.2322,
+      "step": 4194
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.727521929824562e-06,
+      "loss": 0.3988,
+      "step": 4195
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.726973684210526e-06,
+      "loss": 0.8611,
+      "step": 4196
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.726425438596492e-06,
+      "loss": 0.9869,
+      "step": 4197
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.725877192982458e-06,
+      "loss": 1.0349,
+      "step": 4198
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.725328947368422e-06,
+      "loss": 0.7109,
+      "step": 4199
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.724780701754386e-06,
+      "loss": 2.044,
+      "step": 4200
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.724232456140351e-06,
+      "loss": 0.9821,
+      "step": 4201
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.723684210526317e-06,
+      "loss": 1.4743,
+      "step": 4202
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.723135964912281e-06,
+      "loss": 0.3258,
+      "step": 4203
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.722587719298247e-06,
+      "loss": 0.6503,
+      "step": 4204
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.72203947368421e-06,
+      "loss": 0.5627,
+      "step": 4205
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.721491228070176e-06,
+      "loss": 0.7776,
+      "step": 4206
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.72094298245614e-06,
+      "loss": 1.0524,
+      "step": 4207
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.720394736842106e-06,
+      "loss": 0.6615,
+      "step": 4208
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.719846491228072e-06,
+      "loss": 1.035,
+      "step": 4209
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.719298245614036e-06,
+      "loss": 0.5788,
+      "step": 4210
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.71875e-06,
+      "loss": 0.4982,
+      "step": 4211
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.718201754385965e-06,
+      "loss": 0.648,
+      "step": 4212
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.717653508771931e-06,
+      "loss": 0.4763,
+      "step": 4213
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.717105263157897e-06,
+      "loss": 0.8214,
+      "step": 4214
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.71655701754386e-06,
+      "loss": 1.1314,
+      "step": 4215
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.716008771929825e-06,
+      "loss": 0.3646,
+      "step": 4216
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.71546052631579e-06,
+      "loss": 1.3176,
+      "step": 4217
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.714912280701756e-06,
+      "loss": 0.927,
+      "step": 4218
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.71436403508772e-06,
+      "loss": 1.2264,
+      "step": 4219
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.713815789473686e-06,
+      "loss": 1.0167,
+      "step": 4220
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.71326754385965e-06,
+      "loss": 0.9824,
+      "step": 4221
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.712719298245615e-06,
+      "loss": 0.692,
+      "step": 4222
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.71217105263158e-06,
+      "loss": 1.242,
+      "step": 4223
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.711622807017545e-06,
+      "loss": 0.6743,
+      "step": 4224
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.71107456140351e-06,
+      "loss": 1.2106,
+      "step": 4225
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.710526315789474e-06,
+      "loss": 0.4943,
+      "step": 4226
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.709978070175438e-06,
+      "loss": 1.0759,
+      "step": 4227
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.709429824561404e-06,
+      "loss": 0.8811,
+      "step": 4228
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.70888157894737e-06,
+      "loss": 1.1122,
+      "step": 4229
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.708333333333334e-06,
+      "loss": 0.8872,
+      "step": 4230
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.707785087719298e-06,
+      "loss": 0.5318,
+      "step": 4231
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.707236842105263e-06,
+      "loss": 1.3699,
+      "step": 4232
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.706688596491229e-06,
+      "loss": 1.2312,
+      "step": 4233
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.706140350877193e-06,
+      "loss": 0.8726,
+      "step": 4234
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.705592105263159e-06,
+      "loss": 0.5668,
+      "step": 4235
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.705043859649124e-06,
+      "loss": 1.0898,
+      "step": 4236
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.704495614035088e-06,
+      "loss": 0.8687,
+      "step": 4237
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.703947368421052e-06,
+      "loss": 0.5504,
+      "step": 4238
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.703399122807018e-06,
+      "loss": 1.0468,
+      "step": 4239
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.702850877192984e-06,
+      "loss": 0.5995,
+      "step": 4240
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.702302631578948e-06,
+      "loss": 0.8219,
+      "step": 4241
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.701754385964912e-06,
+      "loss": 0.8144,
+      "step": 4242
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.701206140350877e-06,
+      "loss": 1.7011,
+      "step": 4243
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.700657894736843e-06,
+      "loss": 0.9625,
+      "step": 4244
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.700109649122807e-06,
+      "loss": 0.4436,
+      "step": 4245
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.699561403508773e-06,
+      "loss": 0.7563,
+      "step": 4246
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.699013157894738e-06,
+      "loss": 0.8037,
+      "step": 4247
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.698464912280702e-06,
+      "loss": 1.0377,
+      "step": 4248
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.697916666666666e-06,
+      "loss": 1.2162,
+      "step": 4249
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.697368421052632e-06,
+      "loss": 1.1068,
+      "step": 4250
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.696820175438598e-06,
+      "loss": 0.4673,
+      "step": 4251
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.696271929824563e-06,
+      "loss": 0.906,
+      "step": 4252
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.695723684210527e-06,
+      "loss": 0.7863,
+      "step": 4253
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.695175438596491e-06,
+      "loss": 1.7863,
+      "step": 4254
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.694627192982457e-06,
+      "loss": 0.5411,
+      "step": 4255
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.694078947368423e-06,
+      "loss": 0.7837,
+      "step": 4256
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.693530701754387e-06,
+      "loss": 0.6227,
+      "step": 4257
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.69298245614035e-06,
+      "loss": 0.5603,
+      "step": 4258
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.692434210526316e-06,
+      "loss": 0.5036,
+      "step": 4259
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.691885964912282e-06,
+      "loss": 1.0607,
+      "step": 4260
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.691337719298246e-06,
+      "loss": 0.5774,
+      "step": 4261
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.690789473684212e-06,
+      "loss": 1.0829,
+      "step": 4262
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.690241228070177e-06,
+      "loss": 1.0467,
+      "step": 4263
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.689692982456141e-06,
+      "loss": 1.3144,
+      "step": 4264
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.689144736842105e-06,
+      "loss": 0.6753,
+      "step": 4265
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.688596491228071e-06,
+      "loss": 0.498,
+      "step": 4266
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.688048245614037e-06,
+      "loss": 0.5375,
+      "step": 4267
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.6875e-06,
+      "loss": 0.8007,
+      "step": 4268
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.686951754385964e-06,
+      "loss": 1.4951,
+      "step": 4269
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.68640350877193e-06,
+      "loss": 1.0166,
+      "step": 4270
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.685855263157896e-06,
+      "loss": 0.8069,
+      "step": 4271
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.68530701754386e-06,
+      "loss": 0.7501,
+      "step": 4272
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.684758771929825e-06,
+      "loss": 0.8202,
+      "step": 4273
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.68421052631579e-06,
+      "loss": 1.0517,
+      "step": 4274
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.683662280701755e-06,
+      "loss": 1.028,
+      "step": 4275
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.683114035087719e-06,
+      "loss": 0.9616,
+      "step": 4276
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.682565789473685e-06,
+      "loss": 0.9167,
+      "step": 4277
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.68201754385965e-06,
+      "loss": 0.67,
+      "step": 4278
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.681469298245614e-06,
+      "loss": 1.3342,
+      "step": 4279
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.680921052631578e-06,
+      "loss": 0.7082,
+      "step": 4280
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.680372807017544e-06,
+      "loss": 0.4678,
+      "step": 4281
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.67982456140351e-06,
+      "loss": 0.942,
+      "step": 4282
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.679276315789474e-06,
+      "loss": 1.0472,
+      "step": 4283
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.67872807017544e-06,
+      "loss": 0.4534,
+      "step": 4284
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.678179824561403e-06,
+      "loss": 0.6364,
+      "step": 4285
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.677631578947369e-06,
+      "loss": 0.5828,
+      "step": 4286
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.677083333333335e-06,
+      "loss": 0.6082,
+      "step": 4287
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.676535087719299e-06,
+      "loss": 1.8656,
+      "step": 4288
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.675986842105264e-06,
+      "loss": 0.6086,
+      "step": 4289
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.67543859649123e-06,
+      "loss": 1.1867,
+      "step": 4290
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.674890350877194e-06,
+      "loss": 0.719,
+      "step": 4291
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.674342105263158e-06,
+      "loss": 1.0174,
+      "step": 4292
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.673793859649124e-06,
+      "loss": 0.709,
+      "step": 4293
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.67324561403509e-06,
+      "loss": 0.6795,
+      "step": 4294
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.672697368421053e-06,
+      "loss": 0.5597,
+      "step": 4295
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.672149122807017e-06,
+      "loss": 0.701,
+      "step": 4296
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.671600877192983e-06,
+      "loss": 0.8465,
+      "step": 4297
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.671052631578949e-06,
+      "loss": 0.6229,
+      "step": 4298
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.670504385964913e-06,
+      "loss": 0.7982,
+      "step": 4299
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.669956140350878e-06,
+      "loss": 1.1187,
+      "step": 4300
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.669407894736842e-06,
+      "loss": 1.0846,
+      "step": 4301
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.668859649122808e-06,
+      "loss": 0.9714,
+      "step": 4302
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.668311403508772e-06,
+      "loss": 0.9759,
+      "step": 4303
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.667763157894738e-06,
+      "loss": 1.2225,
+      "step": 4304
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.667214912280703e-06,
+      "loss": 0.9779,
+      "step": 4305
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.666666666666667e-06,
+      "loss": 1.1805,
+      "step": 4306
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.666118421052631e-06,
+      "loss": 0.6607,
+      "step": 4307
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.665570175438597e-06,
+      "loss": 1.3265,
+      "step": 4308
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.665021929824563e-06,
+      "loss": 0.9384,
+      "step": 4309
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.664473684210527e-06,
+      "loss": 0.856,
+      "step": 4310
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.663925438596492e-06,
+      "loss": 0.876,
+      "step": 4311
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.663377192982456e-06,
+      "loss": 0.5833,
+      "step": 4312
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.662828947368422e-06,
+      "loss": 0.511,
+      "step": 4313
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.662280701754386e-06,
+      "loss": 1.1864,
+      "step": 4314
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.661732456140352e-06,
+      "loss": 0.5525,
+      "step": 4315
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.661184210526317e-06,
+      "loss": 0.9988,
+      "step": 4316
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.660635964912281e-06,
+      "loss": 1.6479,
+      "step": 4317
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.660087719298245e-06,
+      "loss": 0.8982,
+      "step": 4318
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.65953947368421e-06,
+      "loss": 1.365,
+      "step": 4319
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.658991228070176e-06,
+      "loss": 0.6341,
+      "step": 4320
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.65844298245614e-06,
+      "loss": 0.4042,
+      "step": 4321
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.657894736842106e-06,
+      "loss": 0.962,
+      "step": 4322
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.65734649122807e-06,
+      "loss": 0.7276,
+      "step": 4323
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.656798245614036e-06,
+      "loss": 1.008,
+      "step": 4324
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.656250000000001e-06,
+      "loss": 0.6532,
+      "step": 4325
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.655701754385965e-06,
+      "loss": 1.2976,
+      "step": 4326
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.655153508771931e-06,
+      "loss": 0.6468,
+      "step": 4327
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.654605263157895e-06,
+      "loss": 1.6532,
+      "step": 4328
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.65405701754386e-06,
+      "loss": 0.7593,
+      "step": 4329
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.653508771929825e-06,
+      "loss": 0.7324,
+      "step": 4330
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.65296052631579e-06,
+      "loss": 0.4938,
+      "step": 4331
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.652412280701756e-06,
+      "loss": 0.8113,
+      "step": 4332
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.65186403508772e-06,
+      "loss": 1.1823,
+      "step": 4333
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.651315789473684e-06,
+      "loss": 1.2299,
+      "step": 4334
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.65076754385965e-06,
+      "loss": 1.0264,
+      "step": 4335
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.650219298245615e-06,
+      "loss": 0.886,
+      "step": 4336
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.64967105263158e-06,
+      "loss": 0.5042,
+      "step": 4337
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.649122807017545e-06,
+      "loss": 0.7218,
+      "step": 4338
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.648574561403509e-06,
+      "loss": 0.5965,
+      "step": 4339
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.648026315789475e-06,
+      "loss": 0.7975,
+      "step": 4340
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.647478070175439e-06,
+      "loss": 0.5162,
+      "step": 4341
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.646929824561404e-06,
+      "loss": 0.5998,
+      "step": 4342
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.64638157894737e-06,
+      "loss": 1.5558,
+      "step": 4343
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.645833333333334e-06,
+      "loss": 0.5293,
+      "step": 4344
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.645285087719298e-06,
+      "loss": 0.663,
+      "step": 4345
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.644736842105264e-06,
+      "loss": 0.4245,
+      "step": 4346
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.64418859649123e-06,
+      "loss": 0.5525,
+      "step": 4347
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.643640350877193e-06,
+      "loss": 1.0903,
+      "step": 4348
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.643092105263159e-06,
+      "loss": 0.7472,
+      "step": 4349
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.642543859649123e-06,
+      "loss": 0.6122,
+      "step": 4350
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.641995614035089e-06,
+      "loss": 1.0066,
+      "step": 4351
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.641447368421053e-06,
+      "loss": 0.959,
+      "step": 4352
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.640899122807018e-06,
+      "loss": 1.3734,
+      "step": 4353
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.640350877192984e-06,
+      "loss": 0.988,
+      "step": 4354
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.639802631578948e-06,
+      "loss": 1.4452,
+      "step": 4355
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.639254385964912e-06,
+      "loss": 0.6453,
+      "step": 4356
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.638706140350878e-06,
+      "loss": 1.0064,
+      "step": 4357
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.638157894736843e-06,
+      "loss": 0.9709,
+      "step": 4358
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.637609649122807e-06,
+      "loss": 0.5861,
+      "step": 4359
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.637061403508773e-06,
+      "loss": 0.6066,
+      "step": 4360
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.636513157894737e-06,
+      "loss": 0.8996,
+      "step": 4361
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.635964912280703e-06,
+      "loss": 1.8432,
+      "step": 4362
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.635416666666668e-06,
+      "loss": 1.511,
+      "step": 4363
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.634868421052632e-06,
+      "loss": 1.3865,
+      "step": 4364
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.634320175438598e-06,
+      "loss": 1.0279,
+      "step": 4365
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.633771929824562e-06,
+      "loss": 0.9437,
+      "step": 4366
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.633223684210527e-06,
+      "loss": 0.6698,
+      "step": 4367
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.632675438596491e-06,
+      "loss": 0.7084,
+      "step": 4368
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.632127192982457e-06,
+      "loss": 1.0782,
+      "step": 4369
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.631578947368423e-06,
+      "loss": 0.6429,
+      "step": 4370
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.631030701754387e-06,
+      "loss": 0.9767,
+      "step": 4371
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.63048245614035e-06,
+      "loss": 1.1644,
+      "step": 4372
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.629934210526316e-06,
+      "loss": 0.8741,
+      "step": 4373
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.629385964912281e-06,
+      "loss": 1.5607,
+      "step": 4374
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.628837719298247e-06,
+      "loss": 0.4891,
+      "step": 4375
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.628289473684212e-06,
+      "loss": 0.8679,
+      "step": 4376
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.627741228070176e-06,
+      "loss": 1.0036,
+      "step": 4377
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.6271929824561405e-06,
+      "loss": 0.409,
+      "step": 4378
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.626644736842106e-06,
+      "loss": 0.8919,
+      "step": 4379
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.626096491228071e-06,
+      "loss": 0.907,
+      "step": 4380
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.625548245614036e-06,
+      "loss": 0.8356,
+      "step": 4381
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.625e-06,
+      "loss": 0.9344,
+      "step": 4382
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.6244517543859655e-06,
+      "loss": 0.702,
+      "step": 4383
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.62390350877193e-06,
+      "loss": 1.0854,
+      "step": 4384
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.623355263157895e-06,
+      "loss": 0.6604,
+      "step": 4385
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.622807017543861e-06,
+      "loss": 0.5202,
+      "step": 4386
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.622258771929825e-06,
+      "loss": 1.2573,
+      "step": 4387
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.62171052631579e-06,
+      "loss": 1.3399,
+      "step": 4388
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.621162280701755e-06,
+      "loss": 1.2189,
+      "step": 4389
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.62061403508772e-06,
+      "loss": 0.8074,
+      "step": 4390
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.620065789473685e-06,
+      "loss": 0.7872,
+      "step": 4391
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.619517543859651e-06,
+      "loss": 0.6108,
+      "step": 4392
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.618969298245615e-06,
+      "loss": 0.879,
+      "step": 4393
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.6184210526315794e-06,
+      "loss": 1.1097,
+      "step": 4394
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.617872807017544e-06,
+      "loss": 1.1624,
+      "step": 4395
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.61732456140351e-06,
+      "loss": 1.4107,
+      "step": 4396
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.616776315789475e-06,
+      "loss": 0.9163,
+      "step": 4397
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.616228070175439e-06,
+      "loss": 0.5708,
+      "step": 4398
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.6156798245614036e-06,
+      "loss": 0.6381,
+      "step": 4399
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.615131578947369e-06,
+      "loss": 1.2583,
+      "step": 4400
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.614583333333334e-06,
+      "loss": 1.2305,
+      "step": 4401
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.614035087719299e-06,
+      "loss": 0.4182,
+      "step": 4402
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.613486842105263e-06,
+      "loss": 0.5202,
+      "step": 4403
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.6129385964912285e-06,
+      "loss": 0.759,
+      "step": 4404
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.612390350877193e-06,
+      "loss": 0.4937,
+      "step": 4405
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.611842105263158e-06,
+      "loss": 0.9146,
+      "step": 4406
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.611293859649124e-06,
+      "loss": 0.9307,
+      "step": 4407
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.610745614035089e-06,
+      "loss": 1.072,
+      "step": 4408
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.610197368421053e-06,
+      "loss": 0.8889,
+      "step": 4409
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.609649122807018e-06,
+      "loss": 1.0061,
+      "step": 4410
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.609100877192983e-06,
+      "loss": 0.9255,
+      "step": 4411
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.608552631578948e-06,
+      "loss": 0.6594,
+      "step": 4412
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.608004385964914e-06,
+      "loss": 1.1804,
+      "step": 4413
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.607456140350878e-06,
+      "loss": 0.4689,
+      "step": 4414
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.6069078947368424e-06,
+      "loss": 0.7112,
+      "step": 4415
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.606359649122807e-06,
+      "loss": 0.4822,
+      "step": 4416
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.605811403508773e-06,
+      "loss": 0.9063,
+      "step": 4417
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.605263157894738e-06,
+      "loss": 0.916,
+      "step": 4418
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.604714912280703e-06,
+      "loss": 0.6534,
+      "step": 4419
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.6041666666666666e-06,
+      "loss": 1.367,
+      "step": 4420
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.603618421052632e-06,
+      "loss": 1.3616,
+      "step": 4421
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.603070175438597e-06,
+      "loss": 0.8669,
+      "step": 4422
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.602521929824562e-06,
+      "loss": 0.9641,
+      "step": 4423
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.6019736842105276e-06,
+      "loss": 0.6276,
+      "step": 4424
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.6014254385964915e-06,
+      "loss": 0.814,
+      "step": 4425
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.600877192982456e-06,
+      "loss": 0.9351,
+      "step": 4426
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.600328947368422e-06,
+      "loss": 0.9914,
+      "step": 4427
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.599780701754387e-06,
+      "loss": 0.6142,
+      "step": 4428
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.599232456140352e-06,
+      "loss": 0.8396,
+      "step": 4429
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.598684210526316e-06,
+      "loss": 1.1528,
+      "step": 4430
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.598135964912281e-06,
+      "loss": 0.6195,
+      "step": 4431
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.597587719298246e-06,
+      "loss": 0.3561,
+      "step": 4432
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.597039473684211e-06,
+      "loss": 0.789,
+      "step": 4433
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.596491228070177e-06,
+      "loss": 0.683,
+      "step": 4434
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.5959429824561415e-06,
+      "loss": 1.2223,
+      "step": 4435
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.5953947368421055e-06,
+      "loss": 0.9378,
+      "step": 4436
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.59484649122807e-06,
+      "loss": 0.5926,
+      "step": 4437
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.594298245614036e-06,
+      "loss": 1.8528,
+      "step": 4438
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.593750000000001e-06,
+      "loss": 1.1476,
+      "step": 4439
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.593201754385966e-06,
+      "loss": 0.606,
+      "step": 4440
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.59265350877193e-06,
+      "loss": 1.1105,
+      "step": 4441
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.592105263157895e-06,
+      "loss": 0.9924,
+      "step": 4442
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.59155701754386e-06,
+      "loss": 0.9487,
+      "step": 4443
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.591008771929826e-06,
+      "loss": 0.8217,
+      "step": 4444
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.5904605263157906e-06,
+      "loss": 1.0865,
+      "step": 4445
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.5899122807017545e-06,
+      "loss": 1.0288,
+      "step": 4446
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.589364035087719e-06,
+      "loss": 1.3485,
+      "step": 4447
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.588815789473685e-06,
+      "loss": 1.0231,
+      "step": 4448
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.58826754385965e-06,
+      "loss": 1.3151,
+      "step": 4449
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.587719298245615e-06,
+      "loss": 0.6347,
+      "step": 4450
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.58717105263158e-06,
+      "loss": 0.8902,
+      "step": 4451
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.586622807017544e-06,
+      "loss": 0.821,
+      "step": 4452
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.586074561403509e-06,
+      "loss": 0.7635,
+      "step": 4453
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.585526315789474e-06,
+      "loss": 1.0729,
+      "step": 4454
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.58497807017544e-06,
+      "loss": 0.2919,
+      "step": 4455
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.5844298245614045e-06,
+      "loss": 1.2476,
+      "step": 4456
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.5838815789473685e-06,
+      "loss": 0.8638,
+      "step": 4457
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.583333333333333e-06,
+      "loss": 1.0379,
+      "step": 4458
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.582785087719299e-06,
+      "loss": 0.9746,
+      "step": 4459
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.582236842105264e-06,
+      "loss": 1.2192,
+      "step": 4460
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.581688596491229e-06,
+      "loss": 0.7655,
+      "step": 4461
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.581140350877194e-06,
+      "loss": 0.3962,
+      "step": 4462
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.580592105263158e-06,
+      "loss": 0.8894,
+      "step": 4463
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.580043859649123e-06,
+      "loss": 0.7792,
+      "step": 4464
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.579495614035089e-06,
+      "loss": 1.61,
+      "step": 4465
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.578947368421054e-06,
+      "loss": 0.6618,
+      "step": 4466
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.578399122807018e-06,
+      "loss": 0.728,
+      "step": 4467
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.577850877192982e-06,
+      "loss": 1.2796,
+      "step": 4468
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.577302631578948e-06,
+      "loss": 0.8964,
+      "step": 4469
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.576754385964913e-06,
+      "loss": 0.5565,
+      "step": 4470
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.576206140350878e-06,
+      "loss": 0.8131,
+      "step": 4471
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.575657894736843e-06,
+      "loss": 0.7506,
+      "step": 4472
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.575109649122807e-06,
+      "loss": 0.7265,
+      "step": 4473
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.574561403508772e-06,
+      "loss": 1.4072,
+      "step": 4474
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.574013157894737e-06,
+      "loss": 0.8862,
+      "step": 4475
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.573464912280703e-06,
+      "loss": 1.0047,
+      "step": 4476
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.5729166666666675e-06,
+      "loss": 0.5973,
+      "step": 4477
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.572368421052632e-06,
+      "loss": 0.744,
+      "step": 4478
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.571820175438596e-06,
+      "loss": 0.6374,
+      "step": 4479
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.571271929824562e-06,
+      "loss": 0.9589,
+      "step": 4480
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.570723684210527e-06,
+      "loss": 0.9633,
+      "step": 4481
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.5701754385964925e-06,
+      "loss": 1.0638,
+      "step": 4482
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.569627192982457e-06,
+      "loss": 0.5837,
+      "step": 4483
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.569078947368421e-06,
+      "loss": 0.9012,
+      "step": 4484
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.568530701754386e-06,
+      "loss": 1.4206,
+      "step": 4485
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.567982456140352e-06,
+      "loss": 0.7859,
+      "step": 4486
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.567434210526317e-06,
+      "loss": 1.2151,
+      "step": 4487
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.566885964912281e-06,
+      "loss": 0.9796,
+      "step": 4488
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.566337719298247e-06,
+      "loss": 0.4576,
+      "step": 4489
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.565789473684211e-06,
+      "loss": 0.705,
+      "step": 4490
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.565241228070176e-06,
+      "loss": 0.8622,
+      "step": 4491
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.564692982456141e-06,
+      "loss": 0.4722,
+      "step": 4492
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.564144736842106e-06,
+      "loss": 0.8764,
+      "step": 4493
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.563596491228071e-06,
+      "loss": 0.43,
+      "step": 4494
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.563048245614035e-06,
+      "loss": 1.6173,
+      "step": 4495
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.5625e-06,
+      "loss": 0.7217,
+      "step": 4496
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.561951754385966e-06,
+      "loss": 0.6714,
+      "step": 4497
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.5614035087719305e-06,
+      "loss": 1.4042,
+      "step": 4498
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.560855263157895e-06,
+      "loss": 0.7178,
+      "step": 4499
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.56030701754386e-06,
+      "loss": 0.9836,
+      "step": 4500
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.559758771929825e-06,
+      "loss": 0.6344,
+      "step": 4501
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.55921052631579e-06,
+      "loss": 0.6517,
+      "step": 4502
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.5586622807017555e-06,
+      "loss": 0.3978,
+      "step": 4503
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.55811403508772e-06,
+      "loss": 0.9508,
+      "step": 4504
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.557565789473685e-06,
+      "loss": 1.3863,
+      "step": 4505
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.557017543859649e-06,
+      "loss": 0.4967,
+      "step": 4506
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.556469298245615e-06,
+      "loss": 0.8135,
+      "step": 4507
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.55592105263158e-06,
+      "loss": 0.649,
+      "step": 4508
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.5553728070175444e-06,
+      "loss": 0.9758,
+      "step": 4509
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.55482456140351e-06,
+      "loss": 0.7697,
+      "step": 4510
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.554276315789474e-06,
+      "loss": 1.066,
+      "step": 4511
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.553728070175439e-06,
+      "loss": 0.9348,
+      "step": 4512
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.553179824561404e-06,
+      "loss": 0.66,
+      "step": 4513
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.552631578947369e-06,
+      "loss": 1.0803,
+      "step": 4514
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.552083333333334e-06,
+      "loss": 0.8696,
+      "step": 4515
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.551535087719298e-06,
+      "loss": 0.8939,
+      "step": 4516
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.550986842105263e-06,
+      "loss": 1.6859,
+      "step": 4517
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.550438596491229e-06,
+      "loss": 0.6103,
+      "step": 4518
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.5498903508771935e-06,
+      "loss": 0.8869,
+      "step": 4519
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.549342105263159e-06,
+      "loss": 0.4326,
+      "step": 4520
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.548793859649124e-06,
+      "loss": 1.2311,
+      "step": 4521
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.548245614035088e-06,
+      "loss": 0.961,
+      "step": 4522
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.547697368421053e-06,
+      "loss": 0.8304,
+      "step": 4523
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.5471491228070185e-06,
+      "loss": 0.669,
+      "step": 4524
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.546600877192983e-06,
+      "loss": 0.6154,
+      "step": 4525
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.546052631578948e-06,
+      "loss": 1.1086,
+      "step": 4526
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.545504385964912e-06,
+      "loss": 0.8232,
+      "step": 4527
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.544956140350878e-06,
+      "loss": 1.1561,
+      "step": 4528
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.544407894736843e-06,
+      "loss": 0.6974,
+      "step": 4529
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.5438596491228074e-06,
+      "loss": 0.8591,
+      "step": 4530
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.543311403508773e-06,
+      "loss": 0.7404,
+      "step": 4531
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.542763157894738e-06,
+      "loss": 0.9623,
+      "step": 4532
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.542214912280702e-06,
+      "loss": 1.6693,
+      "step": 4533
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.541666666666667e-06,
+      "loss": 0.5357,
+      "step": 4534
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.541118421052632e-06,
+      "loss": 0.6352,
+      "step": 4535
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.540570175438597e-06,
+      "loss": 1.0207,
+      "step": 4536
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.540021929824563e-06,
+      "loss": 1.2422,
+      "step": 4537
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.539473684210527e-06,
+      "loss": 0.4482,
+      "step": 4538
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.538925438596492e-06,
+      "loss": 1.0758,
+      "step": 4539
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.5383771929824565e-06,
+      "loss": 1.4989,
+      "step": 4540
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.537828947368422e-06,
+      "loss": 1.1475,
+      "step": 4541
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.537280701754387e-06,
+      "loss": 1.0833,
+      "step": 4542
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.536732456140351e-06,
+      "loss": 0.7275,
+      "step": 4543
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.536184210526316e-06,
+      "loss": 1.0697,
+      "step": 4544
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.5356359649122815e-06,
+      "loss": 0.6814,
+      "step": 4545
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.535087719298246e-06,
+      "loss": 0.5995,
+      "step": 4546
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.534539473684211e-06,
+      "loss": 0.3951,
+      "step": 4547
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.533991228070177e-06,
+      "loss": 1.2381,
+      "step": 4548
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.533442982456141e-06,
+      "loss": 0.6889,
+      "step": 4549
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.532894736842106e-06,
+      "loss": 0.7902,
+      "step": 4550
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.5323464912280705e-06,
+      "loss": 1.0847,
+      "step": 4551
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.531798245614036e-06,
+      "loss": 1.2917,
+      "step": 4552
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.531250000000001e-06,
+      "loss": 0.8329,
+      "step": 4553
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.530701754385965e-06,
+      "loss": 0.4502,
+      "step": 4554
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.530153508771931e-06,
+      "loss": 1.0006,
+      "step": 4555
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.5296052631578954e-06,
+      "loss": 0.5009,
+      "step": 4556
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.52905701754386e-06,
+      "loss": 0.8739,
+      "step": 4557
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.528508771929826e-06,
+      "loss": 1.0985,
+      "step": 4558
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.52796052631579e-06,
+      "loss": 0.6725,
+      "step": 4559
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.527412280701755e-06,
+      "loss": 0.7724,
+      "step": 4560
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.5268640350877195e-06,
+      "loss": 1.0954,
+      "step": 4561
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.526315789473685e-06,
+      "loss": 0.567,
+      "step": 4562
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.52576754385965e-06,
+      "loss": 0.7632,
+      "step": 4563
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.525219298245615e-06,
+      "loss": 0.6367,
+      "step": 4564
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.524671052631579e-06,
+      "loss": 0.5923,
+      "step": 4565
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.5241228070175445e-06,
+      "loss": 0.8777,
+      "step": 4566
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.523574561403509e-06,
+      "loss": 0.9702,
+      "step": 4567
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.523026315789474e-06,
+      "loss": 0.9805,
+      "step": 4568
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.52247807017544e-06,
+      "loss": 1.0726,
+      "step": 4569
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.521929824561404e-06,
+      "loss": 0.844,
+      "step": 4570
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.521381578947369e-06,
+      "loss": 1.1262,
+      "step": 4571
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.5208333333333335e-06,
+      "loss": 1.0881,
+      "step": 4572
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.520285087719299e-06,
+      "loss": 1.1786,
+      "step": 4573
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.519736842105264e-06,
+      "loss": 0.8648,
+      "step": 4574
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.51918859649123e-06,
+      "loss": 1.1309,
+      "step": 4575
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.518640350877194e-06,
+      "loss": 0.6315,
+      "step": 4576
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.5180921052631584e-06,
+      "loss": 0.74,
+      "step": 4577
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.517543859649123e-06,
+      "loss": 0.9621,
+      "step": 4578
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.516995614035089e-06,
+      "loss": 1.0585,
+      "step": 4579
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.516447368421054e-06,
+      "loss": 1.0294,
+      "step": 4580
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.515899122807018e-06,
+      "loss": 1.0653,
+      "step": 4581
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.5153508771929826e-06,
+      "loss": 1.1965,
+      "step": 4582
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.514802631578948e-06,
+      "loss": 0.7714,
+      "step": 4583
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.514254385964913e-06,
+      "loss": 0.5371,
+      "step": 4584
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.513706140350878e-06,
+      "loss": 1.0981,
+      "step": 4585
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.513157894736842e-06,
+      "loss": 0.286,
+      "step": 4586
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.5126096491228075e-06,
+      "loss": 0.9395,
+      "step": 4587
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.512061403508772e-06,
+      "loss": 0.414,
+      "step": 4588
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.511513157894737e-06,
+      "loss": 0.5925,
+      "step": 4589
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.510964912280703e-06,
+      "loss": 0.7054,
+      "step": 4590
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.510416666666668e-06,
+      "loss": 1.3712,
+      "step": 4591
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.509868421052632e-06,
+      "loss": 0.4385,
+      "step": 4592
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.509320175438597e-06,
+      "loss": 0.9219,
+      "step": 4593
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.508771929824562e-06,
+      "loss": 0.7257,
+      "step": 4594
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.508223684210527e-06,
+      "loss": 1.1314,
+      "step": 4595
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.507675438596493e-06,
+      "loss": 1.5018,
+      "step": 4596
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.507127192982457e-06,
+      "loss": 1.2574,
+      "step": 4597
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.5065789473684214e-06,
+      "loss": 1.0724,
+      "step": 4598
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.506030701754386e-06,
+      "loss": 0.986,
+      "step": 4599
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.505482456140352e-06,
+      "loss": 1.1611,
+      "step": 4600
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.504934210526317e-06,
+      "loss": 0.5293,
+      "step": 4601
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.504385964912281e-06,
+      "loss": 0.654,
+      "step": 4602
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.5038377192982456e-06,
+      "loss": 0.832,
+      "step": 4603
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.503289473684211e-06,
+      "loss": 0.5892,
+      "step": 4604
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.502741228070176e-06,
+      "loss": 0.7912,
+      "step": 4605
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.502192982456141e-06,
+      "loss": 1.349,
+      "step": 4606
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.5016447368421066e-06,
+      "loss": 0.6697,
+      "step": 4607
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.5010964912280705e-06,
+      "loss": 1.1385,
+      "step": 4608
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.500548245614035e-06,
+      "loss": 1.2116,
+      "step": 4609
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 0.7844,
+      "step": 4610
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.499451754385966e-06,
+      "loss": 1.0189,
+      "step": 4611
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.498903508771931e-06,
+      "loss": 0.9234,
+      "step": 4612
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.498355263157895e-06,
+      "loss": 1.1788,
+      "step": 4613
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.49780701754386e-06,
+      "loss": 1.0465,
+      "step": 4614
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.497258771929825e-06,
+      "loss": 0.5258,
+      "step": 4615
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.49671052631579e-06,
+      "loss": 1.0917,
+      "step": 4616
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.496162280701756e-06,
+      "loss": 1.3661,
+      "step": 4617
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.4956140350877205e-06,
+      "loss": 0.6478,
+      "step": 4618
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.4950657894736845e-06,
+      "loss": 0.9318,
+      "step": 4619
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.494517543859649e-06,
+      "loss": 0.6905,
+      "step": 4620
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.493969298245615e-06,
+      "loss": 1.0638,
+      "step": 4621
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.49342105263158e-06,
+      "loss": 0.6606,
+      "step": 4622
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.492872807017545e-06,
+      "loss": 0.5466,
+      "step": 4623
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.492324561403509e-06,
+      "loss": 1.0557,
+      "step": 4624
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.491776315789474e-06,
+      "loss": 1.2346,
+      "step": 4625
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.491228070175439e-06,
+      "loss": 0.6175,
+      "step": 4626
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.490679824561404e-06,
+      "loss": 0.8098,
+      "step": 4627
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.4901315789473696e-06,
+      "loss": 0.9457,
+      "step": 4628
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.4895833333333336e-06,
+      "loss": 0.7332,
+      "step": 4629
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.489035087719298e-06,
+      "loss": 0.5504,
+      "step": 4630
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.488486842105264e-06,
+      "loss": 0.6265,
+      "step": 4631
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.487938596491229e-06,
+      "loss": 0.823,
+      "step": 4632
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.487390350877194e-06,
+      "loss": 0.9123,
+      "step": 4633
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.486842105263159e-06,
+      "loss": 0.8456,
+      "step": 4634
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.486293859649123e-06,
+      "loss": 0.5643,
+      "step": 4635
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.485745614035088e-06,
+      "loss": 0.9242,
+      "step": 4636
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.485197368421053e-06,
+      "loss": 0.9229,
+      "step": 4637
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.484649122807019e-06,
+      "loss": 1.5271,
+      "step": 4638
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.4841008771929835e-06,
+      "loss": 0.762,
+      "step": 4639
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.4835526315789475e-06,
+      "loss": 1.3839,
+      "step": 4640
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.483004385964912e-06,
+      "loss": 0.7422,
+      "step": 4641
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.482456140350878e-06,
+      "loss": 0.8534,
+      "step": 4642
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.481907894736843e-06,
+      "loss": 0.9113,
+      "step": 4643
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.481359649122808e-06,
+      "loss": 1.1487,
+      "step": 4644
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.480811403508772e-06,
+      "loss": 0.6036,
+      "step": 4645
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.480263157894737e-06,
+      "loss": 1.1034,
+      "step": 4646
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.479714912280702e-06,
+      "loss": 0.8733,
+      "step": 4647
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.479166666666668e-06,
+      "loss": 0.4894,
+      "step": 4648
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.478618421052633e-06,
+      "loss": 0.6827,
+      "step": 4649
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.478070175438597e-06,
+      "loss": 1.4214,
+      "step": 4650
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.477521929824561e-06,
+      "loss": 0.769,
+      "step": 4651
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.476973684210527e-06,
+      "loss": 0.5075,
+      "step": 4652
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.476425438596492e-06,
+      "loss": 1.002,
+      "step": 4653
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.475877192982457e-06,
+      "loss": 0.653,
+      "step": 4654
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.475328947368422e-06,
+      "loss": 0.6077,
+      "step": 4655
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.474780701754386e-06,
+      "loss": 0.8065,
+      "step": 4656
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.474232456140351e-06,
+      "loss": 0.601,
+      "step": 4657
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.473684210526316e-06,
+      "loss": 0.6755,
+      "step": 4658
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.473135964912282e-06,
+      "loss": 0.4909,
+      "step": 4659
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.4725877192982465e-06,
+      "loss": 0.5385,
+      "step": 4660
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.472039473684211e-06,
+      "loss": 0.6059,
+      "step": 4661
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.471491228070175e-06,
+      "loss": 1.1224,
+      "step": 4662
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.470942982456141e-06,
+      "loss": 0.5387,
+      "step": 4663
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.470394736842106e-06,
+      "loss": 1.2114,
+      "step": 4664
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.469846491228071e-06,
+      "loss": 1.1378,
+      "step": 4665
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.469298245614036e-06,
+      "loss": 0.7412,
+      "step": 4666
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.46875e-06,
+      "loss": 0.7863,
+      "step": 4667
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.468201754385965e-06,
+      "loss": 0.6477,
+      "step": 4668
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.467653508771931e-06,
+      "loss": 0.3909,
+      "step": 4669
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.467105263157896e-06,
+      "loss": 1.0288,
+      "step": 4670
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.46655701754386e-06,
+      "loss": 1.1912,
+      "step": 4671
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.466008771929824e-06,
+      "loss": 0.4498,
+      "step": 4672
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.46546052631579e-06,
+      "loss": 1.5564,
+      "step": 4673
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.464912280701755e-06,
+      "loss": 0.4268,
+      "step": 4674
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.46436403508772e-06,
+      "loss": 1.3497,
+      "step": 4675
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.463815789473685e-06,
+      "loss": 0.6341,
+      "step": 4676
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.46326754385965e-06,
+      "loss": 0.6127,
+      "step": 4677
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.462719298245614e-06,
+      "loss": 0.894,
+      "step": 4678
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.462171052631579e-06,
+      "loss": 2.1325,
+      "step": 4679
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.461622807017545e-06,
+      "loss": 0.5542,
+      "step": 4680
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.4610745614035095e-06,
+      "loss": 1.3898,
+      "step": 4681
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.460526315789474e-06,
+      "loss": 0.8937,
+      "step": 4682
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.459978070175438e-06,
+      "loss": 0.8512,
+      "step": 4683
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.459429824561404e-06,
+      "loss": 0.4891,
+      "step": 4684
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.458881578947369e-06,
+      "loss": 0.9081,
+      "step": 4685
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.4583333333333345e-06,
+      "loss": 0.4963,
+      "step": 4686
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.457785087719299e-06,
+      "loss": 1.147,
+      "step": 4687
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.457236842105263e-06,
+      "loss": 1.0937,
+      "step": 4688
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.456688596491228e-06,
+      "loss": 1.1827,
+      "step": 4689
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.456140350877194e-06,
+      "loss": 0.7371,
+      "step": 4690
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.455592105263159e-06,
+      "loss": 1.259,
+      "step": 4691
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.4550438596491234e-06,
+      "loss": 1.2398,
+      "step": 4692
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.454495614035089e-06,
+      "loss": 1.7743,
+      "step": 4693
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.453947368421053e-06,
+      "loss": 1.1873,
+      "step": 4694
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.453399122807018e-06,
+      "loss": 0.7477,
+      "step": 4695
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.452850877192983e-06,
+      "loss": 0.2813,
+      "step": 4696
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.452302631578948e-06,
+      "loss": 1.6899,
+      "step": 4697
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.451754385964913e-06,
+      "loss": 1.5303,
+      "step": 4698
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.451206140350877e-06,
+      "loss": 0.5837,
+      "step": 4699
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.450657894736842e-06,
+      "loss": 1.1599,
+      "step": 4700
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.450109649122808e-06,
+      "loss": 0.5427,
+      "step": 4701
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.4495614035087725e-06,
+      "loss": 0.737,
+      "step": 4702
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.449013157894738e-06,
+      "loss": 0.9166,
+      "step": 4703
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.448464912280703e-06,
+      "loss": 1.4538,
+      "step": 4704
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.447916666666667e-06,
+      "loss": 0.9891,
+      "step": 4705
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.447368421052632e-06,
+      "loss": 0.8152,
+      "step": 4706
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.4468201754385975e-06,
+      "loss": 0.8849,
+      "step": 4707
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.446271929824562e-06,
+      "loss": 1.0891,
+      "step": 4708
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.445723684210527e-06,
+      "loss": 0.5018,
+      "step": 4709
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.445175438596491e-06,
+      "loss": 0.5174,
+      "step": 4710
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.444627192982457e-06,
+      "loss": 1.2237,
+      "step": 4711
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.444078947368422e-06,
+      "loss": 0.6595,
+      "step": 4712
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.4435307017543864e-06,
+      "loss": 0.7233,
+      "step": 4713
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.442982456140352e-06,
+      "loss": 0.5238,
+      "step": 4714
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.442434210526316e-06,
+      "loss": 0.9096,
+      "step": 4715
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.441885964912281e-06,
+      "loss": 0.8436,
+      "step": 4716
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.441337719298246e-06,
+      "loss": 0.8012,
+      "step": 4717
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.440789473684211e-06,
+      "loss": 0.7899,
+      "step": 4718
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.440241228070176e-06,
+      "loss": 1.2412,
+      "step": 4719
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.439692982456141e-06,
+      "loss": 1.1603,
+      "step": 4720
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.439144736842106e-06,
+      "loss": 1.0404,
+      "step": 4721
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.438596491228071e-06,
+      "loss": 0.8575,
+      "step": 4722
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.4380482456140355e-06,
+      "loss": 1.0654,
+      "step": 4723
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.437500000000001e-06,
+      "loss": 0.9032,
+      "step": 4724
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.436951754385966e-06,
+      "loss": 0.5808,
+      "step": 4725
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.43640350877193e-06,
+      "loss": 0.6091,
+      "step": 4726
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.435855263157895e-06,
+      "loss": 1.1642,
+      "step": 4727
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.4353070175438605e-06,
+      "loss": 0.7836,
+      "step": 4728
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.434758771929825e-06,
+      "loss": 0.4331,
+      "step": 4729
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.43421052631579e-06,
+      "loss": 0.5458,
+      "step": 4730
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.433662280701754e-06,
+      "loss": 1.1672,
+      "step": 4731
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.43311403508772e-06,
+      "loss": 0.6304,
+      "step": 4732
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.432565789473685e-06,
+      "loss": 1.4064,
+      "step": 4733
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.4320175438596495e-06,
+      "loss": 0.7683,
+      "step": 4734
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.431469298245615e-06,
+      "loss": 0.9871,
+      "step": 4735
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.43092105263158e-06,
+      "loss": 0.375,
+      "step": 4736
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.430372807017544e-06,
+      "loss": 0.3226,
+      "step": 4737
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.429824561403509e-06,
+      "loss": 0.7308,
+      "step": 4738
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.4292763157894744e-06,
+      "loss": 0.8357,
+      "step": 4739
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.428728070175439e-06,
+      "loss": 0.8823,
+      "step": 4740
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.428179824561405e-06,
+      "loss": 0.7129,
+      "step": 4741
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.427631578947369e-06,
+      "loss": 0.7625,
+      "step": 4742
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.427083333333334e-06,
+      "loss": 1.0005,
+      "step": 4743
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.4265350877192986e-06,
+      "loss": 0.7827,
+      "step": 4744
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.425986842105264e-06,
+      "loss": 0.8867,
+      "step": 4745
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.425438596491229e-06,
+      "loss": 0.7482,
+      "step": 4746
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.424890350877194e-06,
+      "loss": 0.9853,
+      "step": 4747
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.424342105263158e-06,
+      "loss": 1.1261,
+      "step": 4748
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.4237938596491235e-06,
+      "loss": 0.8493,
+      "step": 4749
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.423245614035088e-06,
+      "loss": 0.8933,
+      "step": 4750
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.422697368421053e-06,
+      "loss": 0.5945,
+      "step": 4751
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.422149122807019e-06,
+      "loss": 0.852,
+      "step": 4752
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.421600877192983e-06,
+      "loss": 0.768,
+      "step": 4753
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.421052631578948e-06,
+      "loss": 0.434,
+      "step": 4754
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.4205043859649125e-06,
+      "loss": 1.3897,
+      "step": 4755
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.419956140350878e-06,
+      "loss": 1.8153,
+      "step": 4756
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.419407894736843e-06,
+      "loss": 0.6208,
+      "step": 4757
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.418859649122807e-06,
+      "loss": 0.4156,
+      "step": 4758
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.418311403508773e-06,
+      "loss": 0.7628,
+      "step": 4759
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.4177631578947374e-06,
+      "loss": 0.7463,
+      "step": 4760
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.417214912280702e-06,
+      "loss": 0.849,
+      "step": 4761
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.416666666666668e-06,
+      "loss": 0.6666,
+      "step": 4762
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.416118421052633e-06,
+      "loss": 1.2127,
+      "step": 4763
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.415570175438597e-06,
+      "loss": 1.5696,
+      "step": 4764
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.4150219298245616e-06,
+      "loss": 0.5981,
+      "step": 4765
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.414473684210527e-06,
+      "loss": 0.5593,
+      "step": 4766
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.413925438596492e-06,
+      "loss": 1.1316,
+      "step": 4767
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.413377192982457e-06,
+      "loss": 1.0105,
+      "step": 4768
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.412828947368421e-06,
+      "loss": 1.1614,
+      "step": 4769
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.4122807017543865e-06,
+      "loss": 0.853,
+      "step": 4770
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.411732456140351e-06,
+      "loss": 0.7786,
+      "step": 4771
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.411184210526316e-06,
+      "loss": 0.9543,
+      "step": 4772
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.410635964912282e-06,
+      "loss": 0.8701,
+      "step": 4773
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.410087719298247e-06,
+      "loss": 0.7072,
+      "step": 4774
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.409539473684211e-06,
+      "loss": 0.5788,
+      "step": 4775
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.4089912280701755e-06,
+      "loss": 1.0742,
+      "step": 4776
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.408442982456141e-06,
+      "loss": 0.3705,
+      "step": 4777
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.407894736842106e-06,
+      "loss": 0.6294,
+      "step": 4778
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.407346491228072e-06,
+      "loss": 0.9239,
+      "step": 4779
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.406798245614036e-06,
+      "loss": 0.9778,
+      "step": 4780
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.4062500000000005e-06,
+      "loss": 0.9782,
+      "step": 4781
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.405701754385965e-06,
+      "loss": 0.7774,
+      "step": 4782
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.405153508771931e-06,
+      "loss": 1.0101,
+      "step": 4783
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.404605263157896e-06,
+      "loss": 0.9576,
+      "step": 4784
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.40405701754386e-06,
+      "loss": 0.7679,
+      "step": 4785
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.4035087719298246e-06,
+      "loss": 1.0973,
+      "step": 4786
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.40296052631579e-06,
+      "loss": 1.0453,
+      "step": 4787
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.402412280701755e-06,
+      "loss": 1.0542,
+      "step": 4788
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.40186403508772e-06,
+      "loss": 0.8098,
+      "step": 4789
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.4013157894736856e-06,
+      "loss": 1.1282,
+      "step": 4790
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.4007675438596495e-06,
+      "loss": 0.5768,
+      "step": 4791
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.400219298245614e-06,
+      "loss": 1.3704,
+      "step": 4792
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.399671052631579e-06,
+      "loss": 0.6118,
+      "step": 4793
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.399122807017545e-06,
+      "loss": 0.4177,
+      "step": 4794
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.39857456140351e-06,
+      "loss": 1.7766,
+      "step": 4795
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.398026315789474e-06,
+      "loss": 1.0411,
+      "step": 4796
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.397478070175439e-06,
+      "loss": 0.7027,
+      "step": 4797
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.396929824561404e-06,
+      "loss": 1.2892,
+      "step": 4798
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.396381578947369e-06,
+      "loss": 1.234,
+      "step": 4799
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.395833333333335e-06,
+      "loss": 0.9944,
+      "step": 4800
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.395285087719299e-06,
+      "loss": 1.9289,
+      "step": 4801
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.3947368421052635e-06,
+      "loss": 0.5244,
+      "step": 4802
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.394188596491228e-06,
+      "loss": 0.6277,
+      "step": 4803
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.393640350877194e-06,
+      "loss": 0.7103,
+      "step": 4804
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.393092105263159e-06,
+      "loss": 0.93,
+      "step": 4805
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.392543859649124e-06,
+      "loss": 1.126,
+      "step": 4806
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.391995614035088e-06,
+      "loss": 0.3847,
+      "step": 4807
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.391447368421053e-06,
+      "loss": 0.5735,
+      "step": 4808
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.390899122807018e-06,
+      "loss": 0.9796,
+      "step": 4809
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.390350877192983e-06,
+      "loss": 2.0748,
+      "step": 4810
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.389802631578949e-06,
+      "loss": 0.7298,
+      "step": 4811
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.3892543859649126e-06,
+      "loss": 0.5665,
+      "step": 4812
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.388706140350877e-06,
+      "loss": 1.0068,
+      "step": 4813
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.388157894736843e-06,
+      "loss": 0.9266,
+      "step": 4814
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.387609649122808e-06,
+      "loss": 0.4,
+      "step": 4815
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.387061403508773e-06,
+      "loss": 1.5939,
+      "step": 4816
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.386513157894738e-06,
+      "loss": 0.7506,
+      "step": 4817
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.385964912280702e-06,
+      "loss": 0.8815,
+      "step": 4818
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.385416666666667e-06,
+      "loss": 1.3453,
+      "step": 4819
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.384868421052632e-06,
+      "loss": 0.9704,
+      "step": 4820
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.384320175438598e-06,
+      "loss": 0.6738,
+      "step": 4821
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.3837719298245625e-06,
+      "loss": 0.5996,
+      "step": 4822
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.3832236842105265e-06,
+      "loss": 0.8872,
+      "step": 4823
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.382675438596491e-06,
+      "loss": 0.5208,
+      "step": 4824
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.382127192982457e-06,
+      "loss": 1.0214,
+      "step": 4825
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.381578947368422e-06,
+      "loss": 0.4637,
+      "step": 4826
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.381030701754387e-06,
+      "loss": 0.9191,
+      "step": 4827
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.380482456140351e-06,
+      "loss": 1.3569,
+      "step": 4828
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.379934210526316e-06,
+      "loss": 0.5539,
+      "step": 4829
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.379385964912281e-06,
+      "loss": 0.6662,
+      "step": 4830
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.378837719298246e-06,
+      "loss": 0.7402,
+      "step": 4831
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.378289473684212e-06,
+      "loss": 0.6604,
+      "step": 4832
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.377741228070176e-06,
+      "loss": 0.7676,
+      "step": 4833
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.37719298245614e-06,
+      "loss": 1.3413,
+      "step": 4834
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.376644736842106e-06,
+      "loss": 0.8928,
+      "step": 4835
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.376096491228071e-06,
+      "loss": 0.8606,
+      "step": 4836
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.375548245614036e-06,
+      "loss": 0.7161,
+      "step": 4837
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.375000000000001e-06,
+      "loss": 0.5958,
+      "step": 4838
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.374451754385965e-06,
+      "loss": 0.8366,
+      "step": 4839
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.37390350877193e-06,
+      "loss": 0.9662,
+      "step": 4840
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.373355263157895e-06,
+      "loss": 0.6654,
+      "step": 4841
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.372807017543861e-06,
+      "loss": 1.0317,
+      "step": 4842
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.3722587719298255e-06,
+      "loss": 1.2373,
+      "step": 4843
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.3717105263157895e-06,
+      "loss": 0.7513,
+      "step": 4844
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.371162280701754e-06,
+      "loss": 1.2731,
+      "step": 4845
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.37061403508772e-06,
+      "loss": 1.2032,
+      "step": 4846
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.370065789473685e-06,
+      "loss": 0.9738,
+      "step": 4847
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.36951754385965e-06,
+      "loss": 0.8098,
+      "step": 4848
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.368969298245615e-06,
+      "loss": 0.9359,
+      "step": 4849
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.368421052631579e-06,
+      "loss": 0.5243,
+      "step": 4850
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.367872807017544e-06,
+      "loss": 1.175,
+      "step": 4851
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.36732456140351e-06,
+      "loss": 0.906,
+      "step": 4852
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.366776315789475e-06,
+      "loss": 0.3541,
+      "step": 4853
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.3662280701754394e-06,
+      "loss": 0.836,
+      "step": 4854
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.365679824561403e-06,
+      "loss": 0.8384,
+      "step": 4855
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.365131578947369e-06,
+      "loss": 1.116,
+      "step": 4856
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.364583333333334e-06,
+      "loss": 0.8046,
+      "step": 4857
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.364035087719299e-06,
+      "loss": 0.8457,
+      "step": 4858
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.363486842105264e-06,
+      "loss": 0.5354,
+      "step": 4859
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.362938596491229e-06,
+      "loss": 0.8103,
+      "step": 4860
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.362390350877193e-06,
+      "loss": 1.2136,
+      "step": 4861
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.361842105263158e-06,
+      "loss": 0.5078,
+      "step": 4862
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.361293859649124e-06,
+      "loss": 1.106,
+      "step": 4863
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.3607456140350885e-06,
+      "loss": 0.9914,
+      "step": 4864
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.360197368421053e-06,
+      "loss": 0.5059,
+      "step": 4865
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.359649122807017e-06,
+      "loss": 0.9433,
+      "step": 4866
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.359100877192983e-06,
+      "loss": 0.5516,
+      "step": 4867
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.358552631578948e-06,
+      "loss": 0.6839,
+      "step": 4868
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.358004385964913e-06,
+      "loss": 1.193,
+      "step": 4869
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.357456140350878e-06,
+      "loss": 0.5735,
+      "step": 4870
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.356907894736842e-06,
+      "loss": 1.1742,
+      "step": 4871
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.356359649122807e-06,
+      "loss": 0.6794,
+      "step": 4872
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.355811403508773e-06,
+      "loss": 0.8646,
+      "step": 4873
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.355263157894738e-06,
+      "loss": 0.7119,
+      "step": 4874
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.3547149122807024e-06,
+      "loss": 0.6668,
+      "step": 4875
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.354166666666668e-06,
+      "loss": 1.3529,
+      "step": 4876
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.353618421052632e-06,
+      "loss": 1.0528,
+      "step": 4877
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.353070175438597e-06,
+      "loss": 0.7604,
+      "step": 4878
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.352521929824562e-06,
+      "loss": 0.8787,
+      "step": 4879
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.351973684210527e-06,
+      "loss": 1.5816,
+      "step": 4880
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.351425438596492e-06,
+      "loss": 1.3098,
+      "step": 4881
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.350877192982456e-06,
+      "loss": 0.7961,
+      "step": 4882
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.350328947368421e-06,
+      "loss": 1.4997,
+      "step": 4883
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.349780701754387e-06,
+      "loss": 2.0058,
+      "step": 4884
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.3492324561403515e-06,
+      "loss": 1.2115,
+      "step": 4885
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.348684210526316e-06,
+      "loss": 1.1888,
+      "step": 4886
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.348135964912281e-06,
+      "loss": 1.3195,
+      "step": 4887
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.347587719298246e-06,
+      "loss": 1.1907,
+      "step": 4888
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.347039473684211e-06,
+      "loss": 0.5332,
+      "step": 4889
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.3464912280701765e-06,
+      "loss": 1.2941,
+      "step": 4890
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.345942982456141e-06,
+      "loss": 0.6941,
+      "step": 4891
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.345394736842106e-06,
+      "loss": 1.2567,
+      "step": 4892
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.34484649122807e-06,
+      "loss": 0.954,
+      "step": 4893
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.344298245614036e-06,
+      "loss": 0.9758,
+      "step": 4894
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.343750000000001e-06,
+      "loss": 0.8214,
+      "step": 4895
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.3432017543859655e-06,
+      "loss": 1.4124,
+      "step": 4896
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.342653508771931e-06,
+      "loss": 1.0561,
+      "step": 4897
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.342105263157895e-06,
+      "loss": 0.6706,
+      "step": 4898
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.34155701754386e-06,
+      "loss": 0.9721,
+      "step": 4899
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.341008771929825e-06,
+      "loss": 0.6676,
+      "step": 4900
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.34046052631579e-06,
+      "loss": 0.9746,
+      "step": 4901
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.339912280701755e-06,
+      "loss": 0.75,
+      "step": 4902
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.33936403508772e-06,
+      "loss": 0.7614,
+      "step": 4903
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.338815789473684e-06,
+      "loss": 0.9606,
+      "step": 4904
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.33826754385965e-06,
+      "loss": 0.6111,
+      "step": 4905
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.3377192982456145e-06,
+      "loss": 1.1167,
+      "step": 4906
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.33717105263158e-06,
+      "loss": 0.9223,
+      "step": 4907
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.336622807017545e-06,
+      "loss": 0.7069,
+      "step": 4908
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.336074561403509e-06,
+      "loss": 0.6525,
+      "step": 4909
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.335526315789474e-06,
+      "loss": 0.9492,
+      "step": 4910
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.3349780701754395e-06,
+      "loss": 0.6274,
+      "step": 4911
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.334429824561404e-06,
+      "loss": 0.6438,
+      "step": 4912
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.333881578947369e-06,
+      "loss": 0.9074,
+      "step": 4913
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.333333333333333e-06,
+      "loss": 0.7435,
+      "step": 4914
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.332785087719299e-06,
+      "loss": 0.9255,
+      "step": 4915
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.332236842105264e-06,
+      "loss": 0.8327,
+      "step": 4916
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.3316885964912285e-06,
+      "loss": 0.9535,
+      "step": 4917
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.331140350877194e-06,
+      "loss": 0.6731,
+      "step": 4918
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.330592105263159e-06,
+      "loss": 0.9101,
+      "step": 4919
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.330043859649123e-06,
+      "loss": 1.0987,
+      "step": 4920
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.329495614035088e-06,
+      "loss": 0.6865,
+      "step": 4921
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.3289473684210534e-06,
+      "loss": 0.5832,
+      "step": 4922
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.328399122807018e-06,
+      "loss": 0.5329,
+      "step": 4923
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.327850877192983e-06,
+      "loss": 0.5418,
+      "step": 4924
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.327302631578948e-06,
+      "loss": 1.3052,
+      "step": 4925
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.326754385964913e-06,
+      "loss": 1.1379,
+      "step": 4926
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.3262061403508776e-06,
+      "loss": 0.6685,
+      "step": 4927
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.325657894736843e-06,
+      "loss": 0.2921,
+      "step": 4928
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.325109649122808e-06,
+      "loss": 1.0549,
+      "step": 4929
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.324561403508772e-06,
+      "loss": 1.0952,
+      "step": 4930
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.324013157894737e-06,
+      "loss": 0.9549,
+      "step": 4931
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.3234649122807025e-06,
+      "loss": 0.7636,
+      "step": 4932
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.322916666666667e-06,
+      "loss": 1.0488,
+      "step": 4933
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.322368421052632e-06,
+      "loss": 1.402,
+      "step": 4934
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.321820175438598e-06,
+      "loss": 1.3472,
+      "step": 4935
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.321271929824562e-06,
+      "loss": 0.8993,
+      "step": 4936
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.320723684210527e-06,
+      "loss": 0.7997,
+      "step": 4937
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.3201754385964915e-06,
+      "loss": 0.8259,
+      "step": 4938
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.319627192982457e-06,
+      "loss": 1.2877,
+      "step": 4939
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.319078947368422e-06,
+      "loss": 0.3174,
+      "step": 4940
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.318530701754386e-06,
+      "loss": 0.5027,
+      "step": 4941
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.317982456140351e-06,
+      "loss": 1.2969,
+      "step": 4942
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.3174342105263164e-06,
+      "loss": 0.635,
+      "step": 4943
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.316885964912281e-06,
+      "loss": 1.1278,
+      "step": 4944
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.316337719298247e-06,
+      "loss": 0.5799,
+      "step": 4945
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.315789473684212e-06,
+      "loss": 1.3298,
+      "step": 4946
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.315241228070176e-06,
+      "loss": 1.2662,
+      "step": 4947
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.3146929824561406e-06,
+      "loss": 0.8888,
+      "step": 4948
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.314144736842106e-06,
+      "loss": 0.4784,
+      "step": 4949
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.313596491228071e-06,
+      "loss": 1.6531,
+      "step": 4950
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.313048245614036e-06,
+      "loss": 0.9246,
+      "step": 4951
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.3125e-06,
+      "loss": 0.9641,
+      "step": 4952
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.3119517543859655e-06,
+      "loss": 0.6602,
+      "step": 4953
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.31140350877193e-06,
+      "loss": 0.7941,
+      "step": 4954
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.310855263157895e-06,
+      "loss": 0.835,
+      "step": 4955
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.310307017543861e-06,
+      "loss": 1.2054,
+      "step": 4956
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.309758771929825e-06,
+      "loss": 1.2965,
+      "step": 4957
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.30921052631579e-06,
+      "loss": 0.5798,
+      "step": 4958
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.3086622807017545e-06,
+      "loss": 0.8663,
+      "step": 4959
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.30811403508772e-06,
+      "loss": 0.6102,
+      "step": 4960
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.307565789473685e-06,
+      "loss": 0.5577,
+      "step": 4961
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.30701754385965e-06,
+      "loss": 0.8594,
+      "step": 4962
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.306469298245615e-06,
+      "loss": 0.6299,
+      "step": 4963
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.3059210526315795e-06,
+      "loss": 1.1962,
+      "step": 4964
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.305372807017544e-06,
+      "loss": 0.465,
+      "step": 4965
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.30482456140351e-06,
+      "loss": 0.965,
+      "step": 4966
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.304276315789475e-06,
+      "loss": 1.0772,
+      "step": 4967
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.303728070175439e-06,
+      "loss": 0.7317,
+      "step": 4968
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.303179824561404e-06,
+      "loss": 1.3276,
+      "step": 4969
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.302631578947369e-06,
+      "loss": 1.1709,
+      "step": 4970
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.302083333333334e-06,
+      "loss": 0.583,
+      "step": 4971
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.301535087719299e-06,
+      "loss": 0.501,
+      "step": 4972
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.300986842105263e-06,
+      "loss": 1.5771,
+      "step": 4973
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.3004385964912286e-06,
+      "loss": 1.5154,
+      "step": 4974
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.299890350877193e-06,
+      "loss": 1.1475,
+      "step": 4975
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.299342105263158e-06,
+      "loss": 1.5374,
+      "step": 4976
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.298793859649124e-06,
+      "loss": 1.3278,
+      "step": 4977
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.298245614035089e-06,
+      "loss": 0.5695,
+      "step": 4978
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.297697368421053e-06,
+      "loss": 1.0756,
+      "step": 4979
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.297149122807018e-06,
+      "loss": 0.8778,
+      "step": 4980
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.296600877192983e-06,
+      "loss": 0.6275,
+      "step": 4981
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.296052631578948e-06,
+      "loss": 0.7597,
+      "step": 4982
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.295504385964914e-06,
+      "loss": 0.7851,
+      "step": 4983
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.294956140350878e-06,
+      "loss": 0.5078,
+      "step": 4984
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.2944078947368425e-06,
+      "loss": 0.8144,
+      "step": 4985
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.293859649122807e-06,
+      "loss": 0.874,
+      "step": 4986
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.293311403508773e-06,
+      "loss": 0.9099,
+      "step": 4987
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.292763157894738e-06,
+      "loss": 0.9909,
+      "step": 4988
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.292214912280703e-06,
+      "loss": 1.9046,
+      "step": 4989
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.291666666666667e-06,
+      "loss": 0.9011,
+      "step": 4990
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.291118421052632e-06,
+      "loss": 0.915,
+      "step": 4991
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.290570175438597e-06,
+      "loss": 0.9779,
+      "step": 4992
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.290021929824562e-06,
+      "loss": 0.4563,
+      "step": 4993
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.289473684210528e-06,
+      "loss": 0.6911,
+      "step": 4994
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.2889254385964916e-06,
+      "loss": 0.9455,
+      "step": 4995
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.288377192982456e-06,
+      "loss": 1.3516,
+      "step": 4996
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.287828947368421e-06,
+      "loss": 0.894,
+      "step": 4997
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.287280701754387e-06,
+      "loss": 0.7421,
+      "step": 4998
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.286732456140352e-06,
+      "loss": 0.96,
+      "step": 4999
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.286184210526316e-06,
+      "loss": 0.6592,
+      "step": 5000
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.285635964912281e-06,
+      "loss": 0.706,
+      "step": 5001
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.285087719298246e-06,
+      "loss": 0.6344,
+      "step": 5002
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.284539473684211e-06,
+      "loss": 0.9173,
+      "step": 5003
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.283991228070177e-06,
+      "loss": 0.8717,
+      "step": 5004
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.2834429824561415e-06,
+      "loss": 1.1648,
+      "step": 5005
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.2828947368421055e-06,
+      "loss": 1.2607,
+      "step": 5006
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.28234649122807e-06,
+      "loss": 0.3809,
+      "step": 5007
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.281798245614036e-06,
+      "loss": 0.3577,
+      "step": 5008
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.281250000000001e-06,
+      "loss": 0.8048,
+      "step": 5009
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.280701754385966e-06,
+      "loss": 0.495,
+      "step": 5010
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.28015350877193e-06,
+      "loss": 0.5928,
+      "step": 5011
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.279605263157895e-06,
+      "loss": 0.4863,
+      "step": 5012
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.27905701754386e-06,
+      "loss": 0.9666,
+      "step": 5013
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.278508771929825e-06,
+      "loss": 1.1844,
+      "step": 5014
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.277960526315791e-06,
+      "loss": 0.5212,
+      "step": 5015
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.2774122807017546e-06,
+      "loss": 1.4187,
+      "step": 5016
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.276864035087719e-06,
+      "loss": 0.4881,
+      "step": 5017
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.276315789473685e-06,
+      "loss": 0.9222,
+      "step": 5018
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.27576754385965e-06,
+      "loss": 1.3967,
+      "step": 5019
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.275219298245615e-06,
+      "loss": 0.234,
+      "step": 5020
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.27467105263158e-06,
+      "loss": 0.4083,
+      "step": 5021
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.274122807017544e-06,
+      "loss": 0.7078,
+      "step": 5022
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.273574561403509e-06,
+      "loss": 0.5603,
+      "step": 5023
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.273026315789474e-06,
+      "loss": 0.7375,
+      "step": 5024
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.27247807017544e-06,
+      "loss": 1.441,
+      "step": 5025
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.2719298245614045e-06,
+      "loss": 1.4404,
+      "step": 5026
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.2713815789473685e-06,
+      "loss": 0.6064,
+      "step": 5027
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.270833333333333e-06,
+      "loss": 0.8425,
+      "step": 5028
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.270285087719299e-06,
+      "loss": 0.9769,
+      "step": 5029
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.269736842105264e-06,
+      "loss": 1.0674,
+      "step": 5030
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.269188596491229e-06,
+      "loss": 1.1128,
+      "step": 5031
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.268640350877194e-06,
+      "loss": 1.5088,
+      "step": 5032
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.268092105263158e-06,
+      "loss": 0.5982,
+      "step": 5033
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.267543859649123e-06,
+      "loss": 1.5556,
+      "step": 5034
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.266995614035088e-06,
+      "loss": 0.8106,
+      "step": 5035
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.266447368421054e-06,
+      "loss": 0.6288,
+      "step": 5036
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.2658991228070184e-06,
+      "loss": 0.5302,
+      "step": 5037
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.265350877192982e-06,
+      "loss": 1.3587,
+      "step": 5038
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.264802631578948e-06,
+      "loss": 0.5258,
+      "step": 5039
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.264254385964913e-06,
+      "loss": 0.5947,
+      "step": 5040
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.263706140350878e-06,
+      "loss": 1.5116,
+      "step": 5041
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.263157894736843e-06,
+      "loss": 1.3613,
+      "step": 5042
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.262609649122807e-06,
+      "loss": 2.0283,
+      "step": 5043
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.262061403508772e-06,
+      "loss": 1.0172,
+      "step": 5044
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.261513157894737e-06,
+      "loss": 0.4344,
+      "step": 5045
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.260964912280703e-06,
+      "loss": 0.4715,
+      "step": 5046
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.2604166666666675e-06,
+      "loss": 0.8683,
+      "step": 5047
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.259868421052632e-06,
+      "loss": 0.3542,
+      "step": 5048
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.259320175438596e-06,
+      "loss": 0.853,
+      "step": 5049
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.258771929824562e-06,
+      "loss": 1.1131,
+      "step": 5050
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.258223684210527e-06,
+      "loss": 0.7221,
+      "step": 5051
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.257675438596492e-06,
+      "loss": 0.8281,
+      "step": 5052
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.257127192982457e-06,
+      "loss": 0.897,
+      "step": 5053
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.256578947368421e-06,
+      "loss": 0.4191,
+      "step": 5054
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.256030701754386e-06,
+      "loss": 0.545,
+      "step": 5055
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.255482456140352e-06,
+      "loss": 0.9912,
+      "step": 5056
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.254934210526317e-06,
+      "loss": 0.7825,
+      "step": 5057
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.2543859649122814e-06,
+      "loss": 0.9907,
+      "step": 5058
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.253837719298247e-06,
+      "loss": 0.3423,
+      "step": 5059
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.253289473684211e-06,
+      "loss": 1.1494,
+      "step": 5060
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.252741228070176e-06,
+      "loss": 1.0673,
+      "step": 5061
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.252192982456141e-06,
+      "loss": 1.2042,
+      "step": 5062
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.251644736842106e-06,
+      "loss": 0.8346,
+      "step": 5063
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.251096491228071e-06,
+      "loss": 0.6331,
+      "step": 5064
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.250548245614035e-06,
+      "loss": 0.6313,
+      "step": 5065
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.25e-06,
+      "loss": 1.1292,
+      "step": 5066
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.249451754385966e-06,
+      "loss": 0.7037,
+      "step": 5067
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.2489035087719305e-06,
+      "loss": 1.0482,
+      "step": 5068
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.248355263157895e-06,
+      "loss": 1.2903,
+      "step": 5069
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.247807017543859e-06,
+      "loss": 0.7851,
+      "step": 5070
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.247258771929825e-06,
+      "loss": 1.062,
+      "step": 5071
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.24671052631579e-06,
+      "loss": 0.9801,
+      "step": 5072
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.2461622807017555e-06,
+      "loss": 0.3352,
+      "step": 5073
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.24561403508772e-06,
+      "loss": 1.0273,
+      "step": 5074
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.245065789473685e-06,
+      "loss": 0.8247,
+      "step": 5075
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.244517543859649e-06,
+      "loss": 0.9389,
+      "step": 5076
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.243969298245615e-06,
+      "loss": 1.2996,
+      "step": 5077
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.24342105263158e-06,
+      "loss": 1.1172,
+      "step": 5078
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.2428728070175445e-06,
+      "loss": 0.9049,
+      "step": 5079
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.24232456140351e-06,
+      "loss": 0.8644,
+      "step": 5080
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.241776315789474e-06,
+      "loss": 0.802,
+      "step": 5081
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.241228070175439e-06,
+      "loss": 0.8874,
+      "step": 5082
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.240679824561404e-06,
+      "loss": 0.874,
+      "step": 5083
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.2401315789473694e-06,
+      "loss": 0.7834,
+      "step": 5084
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.239583333333334e-06,
+      "loss": 0.7199,
+      "step": 5085
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.239035087719298e-06,
+      "loss": 0.7638,
+      "step": 5086
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.238486842105263e-06,
+      "loss": 0.6979,
+      "step": 5087
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.237938596491229e-06,
+      "loss": 1.0021,
+      "step": 5088
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.2373903508771936e-06,
+      "loss": 0.5299,
+      "step": 5089
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.236842105263158e-06,
+      "loss": 1.234,
+      "step": 5090
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.236293859649124e-06,
+      "loss": 0.8239,
+      "step": 5091
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.235745614035088e-06,
+      "loss": 0.8882,
+      "step": 5092
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.235197368421053e-06,
+      "loss": 1.4953,
+      "step": 5093
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.2346491228070185e-06,
+      "loss": 0.9537,
+      "step": 5094
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.234100877192983e-06,
+      "loss": 0.5902,
+      "step": 5095
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.233552631578948e-06,
+      "loss": 0.7939,
+      "step": 5096
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.233004385964912e-06,
+      "loss": 0.5364,
+      "step": 5097
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.232456140350878e-06,
+      "loss": 0.5729,
+      "step": 5098
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.231907894736843e-06,
+      "loss": 1.1774,
+      "step": 5099
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.2313596491228075e-06,
+      "loss": 0.5377,
+      "step": 5100
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.230811403508773e-06,
+      "loss": 1.0727,
+      "step": 5101
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.230263157894738e-06,
+      "loss": 0.902,
+      "step": 5102
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.229714912280702e-06,
+      "loss": 0.2908,
+      "step": 5103
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.229166666666667e-06,
+      "loss": 0.9521,
+      "step": 5104
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.2286184210526324e-06,
+      "loss": 1.1441,
+      "step": 5105
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.228070175438597e-06,
+      "loss": 0.5826,
+      "step": 5106
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.227521929824562e-06,
+      "loss": 0.8721,
+      "step": 5107
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.226973684210526e-06,
+      "loss": 1.015,
+      "step": 5108
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.226425438596492e-06,
+      "loss": 0.8397,
+      "step": 5109
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.2258771929824566e-06,
+      "loss": 0.5027,
+      "step": 5110
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.225328947368422e-06,
+      "loss": 0.4722,
+      "step": 5111
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.224780701754387e-06,
+      "loss": 0.8999,
+      "step": 5112
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.224232456140351e-06,
+      "loss": 0.5104,
+      "step": 5113
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.223684210526316e-06,
+      "loss": 0.4962,
+      "step": 5114
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.2231359649122815e-06,
+      "loss": 0.9673,
+      "step": 5115
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.222587719298246e-06,
+      "loss": 1.2267,
+      "step": 5116
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.222039473684211e-06,
+      "loss": 1.0507,
+      "step": 5117
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.221491228070177e-06,
+      "loss": 0.8446,
+      "step": 5118
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.220942982456141e-06,
+      "loss": 1.2098,
+      "step": 5119
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.220394736842106e-06,
+      "loss": 0.8658,
+      "step": 5120
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.2198464912280705e-06,
+      "loss": 0.8949,
+      "step": 5121
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.219298245614036e-06,
+      "loss": 0.4519,
+      "step": 5122
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.218750000000001e-06,
+      "loss": 0.9455,
+      "step": 5123
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.218201754385965e-06,
+      "loss": 0.5999,
+      "step": 5124
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.21765350877193e-06,
+      "loss": 0.4154,
+      "step": 5125
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.2171052631578955e-06,
+      "loss": 0.981,
+      "step": 5126
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.21655701754386e-06,
+      "loss": 0.786,
+      "step": 5127
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.216008771929825e-06,
+      "loss": 0.7456,
+      "step": 5128
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.21546052631579e-06,
+      "loss": 0.6679,
+      "step": 5129
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.214912280701755e-06,
+      "loss": 0.637,
+      "step": 5130
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.2143640350877196e-06,
+      "loss": 0.7968,
+      "step": 5131
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.213815789473685e-06,
+      "loss": 0.6695,
+      "step": 5132
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.21326754385965e-06,
+      "loss": 1.1427,
+      "step": 5133
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.212719298245615e-06,
+      "loss": 0.8946,
+      "step": 5134
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.212171052631579e-06,
+      "loss": 0.9955,
+      "step": 5135
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.2116228070175445e-06,
+      "loss": 0.6723,
+      "step": 5136
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.211074561403509e-06,
+      "loss": 0.7904,
+      "step": 5137
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.210526315789474e-06,
+      "loss": 0.4857,
+      "step": 5138
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.20997807017544e-06,
+      "loss": 0.7243,
+      "step": 5139
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.209429824561404e-06,
+      "loss": 0.6389,
+      "step": 5140
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.208881578947369e-06,
+      "loss": 1.1985,
+      "step": 5141
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.2083333333333335e-06,
+      "loss": 0.5095,
+      "step": 5142
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.207785087719299e-06,
+      "loss": 0.7716,
+      "step": 5143
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.207236842105264e-06,
+      "loss": 0.5929,
+      "step": 5144
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.206688596491229e-06,
+      "loss": 0.6918,
+      "step": 5145
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.206140350877193e-06,
+      "loss": 1.0716,
+      "step": 5146
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.2055921052631585e-06,
+      "loss": 0.8015,
+      "step": 5147
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.205043859649123e-06,
+      "loss": 1.4538,
+      "step": 5148
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.204495614035089e-06,
+      "loss": 0.8768,
+      "step": 5149
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.203947368421054e-06,
+      "loss": 0.8563,
+      "step": 5150
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.203399122807018e-06,
+      "loss": 0.2488,
+      "step": 5151
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.202850877192983e-06,
+      "loss": 0.8814,
+      "step": 5152
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.202302631578948e-06,
+      "loss": 1.3881,
+      "step": 5153
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.201754385964913e-06,
+      "loss": 1.3174,
+      "step": 5154
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.201206140350878e-06,
+      "loss": 0.9262,
+      "step": 5155
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.200657894736842e-06,
+      "loss": 0.7747,
+      "step": 5156
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.2001096491228076e-06,
+      "loss": 0.5343,
+      "step": 5157
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.199561403508772e-06,
+      "loss": 0.2579,
+      "step": 5158
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.199013157894737e-06,
+      "loss": 0.6948,
+      "step": 5159
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.198464912280703e-06,
+      "loss": 1.1257,
+      "step": 5160
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.197916666666668e-06,
+      "loss": 0.5472,
+      "step": 5161
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.197368421052632e-06,
+      "loss": 1.0227,
+      "step": 5162
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.1968201754385965e-06,
+      "loss": 1.2837,
+      "step": 5163
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.196271929824562e-06,
+      "loss": 0.8805,
+      "step": 5164
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.195723684210527e-06,
+      "loss": 0.9131,
+      "step": 5165
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.195175438596493e-06,
+      "loss": 0.8357,
+      "step": 5166
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.194627192982457e-06,
+      "loss": 0.5035,
+      "step": 5167
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.1940789473684215e-06,
+      "loss": 1.3434,
+      "step": 5168
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.193530701754386e-06,
+      "loss": 0.657,
+      "step": 5169
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.192982456140352e-06,
+      "loss": 1.3202,
+      "step": 5170
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.192434210526317e-06,
+      "loss": 1.2901,
+      "step": 5171
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.191885964912281e-06,
+      "loss": 0.9023,
+      "step": 5172
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.191337719298246e-06,
+      "loss": 0.5118,
+      "step": 5173
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.190789473684211e-06,
+      "loss": 0.4495,
+      "step": 5174
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.190241228070176e-06,
+      "loss": 0.8224,
+      "step": 5175
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.189692982456141e-06,
+      "loss": 1.1608,
+      "step": 5176
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.189144736842107e-06,
+      "loss": 0.4401,
+      "step": 5177
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.1885964912280706e-06,
+      "loss": 0.8515,
+      "step": 5178
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.188048245614035e-06,
+      "loss": 1.2012,
+      "step": 5179
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.1875e-06,
+      "loss": 0.9486,
+      "step": 5180
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.186951754385966e-06,
+      "loss": 0.7654,
+      "step": 5181
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.186403508771931e-06,
+      "loss": 1.1468,
+      "step": 5182
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.185855263157895e-06,
+      "loss": 0.6331,
+      "step": 5183
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.18530701754386e-06,
+      "loss": 0.9256,
+      "step": 5184
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.184758771929825e-06,
+      "loss": 0.7648,
+      "step": 5185
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.18421052631579e-06,
+      "loss": 1.0439,
+      "step": 5186
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.183662280701756e-06,
+      "loss": 0.6377,
+      "step": 5187
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.1831140350877205e-06,
+      "loss": 0.9789,
+      "step": 5188
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.1825657894736845e-06,
+      "loss": 1.3778,
+      "step": 5189
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.182017543859649e-06,
+      "loss": 0.8459,
+      "step": 5190
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.181469298245615e-06,
+      "loss": 0.6797,
+      "step": 5191
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.18092105263158e-06,
+      "loss": 0.7567,
+      "step": 5192
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.180372807017545e-06,
+      "loss": 1.0765,
+      "step": 5193
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.179824561403509e-06,
+      "loss": 0.6079,
+      "step": 5194
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.179276315789474e-06,
+      "loss": 1.6805,
+      "step": 5195
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.178728070175439e-06,
+      "loss": 1.0566,
+      "step": 5196
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.178179824561404e-06,
+      "loss": 0.7367,
+      "step": 5197
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.17763157894737e-06,
+      "loss": 1.1461,
+      "step": 5198
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.177083333333334e-06,
+      "loss": 1.2423,
+      "step": 5199
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.176535087719298e-06,
+      "loss": 0.4264,
+      "step": 5200
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.175986842105263e-06,
+      "loss": 0.6157,
+      "step": 5201
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.175438596491229e-06,
+      "loss": 1.19,
+      "step": 5202
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.174890350877194e-06,
+      "loss": 0.4582,
+      "step": 5203
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.174342105263159e-06,
+      "loss": 0.7561,
+      "step": 5204
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.173793859649123e-06,
+      "loss": 1.0121,
+      "step": 5205
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.173245614035088e-06,
+      "loss": 0.5717,
+      "step": 5206
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.172697368421053e-06,
+      "loss": 0.5762,
+      "step": 5207
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.172149122807019e-06,
+      "loss": 0.867,
+      "step": 5208
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.1716008771929835e-06,
+      "loss": 0.4378,
+      "step": 5209
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.1710526315789475e-06,
+      "loss": 1.0011,
+      "step": 5210
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.170504385964912e-06,
+      "loss": 0.4744,
+      "step": 5211
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.169956140350878e-06,
+      "loss": 0.7304,
+      "step": 5212
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.169407894736843e-06,
+      "loss": 1.0163,
+      "step": 5213
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.168859649122808e-06,
+      "loss": 0.7626,
+      "step": 5214
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.168311403508772e-06,
+      "loss": 0.4989,
+      "step": 5215
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.167763157894737e-06,
+      "loss": 0.7903,
+      "step": 5216
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.167214912280702e-06,
+      "loss": 0.593,
+      "step": 5217
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.166666666666667e-06,
+      "loss": 0.5281,
+      "step": 5218
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.166118421052633e-06,
+      "loss": 1.3535,
+      "step": 5219
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.1655701754385974e-06,
+      "loss": 0.7379,
+      "step": 5220
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.165021929824561e-06,
+      "loss": 0.6029,
+      "step": 5221
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.164473684210527e-06,
+      "loss": 0.7028,
+      "step": 5222
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.163925438596492e-06,
+      "loss": 0.6958,
+      "step": 5223
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.163377192982457e-06,
+      "loss": 1.3336,
+      "step": 5224
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.162828947368422e-06,
+      "loss": 0.5333,
+      "step": 5225
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.162280701754386e-06,
+      "loss": 0.5204,
+      "step": 5226
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.161732456140351e-06,
+      "loss": 0.5148,
+      "step": 5227
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.161184210526316e-06,
+      "loss": 1.1247,
+      "step": 5228
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.160635964912282e-06,
+      "loss": 0.6151,
+      "step": 5229
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.1600877192982465e-06,
+      "loss": 0.6988,
+      "step": 5230
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.159539473684211e-06,
+      "loss": 1.4532,
+      "step": 5231
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.158991228070175e-06,
+      "loss": 1.154,
+      "step": 5232
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.158442982456141e-06,
+      "loss": 0.8832,
+      "step": 5233
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.157894736842106e-06,
+      "loss": 0.6271,
+      "step": 5234
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.157346491228071e-06,
+      "loss": 0.5339,
+      "step": 5235
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.156798245614036e-06,
+      "loss": 0.9882,
+      "step": 5236
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.15625e-06,
+      "loss": 0.6989,
+      "step": 5237
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.155701754385965e-06,
+      "loss": 0.6427,
+      "step": 5238
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.155153508771931e-06,
+      "loss": 1.3127,
+      "step": 5239
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.154605263157896e-06,
+      "loss": 0.3742,
+      "step": 5240
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.1540570175438605e-06,
+      "loss": 1.0,
+      "step": 5241
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.1535087719298244e-06,
+      "loss": 0.7793,
+      "step": 5242
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.15296052631579e-06,
+      "loss": 0.8748,
+      "step": 5243
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.152412280701755e-06,
+      "loss": 0.9228,
+      "step": 5244
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.15186403508772e-06,
+      "loss": 0.6386,
+      "step": 5245
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.151315789473685e-06,
+      "loss": 0.5873,
+      "step": 5246
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.15076754385965e-06,
+      "loss": 1.4223,
+      "step": 5247
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.150219298245614e-06,
+      "loss": 0.8368,
+      "step": 5248
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.149671052631579e-06,
+      "loss": 0.6419,
+      "step": 5249
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.149122807017545e-06,
+      "loss": 1.2051,
+      "step": 5250
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.1485745614035095e-06,
+      "loss": 0.944,
+      "step": 5251
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.148026315789474e-06,
+      "loss": 0.4939,
+      "step": 5252
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.147478070175438e-06,
+      "loss": 0.5682,
+      "step": 5253
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.146929824561404e-06,
+      "loss": 0.8137,
+      "step": 5254
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.146381578947369e-06,
+      "loss": 0.7314,
+      "step": 5255
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.145833333333334e-06,
+      "loss": 0.7863,
+      "step": 5256
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.145285087719299e-06,
+      "loss": 1.0988,
+      "step": 5257
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.144736842105263e-06,
+      "loss": 1.9103,
+      "step": 5258
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.144188596491228e-06,
+      "loss": 0.9003,
+      "step": 5259
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.143640350877194e-06,
+      "loss": 1.1758,
+      "step": 5260
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.143092105263159e-06,
+      "loss": 0.7526,
+      "step": 5261
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.1425438596491235e-06,
+      "loss": 0.8722,
+      "step": 5262
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.141995614035089e-06,
+      "loss": 0.469,
+      "step": 5263
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.141447368421053e-06,
+      "loss": 1.0863,
+      "step": 5264
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.140899122807018e-06,
+      "loss": 1.0796,
+      "step": 5265
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.140350877192983e-06,
+      "loss": 0.7472,
+      "step": 5266
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.1398026315789484e-06,
+      "loss": 1.2466,
+      "step": 5267
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.139254385964913e-06,
+      "loss": 1.5324,
+      "step": 5268
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.138706140350877e-06,
+      "loss": 0.5208,
+      "step": 5269
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.138157894736842e-06,
+      "loss": 0.3155,
+      "step": 5270
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.137609649122808e-06,
+      "loss": 0.9172,
+      "step": 5271
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.1370614035087726e-06,
+      "loss": 1.0274,
+      "step": 5272
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.136513157894737e-06,
+      "loss": 0.5154,
+      "step": 5273
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.135964912280703e-06,
+      "loss": 0.6949,
+      "step": 5274
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.135416666666667e-06,
+      "loss": 1.3227,
+      "step": 5275
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.134868421052632e-06,
+      "loss": 0.8146,
+      "step": 5276
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.1343201754385975e-06,
+      "loss": 1.2348,
+      "step": 5277
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.133771929824562e-06,
+      "loss": 0.8611,
+      "step": 5278
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.133223684210527e-06,
+      "loss": 0.8608,
+      "step": 5279
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.132675438596491e-06,
+      "loss": 0.4431,
+      "step": 5280
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.132127192982457e-06,
+      "loss": 0.5427,
+      "step": 5281
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.131578947368422e-06,
+      "loss": 1.2477,
+      "step": 5282
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.1310307017543865e-06,
+      "loss": 0.709,
+      "step": 5283
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.130482456140352e-06,
+      "loss": 0.6933,
+      "step": 5284
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.129934210526316e-06,
+      "loss": 0.8798,
+      "step": 5285
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.129385964912281e-06,
+      "loss": 1.1497,
+      "step": 5286
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.128837719298246e-06,
+      "loss": 0.4408,
+      "step": 5287
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.1282894736842114e-06,
+      "loss": 1.2662,
+      "step": 5288
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.127741228070176e-06,
+      "loss": 0.4452,
+      "step": 5289
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.127192982456141e-06,
+      "loss": 1.5275,
+      "step": 5290
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.126644736842105e-06,
+      "loss": 0.8495,
+      "step": 5291
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.126096491228071e-06,
+      "loss": 0.9471,
+      "step": 5292
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.1255482456140356e-06,
+      "loss": 1.2904,
+      "step": 5293
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.125e-06,
+      "loss": 1.2895,
+      "step": 5294
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.124451754385966e-06,
+      "loss": 0.2581,
+      "step": 5295
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.12390350877193e-06,
+      "loss": 0.6143,
+      "step": 5296
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.123355263157895e-06,
+      "loss": 1.0106,
+      "step": 5297
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.1228070175438605e-06,
+      "loss": 1.1334,
+      "step": 5298
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.122258771929825e-06,
+      "loss": 1.4275,
+      "step": 5299
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.12171052631579e-06,
+      "loss": 0.5379,
+      "step": 5300
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.121162280701754e-06,
+      "loss": 1.0116,
+      "step": 5301
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.12061403508772e-06,
+      "loss": 0.9077,
+      "step": 5302
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.120065789473685e-06,
+      "loss": 0.7751,
+      "step": 5303
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.1195175438596495e-06,
+      "loss": 1.1317,
+      "step": 5304
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.118969298245615e-06,
+      "loss": 1.7426,
+      "step": 5305
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.11842105263158e-06,
+      "loss": 0.9339,
+      "step": 5306
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.117872807017544e-06,
+      "loss": 0.9843,
+      "step": 5307
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.117324561403509e-06,
+      "loss": 0.6966,
+      "step": 5308
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.1167763157894745e-06,
+      "loss": 0.9339,
+      "step": 5309
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.116228070175439e-06,
+      "loss": 0.6725,
+      "step": 5310
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.115679824561404e-06,
+      "loss": 0.3553,
+      "step": 5311
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.115131578947368e-06,
+      "loss": 0.3942,
+      "step": 5312
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.114583333333334e-06,
+      "loss": 1.311,
+      "step": 5313
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.114035087719299e-06,
+      "loss": 0.7138,
+      "step": 5314
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.113486842105264e-06,
+      "loss": 0.7771,
+      "step": 5315
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.112938596491229e-06,
+      "loss": 0.7901,
+      "step": 5316
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.112390350877194e-06,
+      "loss": 0.2656,
+      "step": 5317
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.111842105263158e-06,
+      "loss": 0.7127,
+      "step": 5318
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.1112938596491235e-06,
+      "loss": 0.3993,
+      "step": 5319
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.110745614035088e-06,
+      "loss": 1.3598,
+      "step": 5320
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.110197368421053e-06,
+      "loss": 1.0949,
+      "step": 5321
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.109649122807019e-06,
+      "loss": 0.8869,
+      "step": 5322
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.109100877192983e-06,
+      "loss": 0.8776,
+      "step": 5323
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.108552631578948e-06,
+      "loss": 1.1161,
+      "step": 5324
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.1080043859649125e-06,
+      "loss": 0.571,
+      "step": 5325
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.107456140350878e-06,
+      "loss": 1.2455,
+      "step": 5326
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.106907894736843e-06,
+      "loss": 0.6938,
+      "step": 5327
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.106359649122807e-06,
+      "loss": 0.6493,
+      "step": 5328
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.105811403508772e-06,
+      "loss": 0.8153,
+      "step": 5329
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.1052631578947375e-06,
+      "loss": 0.5657,
+      "step": 5330
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.104714912280702e-06,
+      "loss": 0.778,
+      "step": 5331
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.104166666666668e-06,
+      "loss": 0.6807,
+      "step": 5332
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.103618421052633e-06,
+      "loss": 0.8986,
+      "step": 5333
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.103070175438597e-06,
+      "loss": 0.784,
+      "step": 5334
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.102521929824562e-06,
+      "loss": 1.0172,
+      "step": 5335
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.101973684210527e-06,
+      "loss": 1.7271,
+      "step": 5336
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.101425438596492e-06,
+      "loss": 0.6499,
+      "step": 5337
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.100877192982457e-06,
+      "loss": 0.6764,
+      "step": 5338
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.100328947368421e-06,
+      "loss": 1.0244,
+      "step": 5339
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.0997807017543866e-06,
+      "loss": 1.2326,
+      "step": 5340
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.099232456140351e-06,
+      "loss": 0.8094,
+      "step": 5341
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.098684210526316e-06,
+      "loss": 0.4189,
+      "step": 5342
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.098135964912282e-06,
+      "loss": 1.0272,
+      "step": 5343
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.097587719298247e-06,
+      "loss": 1.364,
+      "step": 5344
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.097039473684211e-06,
+      "loss": 1.7154,
+      "step": 5345
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.0964912280701755e-06,
+      "loss": 0.7662,
+      "step": 5346
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.095942982456141e-06,
+      "loss": 0.3808,
+      "step": 5347
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.095394736842106e-06,
+      "loss": 1.1874,
+      "step": 5348
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.094846491228071e-06,
+      "loss": 0.6671,
+      "step": 5349
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.094298245614036e-06,
+      "loss": 0.3782,
+      "step": 5350
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.0937500000000005e-06,
+      "loss": 0.6705,
+      "step": 5351
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.093201754385965e-06,
+      "loss": 0.7295,
+      "step": 5352
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.092653508771931e-06,
+      "loss": 0.2881,
+      "step": 5353
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.092105263157896e-06,
+      "loss": 1.5007,
+      "step": 5354
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.09155701754386e-06,
+      "loss": 1.6142,
+      "step": 5355
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.091008771929825e-06,
+      "loss": 1.2406,
+      "step": 5356
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.09046052631579e-06,
+      "loss": 0.4803,
+      "step": 5357
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.089912280701755e-06,
+      "loss": 0.7737,
+      "step": 5358
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.08936403508772e-06,
+      "loss": 1.5304,
+      "step": 5359
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.088815789473686e-06,
+      "loss": 0.5906,
+      "step": 5360
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.0882675438596496e-06,
+      "loss": 0.918,
+      "step": 5361
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.087719298245614e-06,
+      "loss": 0.2696,
+      "step": 5362
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.087171052631579e-06,
+      "loss": 1.3642,
+      "step": 5363
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.086622807017545e-06,
+      "loss": 1.1192,
+      "step": 5364
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.08607456140351e-06,
+      "loss": 0.977,
+      "step": 5365
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.085526315789474e-06,
+      "loss": 0.7788,
+      "step": 5366
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.0849780701754385e-06,
+      "loss": 0.7027,
+      "step": 5367
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.084429824561404e-06,
+      "loss": 1.3019,
+      "step": 5368
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.083881578947369e-06,
+      "loss": 0.5467,
+      "step": 5369
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.083333333333335e-06,
+      "loss": 0.9803,
+      "step": 5370
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.082785087719299e-06,
+      "loss": 0.7612,
+      "step": 5371
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.0822368421052635e-06,
+      "loss": 0.6684,
+      "step": 5372
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.081688596491228e-06,
+      "loss": 0.7494,
+      "step": 5373
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.081140350877194e-06,
+      "loss": 0.6962,
+      "step": 5374
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.080592105263159e-06,
+      "loss": 1.5699,
+      "step": 5375
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.080043859649124e-06,
+      "loss": 0.8156,
+      "step": 5376
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.079495614035088e-06,
+      "loss": 0.4199,
+      "step": 5377
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.078947368421053e-06,
+      "loss": 0.4433,
+      "step": 5378
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.078399122807018e-06,
+      "loss": 1.2408,
+      "step": 5379
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.077850877192983e-06,
+      "loss": 1.0079,
+      "step": 5380
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.077302631578949e-06,
+      "loss": 0.6322,
+      "step": 5381
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.076754385964913e-06,
+      "loss": 1.622,
+      "step": 5382
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.076206140350877e-06,
+      "loss": 1.2704,
+      "step": 5383
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.075657894736842e-06,
+      "loss": 0.3853,
+      "step": 5384
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.075109649122808e-06,
+      "loss": 1.0571,
+      "step": 5385
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.074561403508773e-06,
+      "loss": 0.7822,
+      "step": 5386
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.0740131578947376e-06,
+      "loss": 0.7532,
+      "step": 5387
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.073464912280702e-06,
+      "loss": 0.9742,
+      "step": 5388
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.072916666666667e-06,
+      "loss": 1.1374,
+      "step": 5389
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.072368421052632e-06,
+      "loss": 1.116,
+      "step": 5390
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.071820175438598e-06,
+      "loss": 0.7512,
+      "step": 5391
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.0712719298245625e-06,
+      "loss": 0.8978,
+      "step": 5392
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.0707236842105265e-06,
+      "loss": 0.8118,
+      "step": 5393
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.070175438596491e-06,
+      "loss": 0.9839,
+      "step": 5394
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.069627192982457e-06,
+      "loss": 0.5338,
+      "step": 5395
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.069078947368422e-06,
+      "loss": 1.0632,
+      "step": 5396
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.068530701754387e-06,
+      "loss": 0.825,
+      "step": 5397
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.067982456140351e-06,
+      "loss": 0.6335,
+      "step": 5398
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.067434210526316e-06,
+      "loss": 0.8634,
+      "step": 5399
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.066885964912281e-06,
+      "loss": 0.6853,
+      "step": 5400
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.066337719298246e-06,
+      "loss": 0.6169,
+      "step": 5401
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.065789473684212e-06,
+      "loss": 0.6997,
+      "step": 5402
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.0652412280701764e-06,
+      "loss": 0.7391,
+      "step": 5403
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.06469298245614e-06,
+      "loss": 1.4546,
+      "step": 5404
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.064144736842105e-06,
+      "loss": 0.7256,
+      "step": 5405
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.063596491228071e-06,
+      "loss": 0.1819,
+      "step": 5406
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.063048245614036e-06,
+      "loss": 0.5073,
+      "step": 5407
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.062500000000001e-06,
+      "loss": 0.9138,
+      "step": 5408
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.061951754385965e-06,
+      "loss": 0.6097,
+      "step": 5409
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.06140350877193e-06,
+      "loss": 0.4699,
+      "step": 5410
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.060855263157895e-06,
+      "loss": 0.5376,
+      "step": 5411
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.060307017543861e-06,
+      "loss": 1.5703,
+      "step": 5412
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.0597587719298255e-06,
+      "loss": 0.7543,
+      "step": 5413
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.0592105263157895e-06,
+      "loss": 1.0022,
+      "step": 5414
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.058662280701754e-06,
+      "loss": 0.9276,
+      "step": 5415
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.05811403508772e-06,
+      "loss": 1.2012,
+      "step": 5416
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.057565789473685e-06,
+      "loss": 0.9315,
+      "step": 5417
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.05701754385965e-06,
+      "loss": 0.6094,
+      "step": 5418
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.056469298245615e-06,
+      "loss": 0.5564,
+      "step": 5419
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.055921052631579e-06,
+      "loss": 0.5275,
+      "step": 5420
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.055372807017544e-06,
+      "loss": 1.0384,
+      "step": 5421
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.054824561403509e-06,
+      "loss": 0.5823,
+      "step": 5422
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.054276315789475e-06,
+      "loss": 0.8089,
+      "step": 5423
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.0537280701754395e-06,
+      "loss": 1.2051,
+      "step": 5424
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.0531798245614034e-06,
+      "loss": 0.9888,
+      "step": 5425
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.052631578947369e-06,
+      "loss": 0.6759,
+      "step": 5426
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.052083333333334e-06,
+      "loss": 1.0997,
+      "step": 5427
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.051535087719299e-06,
+      "loss": 1.0043,
+      "step": 5428
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.0509868421052644e-06,
+      "loss": 0.5824,
+      "step": 5429
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.050438596491229e-06,
+      "loss": 1.73,
+      "step": 5430
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.049890350877193e-06,
+      "loss": 1.088,
+      "step": 5431
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.049342105263158e-06,
+      "loss": 1.0172,
+      "step": 5432
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.048793859649124e-06,
+      "loss": 0.9069,
+      "step": 5433
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.0482456140350885e-06,
+      "loss": 0.5897,
+      "step": 5434
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.047697368421053e-06,
+      "loss": 0.892,
+      "step": 5435
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.047149122807017e-06,
+      "loss": 0.6317,
+      "step": 5436
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.046600877192983e-06,
+      "loss": 1.0308,
+      "step": 5437
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.046052631578948e-06,
+      "loss": 1.0761,
+      "step": 5438
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.045504385964913e-06,
+      "loss": 0.7333,
+      "step": 5439
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.044956140350878e-06,
+      "loss": 0.8377,
+      "step": 5440
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.044407894736842e-06,
+      "loss": 0.4258,
+      "step": 5441
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.043859649122807e-06,
+      "loss": 0.9353,
+      "step": 5442
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.043311403508773e-06,
+      "loss": 0.94,
+      "step": 5443
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.042763157894738e-06,
+      "loss": 1.5407,
+      "step": 5444
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.0422149122807025e-06,
+      "loss": 0.6382,
+      "step": 5445
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.041666666666668e-06,
+      "loss": 0.6171,
+      "step": 5446
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.041118421052632e-06,
+      "loss": 0.4883,
+      "step": 5447
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.040570175438597e-06,
+      "loss": 0.5871,
+      "step": 5448
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.040021929824562e-06,
+      "loss": 0.4682,
+      "step": 5449
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.0394736842105274e-06,
+      "loss": 0.984,
+      "step": 5450
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.038925438596492e-06,
+      "loss": 0.4922,
+      "step": 5451
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.038377192982456e-06,
+      "loss": 0.4093,
+      "step": 5452
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.037828947368421e-06,
+      "loss": 1.0034,
+      "step": 5453
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.037280701754387e-06,
+      "loss": 0.9704,
+      "step": 5454
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.0367324561403516e-06,
+      "loss": 0.919,
+      "step": 5455
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.036184210526316e-06,
+      "loss": 0.9914,
+      "step": 5456
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.03563596491228e-06,
+      "loss": 0.6448,
+      "step": 5457
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.035087719298246e-06,
+      "loss": 0.3934,
+      "step": 5458
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.034539473684211e-06,
+      "loss": 0.9653,
+      "step": 5459
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.033991228070176e-06,
+      "loss": 0.922,
+      "step": 5460
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.033442982456141e-06,
+      "loss": 1.3258,
+      "step": 5461
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.032894736842106e-06,
+      "loss": 1.028,
+      "step": 5462
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.03234649122807e-06,
+      "loss": 0.8629,
+      "step": 5463
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.031798245614036e-06,
+      "loss": 0.8348,
+      "step": 5464
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.031250000000001e-06,
+      "loss": 1.0872,
+      "step": 5465
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.0307017543859655e-06,
+      "loss": 1.0288,
+      "step": 5466
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.030153508771931e-06,
+      "loss": 1.5243,
+      "step": 5467
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.029605263157895e-06,
+      "loss": 0.7204,
+      "step": 5468
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.02905701754386e-06,
+      "loss": 0.9043,
+      "step": 5469
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.028508771929825e-06,
+      "loss": 0.4744,
+      "step": 5470
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.0279605263157904e-06,
+      "loss": 1.1137,
+      "step": 5471
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.027412280701755e-06,
+      "loss": 0.9253,
+      "step": 5472
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.02686403508772e-06,
+      "loss": 1.1918,
+      "step": 5473
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.026315789473684e-06,
+      "loss": 0.7233,
+      "step": 5474
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.02576754385965e-06,
+      "loss": 0.5401,
+      "step": 5475
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.0252192982456146e-06,
+      "loss": 1.3533,
+      "step": 5476
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.024671052631579e-06,
+      "loss": 1.3559,
+      "step": 5477
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.024122807017545e-06,
+      "loss": 0.7185,
+      "step": 5478
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.023574561403509e-06,
+      "loss": 0.8444,
+      "step": 5479
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.023026315789474e-06,
+      "loss": 1.1125,
+      "step": 5480
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.0224780701754395e-06,
+      "loss": 1.6547,
+      "step": 5481
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.021929824561404e-06,
+      "loss": 0.868,
+      "step": 5482
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.021381578947369e-06,
+      "loss": 1.295,
+      "step": 5483
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.020833333333333e-06,
+      "loss": 0.9191,
+      "step": 5484
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.020285087719299e-06,
+      "loss": 0.7127,
+      "step": 5485
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.019736842105264e-06,
+      "loss": 0.5601,
+      "step": 5486
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.0191885964912285e-06,
+      "loss": 2.1834,
+      "step": 5487
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.6715193766659832,
+      "eval_loss": 0.8014807105064392,
+      "eval_roc_auc": 0.9092854873689875,
+      "eval_runtime": 181.2274,
+      "eval_samples_per_second": 26.911,
+      "eval_steps_per_second": 3.366,
+      "step": 5487
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.018640350877194e-06,
+      "loss": 0.5459,
+      "step": 5488
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.018092105263159e-06,
+      "loss": 0.742,
+      "step": 5489
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.017543859649123e-06,
+      "loss": 1.2274,
+      "step": 5490
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.016995614035088e-06,
+      "loss": 1.1241,
+      "step": 5491
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.0164473684210535e-06,
+      "loss": 0.731,
+      "step": 5492
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.015899122807018e-06,
+      "loss": 0.3796,
+      "step": 5493
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.015350877192983e-06,
+      "loss": 0.8285,
+      "step": 5494
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.014802631578947e-06,
+      "loss": 0.4094,
+      "step": 5495
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.014254385964913e-06,
+      "loss": 1.146,
+      "step": 5496
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 7.013706140350878e-06,
+      "loss": 1.2944,
+      "step": 5497
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 7.013157894736842e-06,
+      "loss": 1.003,
+      "step": 5498
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 7.012609649122808e-06,
+      "loss": 0.7502,
+      "step": 5499
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 7.012061403508772e-06,
+      "loss": 0.4634,
+      "step": 5500
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 7.011513157894737e-06,
+      "loss": 1.3168,
+      "step": 5501
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 7.0109649122807026e-06,
+      "loss": 0.4785,
+      "step": 5502
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 7.010416666666667e-06,
+      "loss": 0.5067,
+      "step": 5503
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 7.009868421052632e-06,
+      "loss": 1.0109,
+      "step": 5504
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 7.009320175438598e-06,
+      "loss": 1.1232,
+      "step": 5505
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 7.008771929824562e-06,
+      "loss": 1.2151,
+      "step": 5506
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 7.008223684210527e-06,
+      "loss": 1.0871,
+      "step": 5507
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 7.0076754385964915e-06,
+      "loss": 1.0138,
+      "step": 5508
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 7.007127192982457e-06,
+      "loss": 0.8839,
+      "step": 5509
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 7.006578947368422e-06,
+      "loss": 0.3311,
+      "step": 5510
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 7.006030701754386e-06,
+      "loss": 0.651,
+      "step": 5511
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 7.005482456140351e-06,
+      "loss": 0.6244,
+      "step": 5512
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 7.0049342105263165e-06,
+      "loss": 0.5926,
+      "step": 5513
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 7.004385964912281e-06,
+      "loss": 1.0652,
+      "step": 5514
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 7.003837719298246e-06,
+      "loss": 0.5671,
+      "step": 5515
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 7.003289473684212e-06,
+      "loss": 0.6877,
+      "step": 5516
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 7.002741228070176e-06,
+      "loss": 1.6047,
+      "step": 5517
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 7.002192982456141e-06,
+      "loss": 0.8664,
+      "step": 5518
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 7.001644736842106e-06,
+      "loss": 0.6788,
+      "step": 5519
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 7.001096491228071e-06,
+      "loss": 0.6384,
+      "step": 5520
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 7.000548245614036e-06,
+      "loss": 0.7635,
+      "step": 5521
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 7e-06,
+      "loss": 0.9824,
+      "step": 5522
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 6.9994517543859656e-06,
+      "loss": 1.1602,
+      "step": 5523
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 6.99890350877193e-06,
+      "loss": 1.2765,
+      "step": 5524
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 6.998355263157895e-06,
+      "loss": 1.435,
+      "step": 5525
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 6.997807017543861e-06,
+      "loss": 0.7693,
+      "step": 5526
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 6.997258771929825e-06,
+      "loss": 1.0743,
+      "step": 5527
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 6.99671052631579e-06,
+      "loss": 0.5404,
+      "step": 5528
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 6.9961622807017545e-06,
+      "loss": 1.2679,
+      "step": 5529
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 6.99561403508772e-06,
+      "loss": 0.2997,
+      "step": 5530
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 6.995065789473685e-06,
+      "loss": 1.8415,
+      "step": 5531
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 6.99451754385965e-06,
+      "loss": 0.6466,
+      "step": 5532
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.993969298245614e-06,
+      "loss": 1.1735,
+      "step": 5533
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.9934210526315795e-06,
+      "loss": 0.7844,
+      "step": 5534
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.992872807017544e-06,
+      "loss": 1.0333,
+      "step": 5535
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.99232456140351e-06,
+      "loss": 1.532,
+      "step": 5536
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.991776315789475e-06,
+      "loss": 0.8992,
+      "step": 5537
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.991228070175439e-06,
+      "loss": 0.4408,
+      "step": 5538
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.990679824561404e-06,
+      "loss": 0.9106,
+      "step": 5539
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.990131578947369e-06,
+      "loss": 0.533,
+      "step": 5540
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.989583333333334e-06,
+      "loss": 1.268,
+      "step": 5541
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.989035087719299e-06,
+      "loss": 0.7833,
+      "step": 5542
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.988486842105263e-06,
+      "loss": 0.4894,
+      "step": 5543
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.987938596491229e-06,
+      "loss": 0.5422,
+      "step": 5544
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.987390350877193e-06,
+      "loss": 0.7777,
+      "step": 5545
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.986842105263158e-06,
+      "loss": 0.5797,
+      "step": 5546
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.986293859649124e-06,
+      "loss": 0.29,
+      "step": 5547
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.985745614035089e-06,
+      "loss": 1.7897,
+      "step": 5548
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.985197368421053e-06,
+      "loss": 0.4569,
+      "step": 5549
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.9846491228070175e-06,
+      "loss": 1.0694,
+      "step": 5550
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.984100877192983e-06,
+      "loss": 0.8561,
+      "step": 5551
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.983552631578948e-06,
+      "loss": 0.4239,
+      "step": 5552
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.983004385964913e-06,
+      "loss": 0.5462,
+      "step": 5553
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.982456140350878e-06,
+      "loss": 0.3059,
+      "step": 5554
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.9819078947368425e-06,
+      "loss": 0.4714,
+      "step": 5555
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.981359649122807e-06,
+      "loss": 0.4796,
+      "step": 5556
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.980811403508773e-06,
+      "loss": 1.3329,
+      "step": 5557
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.980263157894738e-06,
+      "loss": 0.401,
+      "step": 5558
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.979714912280703e-06,
+      "loss": 1.1483,
+      "step": 5559
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.979166666666667e-06,
+      "loss": 0.3849,
+      "step": 5560
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.978618421052632e-06,
+      "loss": 0.6213,
+      "step": 5561
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.978070175438597e-06,
+      "loss": 0.7618,
+      "step": 5562
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.977521929824562e-06,
+      "loss": 0.668,
+      "step": 5563
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.976973684210528e-06,
+      "loss": 0.375,
+      "step": 5564
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.976425438596492e-06,
+      "loss": 0.8936,
+      "step": 5565
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.975877192982456e-06,
+      "loss": 0.7099,
+      "step": 5566
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.975328947368421e-06,
+      "loss": 0.8302,
+      "step": 5567
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.974780701754387e-06,
+      "loss": 0.7347,
+      "step": 5568
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.974232456140352e-06,
+      "loss": 0.9411,
+      "step": 5569
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.973684210526316e-06,
+      "loss": 0.9544,
+      "step": 5570
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.9731359649122805e-06,
+      "loss": 0.8685,
+      "step": 5571
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.972587719298246e-06,
+      "loss": 0.5053,
+      "step": 5572
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.972039473684211e-06,
+      "loss": 0.9657,
+      "step": 5573
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.971491228070177e-06,
+      "loss": 0.6022,
+      "step": 5574
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.9709429824561415e-06,
+      "loss": 0.3245,
+      "step": 5575
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.9703947368421055e-06,
+      "loss": 1.0471,
+      "step": 5576
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.96984649122807e-06,
+      "loss": 0.5627,
+      "step": 5577
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.969298245614036e-06,
+      "loss": 0.6448,
+      "step": 5578
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.968750000000001e-06,
+      "loss": 0.8245,
+      "step": 5579
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.968201754385966e-06,
+      "loss": 1.2883,
+      "step": 5580
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.96765350877193e-06,
+      "loss": 0.8936,
+      "step": 5581
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.967105263157895e-06,
+      "loss": 0.2622,
+      "step": 5582
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.96655701754386e-06,
+      "loss": 0.8391,
+      "step": 5583
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.966008771929825e-06,
+      "loss": 0.4841,
+      "step": 5584
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.965460526315791e-06,
+      "loss": 0.596,
+      "step": 5585
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.964912280701755e-06,
+      "loss": 0.3967,
+      "step": 5586
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.9643640350877194e-06,
+      "loss": 0.408,
+      "step": 5587
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.963815789473684e-06,
+      "loss": 1.1317,
+      "step": 5588
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.96326754385965e-06,
+      "loss": 1.171,
+      "step": 5589
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.962719298245615e-06,
+      "loss": 0.5723,
+      "step": 5590
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.96217105263158e-06,
+      "loss": 0.5756,
+      "step": 5591
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.961622807017544e-06,
+      "loss": 0.4266,
+      "step": 5592
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.961074561403509e-06,
+      "loss": 1.076,
+      "step": 5593
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.960526315789474e-06,
+      "loss": 0.6026,
+      "step": 5594
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.95997807017544e-06,
+      "loss": 0.228,
+      "step": 5595
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.9594298245614045e-06,
+      "loss": 0.4317,
+      "step": 5596
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.9588815789473685e-06,
+      "loss": 1.5221,
+      "step": 5597
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.958333333333333e-06,
+      "loss": 0.8669,
+      "step": 5598
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.957785087719299e-06,
+      "loss": 0.8784,
+      "step": 5599
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.957236842105264e-06,
+      "loss": 0.8897,
+      "step": 5600
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.956688596491229e-06,
+      "loss": 0.966,
+      "step": 5601
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.956140350877194e-06,
+      "loss": 0.4482,
+      "step": 5602
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.955592105263158e-06,
+      "loss": 0.7459,
+      "step": 5603
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.955043859649123e-06,
+      "loss": 0.6761,
+      "step": 5604
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.954495614035088e-06,
+      "loss": 0.9072,
+      "step": 5605
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.953947368421054e-06,
+      "loss": 0.5298,
+      "step": 5606
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.9533991228070185e-06,
+      "loss": 0.9377,
+      "step": 5607
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.9528508771929824e-06,
+      "loss": 0.3434,
+      "step": 5608
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.952302631578948e-06,
+      "loss": 0.9992,
+      "step": 5609
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.951754385964913e-06,
+      "loss": 0.8668,
+      "step": 5610
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.951206140350878e-06,
+      "loss": 0.2293,
+      "step": 5611
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.9506578947368434e-06,
+      "loss": 0.6705,
+      "step": 5612
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.950109649122807e-06,
+      "loss": 0.4631,
+      "step": 5613
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.949561403508772e-06,
+      "loss": 1.0398,
+      "step": 5614
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.949013157894737e-06,
+      "loss": 1.3818,
+      "step": 5615
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.948464912280703e-06,
+      "loss": 0.8028,
+      "step": 5616
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.9479166666666676e-06,
+      "loss": 0.6199,
+      "step": 5617
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.947368421052632e-06,
+      "loss": 0.901,
+      "step": 5618
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.946820175438596e-06,
+      "loss": 0.8751,
+      "step": 5619
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.946271929824562e-06,
+      "loss": 0.2266,
+      "step": 5620
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.945723684210527e-06,
+      "loss": 0.9236,
+      "step": 5621
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.945175438596492e-06,
+      "loss": 0.6505,
+      "step": 5622
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.944627192982457e-06,
+      "loss": 1.2028,
+      "step": 5623
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.944078947368421e-06,
+      "loss": 0.2965,
+      "step": 5624
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.943530701754386e-06,
+      "loss": 0.4646,
+      "step": 5625
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.942982456140351e-06,
+      "loss": 0.6622,
+      "step": 5626
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.942434210526317e-06,
+      "loss": 0.7316,
+      "step": 5627
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.9418859649122815e-06,
+      "loss": 0.2212,
+      "step": 5628
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.941337719298247e-06,
+      "loss": 0.7166,
+      "step": 5629
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.940789473684211e-06,
+      "loss": 0.4853,
+      "step": 5630
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.940241228070176e-06,
+      "loss": 1.0563,
+      "step": 5631
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.939692982456141e-06,
+      "loss": 0.4142,
+      "step": 5632
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.9391447368421064e-06,
+      "loss": 0.6239,
+      "step": 5633
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.938596491228071e-06,
+      "loss": 1.0387,
+      "step": 5634
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.938048245614035e-06,
+      "loss": 1.3764,
+      "step": 5635
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.9375e-06,
+      "loss": 0.7162,
+      "step": 5636
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.936951754385966e-06,
+      "loss": 1.0339,
+      "step": 5637
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.9364035087719306e-06,
+      "loss": 0.8512,
+      "step": 5638
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.935855263157895e-06,
+      "loss": 0.582,
+      "step": 5639
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.935307017543859e-06,
+      "loss": 0.5808,
+      "step": 5640
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.934758771929825e-06,
+      "loss": 1.3053,
+      "step": 5641
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.93421052631579e-06,
+      "loss": 0.3281,
+      "step": 5642
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.933662280701755e-06,
+      "loss": 0.5521,
+      "step": 5643
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.93311403508772e-06,
+      "loss": 0.7424,
+      "step": 5644
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.932565789473685e-06,
+      "loss": 0.5549,
+      "step": 5645
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.932017543859649e-06,
+      "loss": 0.9916,
+      "step": 5646
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.931469298245615e-06,
+      "loss": 0.7736,
+      "step": 5647
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.93092105263158e-06,
+      "loss": 0.8461,
+      "step": 5648
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.9303728070175445e-06,
+      "loss": 1.1946,
+      "step": 5649
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.92982456140351e-06,
+      "loss": 0.4564,
+      "step": 5650
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.929276315789474e-06,
+      "loss": 0.5899,
+      "step": 5651
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.928728070175439e-06,
+      "loss": 1.2474,
+      "step": 5652
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.928179824561404e-06,
+      "loss": 0.8188,
+      "step": 5653
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.9276315789473695e-06,
+      "loss": 1.7644,
+      "step": 5654
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.927083333333334e-06,
+      "loss": 1.6574,
+      "step": 5655
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.926535087719298e-06,
+      "loss": 0.7457,
+      "step": 5656
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.925986842105263e-06,
+      "loss": 1.432,
+      "step": 5657
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.925438596491229e-06,
+      "loss": 0.5805,
+      "step": 5658
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.924890350877194e-06,
+      "loss": 1.1324,
+      "step": 5659
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.924342105263158e-06,
+      "loss": 0.8925,
+      "step": 5660
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.923793859649124e-06,
+      "loss": 0.3609,
+      "step": 5661
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.923245614035088e-06,
+      "loss": 0.8249,
+      "step": 5662
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.922697368421053e-06,
+      "loss": 0.5679,
+      "step": 5663
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.922149122807018e-06,
+      "loss": 1.3732,
+      "step": 5664
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.921600877192983e-06,
+      "loss": 0.8087,
+      "step": 5665
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.921052631578948e-06,
+      "loss": 0.4965,
+      "step": 5666
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.920504385964912e-06,
+      "loss": 1.285,
+      "step": 5667
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.919956140350878e-06,
+      "loss": 0.3464,
+      "step": 5668
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.919407894736843e-06,
+      "loss": 1.0495,
+      "step": 5669
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.9188596491228075e-06,
+      "loss": 0.4012,
+      "step": 5670
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.918311403508773e-06,
+      "loss": 0.5578,
+      "step": 5671
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.917763157894738e-06,
+      "loss": 1.1147,
+      "step": 5672
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.917214912280702e-06,
+      "loss": 1.3537,
+      "step": 5673
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.916666666666667e-06,
+      "loss": 1.0609,
+      "step": 5674
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.9161184210526325e-06,
+      "loss": 0.8635,
+      "step": 5675
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.915570175438597e-06,
+      "loss": 1.0344,
+      "step": 5676
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.915021929824562e-06,
+      "loss": 0.9019,
+      "step": 5677
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.914473684210526e-06,
+      "loss": 0.8421,
+      "step": 5678
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.913925438596492e-06,
+      "loss": 1.2648,
+      "step": 5679
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.913377192982457e-06,
+      "loss": 0.986,
+      "step": 5680
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.912828947368421e-06,
+      "loss": 0.739,
+      "step": 5681
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.912280701754387e-06,
+      "loss": 0.6299,
+      "step": 5682
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.911732456140351e-06,
+      "loss": 0.642,
+      "step": 5683
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.911184210526316e-06,
+      "loss": 0.21,
+      "step": 5684
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.9106359649122816e-06,
+      "loss": 0.3821,
+      "step": 5685
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.910087719298246e-06,
+      "loss": 1.5843,
+      "step": 5686
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.909539473684211e-06,
+      "loss": 1.1891,
+      "step": 5687
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.908991228070177e-06,
+      "loss": 1.105,
+      "step": 5688
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.908442982456141e-06,
+      "loss": 1.2894,
+      "step": 5689
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.907894736842106e-06,
+      "loss": 0.8778,
+      "step": 5690
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.9073464912280705e-06,
+      "loss": 0.3019,
+      "step": 5691
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.906798245614036e-06,
+      "loss": 0.6271,
+      "step": 5692
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.906250000000001e-06,
+      "loss": 0.8595,
+      "step": 5693
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.905701754385965e-06,
+      "loss": 1.2653,
+      "step": 5694
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.90515350877193e-06,
+      "loss": 0.5375,
+      "step": 5695
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.9046052631578955e-06,
+      "loss": 1.0007,
+      "step": 5696
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.90405701754386e-06,
+      "loss": 0.8935,
+      "step": 5697
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.903508771929825e-06,
+      "loss": 0.4519,
+      "step": 5698
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.902960526315789e-06,
+      "loss": 0.8263,
+      "step": 5699
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.902412280701755e-06,
+      "loss": 0.3525,
+      "step": 5700
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.90186403508772e-06,
+      "loss": 0.5987,
+      "step": 5701
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.901315789473685e-06,
+      "loss": 0.4892,
+      "step": 5702
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.90076754385965e-06,
+      "loss": 1.2832,
+      "step": 5703
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.900219298245615e-06,
+      "loss": 0.6377,
+      "step": 5704
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.899671052631579e-06,
+      "loss": 1.183,
+      "step": 5705
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.8991228070175446e-06,
+      "loss": 0.5248,
+      "step": 5706
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.898574561403509e-06,
+      "loss": 1.6308,
+      "step": 5707
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.898026315789474e-06,
+      "loss": 0.349,
+      "step": 5708
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.89747807017544e-06,
+      "loss": 0.6982,
+      "step": 5709
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.896929824561404e-06,
+      "loss": 0.6723,
+      "step": 5710
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.896381578947369e-06,
+      "loss": 1.4079,
+      "step": 5711
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.8958333333333335e-06,
+      "loss": 0.4316,
+      "step": 5712
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.895285087719299e-06,
+      "loss": 0.6101,
+      "step": 5713
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.894736842105264e-06,
+      "loss": 0.4647,
+      "step": 5714
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.894188596491229e-06,
+      "loss": 0.882,
+      "step": 5715
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.893640350877193e-06,
+      "loss": 0.9619,
+      "step": 5716
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.8930921052631585e-06,
+      "loss": 1.0697,
+      "step": 5717
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.892543859649123e-06,
+      "loss": 0.9497,
+      "step": 5718
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.891995614035088e-06,
+      "loss": 0.9783,
+      "step": 5719
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.891447368421054e-06,
+      "loss": 0.3938,
+      "step": 5720
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.890899122807018e-06,
+      "loss": 0.6542,
+      "step": 5721
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.890350877192983e-06,
+      "loss": 0.8094,
+      "step": 5722
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.889802631578948e-06,
+      "loss": 0.6835,
+      "step": 5723
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.889254385964913e-06,
+      "loss": 0.319,
+      "step": 5724
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.888706140350878e-06,
+      "loss": 1.0762,
+      "step": 5725
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.888157894736842e-06,
+      "loss": 0.5274,
+      "step": 5726
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.887609649122808e-06,
+      "loss": 0.3087,
+      "step": 5727
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.887061403508772e-06,
+      "loss": 0.8579,
+      "step": 5728
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.886513157894737e-06,
+      "loss": 0.7698,
+      "step": 5729
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.885964912280703e-06,
+      "loss": 0.6216,
+      "step": 5730
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.885416666666668e-06,
+      "loss": 0.6095,
+      "step": 5731
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.884868421052632e-06,
+      "loss": 0.5196,
+      "step": 5732
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.8843201754385965e-06,
+      "loss": 1.1867,
+      "step": 5733
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.883771929824562e-06,
+      "loss": 1.2522,
+      "step": 5734
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.883223684210527e-06,
+      "loss": 0.8969,
+      "step": 5735
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.882675438596492e-06,
+      "loss": 0.317,
+      "step": 5736
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.882127192982456e-06,
+      "loss": 0.3542,
+      "step": 5737
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.8815789473684215e-06,
+      "loss": 0.5295,
+      "step": 5738
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.881030701754386e-06,
+      "loss": 0.6582,
+      "step": 5739
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.880482456140352e-06,
+      "loss": 1.0774,
+      "step": 5740
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.879934210526317e-06,
+      "loss": 0.7154,
+      "step": 5741
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.879385964912281e-06,
+      "loss": 0.5505,
+      "step": 5742
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.878837719298246e-06,
+      "loss": 0.4561,
+      "step": 5743
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.878289473684211e-06,
+      "loss": 1.3223,
+      "step": 5744
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.877741228070176e-06,
+      "loss": 0.4756,
+      "step": 5745
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.877192982456141e-06,
+      "loss": 1.3167,
+      "step": 5746
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.876644736842107e-06,
+      "loss": 0.9226,
+      "step": 5747
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.876096491228071e-06,
+      "loss": 0.6545,
+      "step": 5748
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.875548245614035e-06,
+      "loss": 0.6721,
+      "step": 5749
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.875e-06,
+      "loss": 0.9261,
+      "step": 5750
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.874451754385966e-06,
+      "loss": 0.7891,
+      "step": 5751
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.873903508771931e-06,
+      "loss": 0.7515,
+      "step": 5752
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.873355263157895e-06,
+      "loss": 0.8013,
+      "step": 5753
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.8728070175438595e-06,
+      "loss": 0.4827,
+      "step": 5754
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.872258771929825e-06,
+      "loss": 0.8212,
+      "step": 5755
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.87171052631579e-06,
+      "loss": 1.0625,
+      "step": 5756
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.871162280701755e-06,
+      "loss": 1.4908,
+      "step": 5757
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.8706140350877205e-06,
+      "loss": 0.8645,
+      "step": 5758
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.8700657894736845e-06,
+      "loss": 0.8549,
+      "step": 5759
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.869517543859649e-06,
+      "loss": 0.5514,
+      "step": 5760
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.868969298245615e-06,
+      "loss": 0.9741,
+      "step": 5761
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.86842105263158e-06,
+      "loss": 0.6294,
+      "step": 5762
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.867872807017545e-06,
+      "loss": 1.3653,
+      "step": 5763
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.867324561403509e-06,
+      "loss": 1.0883,
+      "step": 5764
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.866776315789474e-06,
+      "loss": 1.4986,
+      "step": 5765
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.866228070175439e-06,
+      "loss": 0.9021,
+      "step": 5766
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.865679824561404e-06,
+      "loss": 0.8918,
+      "step": 5767
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.86513157894737e-06,
+      "loss": 0.574,
+      "step": 5768
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.864583333333334e-06,
+      "loss": 1.0318,
+      "step": 5769
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.8640350877192984e-06,
+      "loss": 1.026,
+      "step": 5770
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.863486842105263e-06,
+      "loss": 0.4734,
+      "step": 5771
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.862938596491229e-06,
+      "loss": 0.6529,
+      "step": 5772
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.862390350877194e-06,
+      "loss": 0.6934,
+      "step": 5773
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.861842105263159e-06,
+      "loss": 0.3787,
+      "step": 5774
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.861293859649123e-06,
+      "loss": 1.0171,
+      "step": 5775
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.860745614035088e-06,
+      "loss": 1.3287,
+      "step": 5776
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.860197368421053e-06,
+      "loss": 1.002,
+      "step": 5777
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.859649122807019e-06,
+      "loss": 0.4162,
+      "step": 5778
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.8591008771929835e-06,
+      "loss": 0.9123,
+      "step": 5779
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.8585526315789475e-06,
+      "loss": 0.5041,
+      "step": 5780
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.858004385964912e-06,
+      "loss": 0.5269,
+      "step": 5781
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.857456140350878e-06,
+      "loss": 0.5326,
+      "step": 5782
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.856907894736843e-06,
+      "loss": 0.7176,
+      "step": 5783
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.856359649122808e-06,
+      "loss": 1.3592,
+      "step": 5784
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.855811403508772e-06,
+      "loss": 0.3671,
+      "step": 5785
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.855263157894737e-06,
+      "loss": 0.6491,
+      "step": 5786
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.854714912280702e-06,
+      "loss": 0.9524,
+      "step": 5787
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.854166666666667e-06,
+      "loss": 0.8194,
+      "step": 5788
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.853618421052633e-06,
+      "loss": 0.5093,
+      "step": 5789
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.8530701754385975e-06,
+      "loss": 0.7303,
+      "step": 5790
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.8525219298245614e-06,
+      "loss": 0.5755,
+      "step": 5791
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.851973684210526e-06,
+      "loss": 0.7383,
+      "step": 5792
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.851425438596492e-06,
+      "loss": 0.5116,
+      "step": 5793
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.850877192982457e-06,
+      "loss": 0.4567,
+      "step": 5794
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.8503289473684224e-06,
+      "loss": 0.6628,
+      "step": 5795
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.849780701754386e-06,
+      "loss": 0.9914,
+      "step": 5796
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.849232456140351e-06,
+      "loss": 0.7504,
+      "step": 5797
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.848684210526316e-06,
+      "loss": 0.8205,
+      "step": 5798
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.848135964912282e-06,
+      "loss": 0.354,
+      "step": 5799
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.8475877192982466e-06,
+      "loss": 0.7944,
+      "step": 5800
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.847039473684211e-06,
+      "loss": 0.5257,
+      "step": 5801
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.846491228070175e-06,
+      "loss": 0.8076,
+      "step": 5802
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.845942982456141e-06,
+      "loss": 0.9185,
+      "step": 5803
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.845394736842106e-06,
+      "loss": 0.7867,
+      "step": 5804
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.844846491228071e-06,
+      "loss": 0.427,
+      "step": 5805
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.844298245614036e-06,
+      "loss": 0.7839,
+      "step": 5806
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.84375e-06,
+      "loss": 0.4703,
+      "step": 5807
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.843201754385965e-06,
+      "loss": 1.2988,
+      "step": 5808
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.84265350877193e-06,
+      "loss": 0.9464,
+      "step": 5809
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.842105263157896e-06,
+      "loss": 0.8536,
+      "step": 5810
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.8415570175438605e-06,
+      "loss": 0.6989,
+      "step": 5811
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.8410087719298245e-06,
+      "loss": 1.0172,
+      "step": 5812
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.84046052631579e-06,
+      "loss": 0.8354,
+      "step": 5813
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.839912280701755e-06,
+      "loss": 0.4691,
+      "step": 5814
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.83936403508772e-06,
+      "loss": 0.9338,
+      "step": 5815
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.8388157894736854e-06,
+      "loss": 0.3533,
+      "step": 5816
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.83826754385965e-06,
+      "loss": 0.285,
+      "step": 5817
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.837719298245614e-06,
+      "loss": 0.7472,
+      "step": 5818
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.837171052631579e-06,
+      "loss": 1.0526,
+      "step": 5819
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.836622807017545e-06,
+      "loss": 1.5308,
+      "step": 5820
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.8360745614035096e-06,
+      "loss": 1.0423,
+      "step": 5821
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.835526315789474e-06,
+      "loss": 0.7575,
+      "step": 5822
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.834978070175438e-06,
+      "loss": 0.4156,
+      "step": 5823
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.834429824561404e-06,
+      "loss": 0.2783,
+      "step": 5824
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.833881578947369e-06,
+      "loss": 0.6864,
+      "step": 5825
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.833333333333334e-06,
+      "loss": 0.7613,
+      "step": 5826
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.832785087719299e-06,
+      "loss": 0.4199,
+      "step": 5827
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.832236842105263e-06,
+      "loss": 0.8412,
+      "step": 5828
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.831688596491228e-06,
+      "loss": 1.0931,
+      "step": 5829
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.831140350877193e-06,
+      "loss": 0.5783,
+      "step": 5830
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.830592105263159e-06,
+      "loss": 0.6557,
+      "step": 5831
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.8300438596491235e-06,
+      "loss": 1.0206,
+      "step": 5832
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.829495614035089e-06,
+      "loss": 0.7435,
+      "step": 5833
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.828947368421053e-06,
+      "loss": 0.3819,
+      "step": 5834
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.828399122807018e-06,
+      "loss": 0.7707,
+      "step": 5835
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.827850877192983e-06,
+      "loss": 0.5468,
+      "step": 5836
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.8273026315789485e-06,
+      "loss": 0.5988,
+      "step": 5837
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.826754385964913e-06,
+      "loss": 0.5648,
+      "step": 5838
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.826206140350877e-06,
+      "loss": 0.9183,
+      "step": 5839
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.825657894736842e-06,
+      "loss": 0.4214,
+      "step": 5840
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.825109649122808e-06,
+      "loss": 0.6867,
+      "step": 5841
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.824561403508773e-06,
+      "loss": 0.7301,
+      "step": 5842
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.824013157894737e-06,
+      "loss": 0.969,
+      "step": 5843
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.823464912280703e-06,
+      "loss": 0.5544,
+      "step": 5844
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.822916666666667e-06,
+      "loss": 0.824,
+      "step": 5845
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.822368421052632e-06,
+      "loss": 1.0395,
+      "step": 5846
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.821820175438597e-06,
+      "loss": 0.8928,
+      "step": 5847
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.821271929824562e-06,
+      "loss": 0.3743,
+      "step": 5848
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.820723684210527e-06,
+      "loss": 0.7062,
+      "step": 5849
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.820175438596491e-06,
+      "loss": 0.7278,
+      "step": 5850
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.819627192982457e-06,
+      "loss": 0.9543,
+      "step": 5851
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.819078947368422e-06,
+      "loss": 0.948,
+      "step": 5852
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.8185307017543865e-06,
+      "loss": 0.9226,
+      "step": 5853
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.817982456140352e-06,
+      "loss": 0.3369,
+      "step": 5854
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.817434210526316e-06,
+      "loss": 0.9605,
+      "step": 5855
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.816885964912281e-06,
+      "loss": 0.9449,
+      "step": 5856
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.816337719298246e-06,
+      "loss": 0.6416,
+      "step": 5857
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.8157894736842115e-06,
+      "loss": 0.5029,
+      "step": 5858
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.815241228070176e-06,
+      "loss": 0.7268,
+      "step": 5859
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.814692982456141e-06,
+      "loss": 0.9399,
+      "step": 5860
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.814144736842105e-06,
+      "loss": 0.3117,
+      "step": 5861
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.813596491228071e-06,
+      "loss": 0.5216,
+      "step": 5862
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.813048245614036e-06,
+      "loss": 0.8233,
+      "step": 5863
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.8125e-06,
+      "loss": 0.7747,
+      "step": 5864
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.811951754385966e-06,
+      "loss": 0.6286,
+      "step": 5865
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.81140350877193e-06,
+      "loss": 0.5773,
+      "step": 5866
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.810855263157895e-06,
+      "loss": 0.8727,
+      "step": 5867
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.8103070175438606e-06,
+      "loss": 0.5686,
+      "step": 5868
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.809758771929825e-06,
+      "loss": 0.5405,
+      "step": 5869
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.80921052631579e-06,
+      "loss": 0.3803,
+      "step": 5870
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.808662280701754e-06,
+      "loss": 1.8056,
+      "step": 5871
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.80811403508772e-06,
+      "loss": 0.6467,
+      "step": 5872
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.807565789473685e-06,
+      "loss": 1.0719,
+      "step": 5873
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.8070175438596495e-06,
+      "loss": 0.8119,
+      "step": 5874
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.806469298245615e-06,
+      "loss": 0.6308,
+      "step": 5875
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.80592105263158e-06,
+      "loss": 0.5795,
+      "step": 5876
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.805372807017544e-06,
+      "loss": 1.0321,
+      "step": 5877
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.804824561403509e-06,
+      "loss": 1.041,
+      "step": 5878
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.8042763157894745e-06,
+      "loss": 0.9263,
+      "step": 5879
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.803728070175439e-06,
+      "loss": 0.983,
+      "step": 5880
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.803179824561404e-06,
+      "loss": 1.0106,
+      "step": 5881
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.802631578947368e-06,
+      "loss": 0.4945,
+      "step": 5882
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.802083333333334e-06,
+      "loss": 0.8628,
+      "step": 5883
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.801535087719299e-06,
+      "loss": 1.0121,
+      "step": 5884
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.8009868421052634e-06,
+      "loss": 0.6949,
+      "step": 5885
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.800438596491229e-06,
+      "loss": 0.2786,
+      "step": 5886
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.799890350877194e-06,
+      "loss": 0.6561,
+      "step": 5887
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.799342105263158e-06,
+      "loss": 0.7268,
+      "step": 5888
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.798793859649124e-06,
+      "loss": 0.5162,
+      "step": 5889
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.798245614035088e-06,
+      "loss": 0.7658,
+      "step": 5890
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.797697368421053e-06,
+      "loss": 0.7774,
+      "step": 5891
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.797149122807019e-06,
+      "loss": 0.4408,
+      "step": 5892
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.796600877192983e-06,
+      "loss": 0.2595,
+      "step": 5893
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.796052631578948e-06,
+      "loss": 0.4437,
+      "step": 5894
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.7955043859649125e-06,
+      "loss": 0.7716,
+      "step": 5895
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.794956140350878e-06,
+      "loss": 0.7942,
+      "step": 5896
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.794407894736843e-06,
+      "loss": 1.1839,
+      "step": 5897
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.793859649122807e-06,
+      "loss": 0.8505,
+      "step": 5898
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.793311403508772e-06,
+      "loss": 0.9721,
+      "step": 5899
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.7927631578947375e-06,
+      "loss": 0.9814,
+      "step": 5900
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.792214912280702e-06,
+      "loss": 0.5981,
+      "step": 5901
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.791666666666667e-06,
+      "loss": 0.8456,
+      "step": 5902
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.791118421052633e-06,
+      "loss": 1.3838,
+      "step": 5903
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.790570175438597e-06,
+      "loss": 0.6339,
+      "step": 5904
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.790021929824562e-06,
+      "loss": 0.54,
+      "step": 5905
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.789473684210527e-06,
+      "loss": 0.7032,
+      "step": 5906
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.788925438596492e-06,
+      "loss": 1.3292,
+      "step": 5907
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.788377192982457e-06,
+      "loss": 0.7375,
+      "step": 5908
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.787828947368421e-06,
+      "loss": 0.7224,
+      "step": 5909
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.787280701754387e-06,
+      "loss": 1.2044,
+      "step": 5910
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.786732456140351e-06,
+      "loss": 0.5114,
+      "step": 5911
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.786184210526316e-06,
+      "loss": 0.818,
+      "step": 5912
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.785635964912282e-06,
+      "loss": 0.5964,
+      "step": 5913
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.785087719298247e-06,
+      "loss": 0.8255,
+      "step": 5914
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.784539473684211e-06,
+      "loss": 1.176,
+      "step": 5915
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.7839912280701755e-06,
+      "loss": 0.737,
+      "step": 5916
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.783442982456141e-06,
+      "loss": 1.494,
+      "step": 5917
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.782894736842106e-06,
+      "loss": 1.6121,
+      "step": 5918
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.782346491228071e-06,
+      "loss": 0.4245,
+      "step": 5919
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.781798245614035e-06,
+      "loss": 0.5263,
+      "step": 5920
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.7812500000000005e-06,
+      "loss": 1.2604,
+      "step": 5921
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.780701754385965e-06,
+      "loss": 1.666,
+      "step": 5922
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.78015350877193e-06,
+      "loss": 1.117,
+      "step": 5923
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.779605263157896e-06,
+      "loss": 1.6133,
+      "step": 5924
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.77905701754386e-06,
+      "loss": 0.8736,
+      "step": 5925
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.778508771929825e-06,
+      "loss": 0.6376,
+      "step": 5926
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.77796052631579e-06,
+      "loss": 0.8697,
+      "step": 5927
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.777412280701755e-06,
+      "loss": 0.6477,
+      "step": 5928
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.77686403508772e-06,
+      "loss": 0.357,
+      "step": 5929
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.776315789473686e-06,
+      "loss": 0.7821,
+      "step": 5930
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.77576754385965e-06,
+      "loss": 1.0781,
+      "step": 5931
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.7752192982456144e-06,
+      "loss": 0.9902,
+      "step": 5932
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.774671052631579e-06,
+      "loss": 0.4834,
+      "step": 5933
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.774122807017545e-06,
+      "loss": 0.5915,
+      "step": 5934
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.77357456140351e-06,
+      "loss": 0.9919,
+      "step": 5935
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.773026315789474e-06,
+      "loss": 0.353,
+      "step": 5936
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.7724780701754385e-06,
+      "loss": 0.474,
+      "step": 5937
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.771929824561404e-06,
+      "loss": 0.4632,
+      "step": 5938
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.771381578947369e-06,
+      "loss": 0.7808,
+      "step": 5939
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.770833333333334e-06,
+      "loss": 0.8162,
+      "step": 5940
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.770285087719298e-06,
+      "loss": 1.0974,
+      "step": 5941
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.7697368421052635e-06,
+      "loss": 0.6195,
+      "step": 5942
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.769188596491228e-06,
+      "loss": 0.7684,
+      "step": 5943
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.768640350877194e-06,
+      "loss": 0.898,
+      "step": 5944
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.768092105263159e-06,
+      "loss": 1.223,
+      "step": 5945
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.767543859649124e-06,
+      "loss": 0.3093,
+      "step": 5946
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.766995614035088e-06,
+      "loss": 0.4237,
+      "step": 5947
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.766447368421053e-06,
+      "loss": 0.6096,
+      "step": 5948
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.765899122807018e-06,
+      "loss": 0.6715,
+      "step": 5949
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.765350877192983e-06,
+      "loss": 0.7068,
+      "step": 5950
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.764802631578949e-06,
+      "loss": 0.8777,
+      "step": 5951
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.764254385964913e-06,
+      "loss": 1.0019,
+      "step": 5952
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.7637061403508774e-06,
+      "loss": 1.1405,
+      "step": 5953
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.763157894736842e-06,
+      "loss": 0.7154,
+      "step": 5954
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.762609649122808e-06,
+      "loss": 0.8678,
+      "step": 5955
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.762061403508773e-06,
+      "loss": 1.0167,
+      "step": 5956
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.761513157894738e-06,
+      "loss": 1.2675,
+      "step": 5957
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.7609649122807016e-06,
+      "loss": 1.0217,
+      "step": 5958
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.760416666666667e-06,
+      "loss": 1.2135,
+      "step": 5959
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.759868421052632e-06,
+      "loss": 1.0904,
+      "step": 5960
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.759320175438598e-06,
+      "loss": 0.7151,
+      "step": 5961
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.7587719298245626e-06,
+      "loss": 0.4475,
+      "step": 5962
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.7582236842105265e-06,
+      "loss": 1.0152,
+      "step": 5963
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.757675438596491e-06,
+      "loss": 0.1415,
+      "step": 5964
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.757127192982457e-06,
+      "loss": 1.3074,
+      "step": 5965
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.756578947368422e-06,
+      "loss": 0.815,
+      "step": 5966
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.756030701754387e-06,
+      "loss": 0.6736,
+      "step": 5967
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.755482456140351e-06,
+      "loss": 1.3243,
+      "step": 5968
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.754934210526316e-06,
+      "loss": 0.9323,
+      "step": 5969
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.754385964912281e-06,
+      "loss": 0.4844,
+      "step": 5970
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.753837719298246e-06,
+      "loss": 1.0794,
+      "step": 5971
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.753289473684212e-06,
+      "loss": 0.7974,
+      "step": 5972
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.7527412280701765e-06,
+      "loss": 1.0691,
+      "step": 5973
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.7521929824561404e-06,
+      "loss": 0.5008,
+      "step": 5974
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.751644736842105e-06,
+      "loss": 0.8864,
+      "step": 5975
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.751096491228071e-06,
+      "loss": 0.2245,
+      "step": 5976
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.750548245614036e-06,
+      "loss": 0.5407,
+      "step": 5977
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.750000000000001e-06,
+      "loss": 0.5306,
+      "step": 5978
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.749451754385965e-06,
+      "loss": 0.5248,
+      "step": 5979
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.74890350877193e-06,
+      "loss": 0.7549,
+      "step": 5980
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.748355263157895e-06,
+      "loss": 0.72,
+      "step": 5981
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.747807017543861e-06,
+      "loss": 0.6325,
+      "step": 5982
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.7472587719298256e-06,
+      "loss": 0.8999,
+      "step": 5983
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.7467105263157895e-06,
+      "loss": 0.8125,
+      "step": 5984
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.746162280701754e-06,
+      "loss": 1.4559,
+      "step": 5985
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.74561403508772e-06,
+      "loss": 0.7379,
+      "step": 5986
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.745065789473685e-06,
+      "loss": 0.9404,
+      "step": 5987
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.74451754385965e-06,
+      "loss": 1.1317,
+      "step": 5988
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.743969298245615e-06,
+      "loss": 0.4938,
+      "step": 5989
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.743421052631579e-06,
+      "loss": 0.4966,
+      "step": 5990
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.742872807017544e-06,
+      "loss": 1.4467,
+      "step": 5991
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.742324561403509e-06,
+      "loss": 0.5421,
+      "step": 5992
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.741776315789475e-06,
+      "loss": 0.558,
+      "step": 5993
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.7412280701754395e-06,
+      "loss": 0.9055,
+      "step": 5994
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.7406798245614035e-06,
+      "loss": 0.5796,
+      "step": 5995
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.740131578947368e-06,
+      "loss": 0.9142,
+      "step": 5996
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.739583333333334e-06,
+      "loss": 0.759,
+      "step": 5997
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.739035087719299e-06,
+      "loss": 0.5257,
+      "step": 5998
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.7384868421052645e-06,
+      "loss": 1.0408,
+      "step": 5999
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.737938596491229e-06,
+      "loss": 0.6374,
+      "step": 6000
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.737390350877193e-06,
+      "loss": 1.156,
+      "step": 6001
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.736842105263158e-06,
+      "loss": 0.6135,
+      "step": 6002
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.736293859649124e-06,
+      "loss": 0.6388,
+      "step": 6003
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.735745614035089e-06,
+      "loss": 0.7168,
+      "step": 6004
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.735197368421053e-06,
+      "loss": 1.2015,
+      "step": 6005
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.734649122807017e-06,
+      "loss": 0.9935,
+      "step": 6006
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.734100877192983e-06,
+      "loss": 0.4447,
+      "step": 6007
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.733552631578948e-06,
+      "loss": 1.1438,
+      "step": 6008
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.733004385964913e-06,
+      "loss": 0.2976,
+      "step": 6009
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.732456140350878e-06,
+      "loss": 0.9968,
+      "step": 6010
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.731907894736842e-06,
+      "loss": 0.8208,
+      "step": 6011
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.731359649122807e-06,
+      "loss": 0.6391,
+      "step": 6012
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.730811403508772e-06,
+      "loss": 0.677,
+      "step": 6013
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.730263157894738e-06,
+      "loss": 0.5957,
+      "step": 6014
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.7297149122807025e-06,
+      "loss": 0.5958,
+      "step": 6015
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.729166666666667e-06,
+      "loss": 1.0208,
+      "step": 6016
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.728618421052632e-06,
+      "loss": 1.0972,
+      "step": 6017
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.728070175438597e-06,
+      "loss": 0.5771,
+      "step": 6018
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.727521929824562e-06,
+      "loss": 0.5148,
+      "step": 6019
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.7269736842105275e-06,
+      "loss": 0.2749,
+      "step": 6020
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.726425438596492e-06,
+      "loss": 1.2995,
+      "step": 6021
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.725877192982456e-06,
+      "loss": 0.8469,
+      "step": 6022
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.725328947368421e-06,
+      "loss": 0.5952,
+      "step": 6023
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.724780701754387e-06,
+      "loss": 0.7777,
+      "step": 6024
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.724232456140352e-06,
+      "loss": 0.7633,
+      "step": 6025
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.723684210526316e-06,
+      "loss": 0.9193,
+      "step": 6026
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.72313596491228e-06,
+      "loss": 0.4678,
+      "step": 6027
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.722587719298246e-06,
+      "loss": 0.4684,
+      "step": 6028
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.722039473684211e-06,
+      "loss": 1.0895,
+      "step": 6029
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.721491228070176e-06,
+      "loss": 1.414,
+      "step": 6030
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.720942982456141e-06,
+      "loss": 0.5544,
+      "step": 6031
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.720394736842106e-06,
+      "loss": 0.5917,
+      "step": 6032
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.71984649122807e-06,
+      "loss": 0.7689,
+      "step": 6033
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.719298245614035e-06,
+      "loss": 0.6487,
+      "step": 6034
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.718750000000001e-06,
+      "loss": 1.08,
+      "step": 6035
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.7182017543859655e-06,
+      "loss": 0.7015,
+      "step": 6036
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.717653508771931e-06,
+      "loss": 0.4811,
+      "step": 6037
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.717105263157895e-06,
+      "loss": 0.5988,
+      "step": 6038
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.71655701754386e-06,
+      "loss": 0.6984,
+      "step": 6039
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.716008771929825e-06,
+      "loss": 1.0248,
+      "step": 6040
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.7154605263157905e-06,
+      "loss": 0.8498,
+      "step": 6041
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.714912280701755e-06,
+      "loss": 0.6678,
+      "step": 6042
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.71436403508772e-06,
+      "loss": 0.2756,
+      "step": 6043
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.713815789473684e-06,
+      "loss": 0.6901,
+      "step": 6044
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.71326754385965e-06,
+      "loss": 0.9441,
+      "step": 6045
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.712719298245615e-06,
+      "loss": 0.8844,
+      "step": 6046
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.7121710526315794e-06,
+      "loss": 0.9076,
+      "step": 6047
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.711622807017545e-06,
+      "loss": 0.3548,
+      "step": 6048
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.711074561403509e-06,
+      "loss": 0.9951,
+      "step": 6049
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.710526315789474e-06,
+      "loss": 1.0109,
+      "step": 6050
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.709978070175439e-06,
+      "loss": 0.9834,
+      "step": 6051
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.709429824561404e-06,
+      "loss": 0.3656,
+      "step": 6052
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.708881578947369e-06,
+      "loss": 0.7452,
+      "step": 6053
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.708333333333333e-06,
+      "loss": 0.8629,
+      "step": 6054
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.707785087719299e-06,
+      "loss": 0.3245,
+      "step": 6055
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.707236842105264e-06,
+      "loss": 0.6072,
+      "step": 6056
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.7066885964912285e-06,
+      "loss": 0.2592,
+      "step": 6057
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.706140350877194e-06,
+      "loss": 0.8626,
+      "step": 6058
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.705592105263159e-06,
+      "loss": 0.2621,
+      "step": 6059
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.705043859649123e-06,
+      "loss": 0.5365,
+      "step": 6060
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.704495614035088e-06,
+      "loss": 0.9429,
+      "step": 6061
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.7039473684210535e-06,
+      "loss": 1.5787,
+      "step": 6062
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.703399122807018e-06,
+      "loss": 0.8365,
+      "step": 6063
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.702850877192983e-06,
+      "loss": 0.2783,
+      "step": 6064
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.702302631578947e-06,
+      "loss": 0.2949,
+      "step": 6065
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.701754385964913e-06,
+      "loss": 1.1369,
+      "step": 6066
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.701206140350878e-06,
+      "loss": 0.2177,
+      "step": 6067
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.7006578947368424e-06,
+      "loss": 0.9751,
+      "step": 6068
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.700109649122808e-06,
+      "loss": 0.7094,
+      "step": 6069
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.699561403508772e-06,
+      "loss": 0.9319,
+      "step": 6070
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.699013157894737e-06,
+      "loss": 1.1006,
+      "step": 6071
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.698464912280703e-06,
+      "loss": 1.2309,
+      "step": 6072
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.697916666666667e-06,
+      "loss": 0.5909,
+      "step": 6073
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.697368421052632e-06,
+      "loss": 1.1452,
+      "step": 6074
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.696820175438598e-06,
+      "loss": 0.7862,
+      "step": 6075
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.696271929824562e-06,
+      "loss": 0.7452,
+      "step": 6076
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.695723684210527e-06,
+      "loss": 0.2749,
+      "step": 6077
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.6951754385964915e-06,
+      "loss": 1.4286,
+      "step": 6078
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.694627192982457e-06,
+      "loss": 1.1424,
+      "step": 6079
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.694078947368422e-06,
+      "loss": 1.2039,
+      "step": 6080
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.693530701754386e-06,
+      "loss": 1.2035,
+      "step": 6081
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.692982456140351e-06,
+      "loss": 0.905,
+      "step": 6082
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.6924342105263165e-06,
+      "loss": 0.8965,
+      "step": 6083
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.691885964912281e-06,
+      "loss": 1.1683,
+      "step": 6084
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.691337719298246e-06,
+      "loss": 1.0997,
+      "step": 6085
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.690789473684212e-06,
+      "loss": 0.2855,
+      "step": 6086
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.690241228070176e-06,
+      "loss": 0.8644,
+      "step": 6087
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.689692982456141e-06,
+      "loss": 0.5433,
+      "step": 6088
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.6891447368421054e-06,
+      "loss": 1.4592,
+      "step": 6089
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.688596491228071e-06,
+      "loss": 0.9062,
+      "step": 6090
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.688048245614036e-06,
+      "loss": 0.9674,
+      "step": 6091
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.6875e-06,
+      "loss": 1.1366,
+      "step": 6092
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.686951754385966e-06,
+      "loss": 0.8447,
+      "step": 6093
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.68640350877193e-06,
+      "loss": 0.5285,
+      "step": 6094
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.685855263157895e-06,
+      "loss": 1.9269,
+      "step": 6095
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.685307017543861e-06,
+      "loss": 0.8522,
+      "step": 6096
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.684758771929825e-06,
+      "loss": 0.4851,
+      "step": 6097
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.68421052631579e-06,
+      "loss": 0.8068,
+      "step": 6098
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.6836622807017545e-06,
+      "loss": 0.8993,
+      "step": 6099
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.68311403508772e-06,
+      "loss": 0.8993,
+      "step": 6100
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.682565789473685e-06,
+      "loss": 0.5587,
+      "step": 6101
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.68201754385965e-06,
+      "loss": 0.6299,
+      "step": 6102
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.681469298245614e-06,
+      "loss": 0.8251,
+      "step": 6103
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.6809210526315795e-06,
+      "loss": 0.5264,
+      "step": 6104
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.680372807017544e-06,
+      "loss": 0.6229,
+      "step": 6105
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.679824561403509e-06,
+      "loss": 0.6293,
+      "step": 6106
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.679276315789475e-06,
+      "loss": 0.5477,
+      "step": 6107
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.678728070175439e-06,
+      "loss": 1.2058,
+      "step": 6108
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.678179824561404e-06,
+      "loss": 1.0936,
+      "step": 6109
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.677631578947369e-06,
+      "loss": 0.8216,
+      "step": 6110
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.677083333333334e-06,
+      "loss": 0.7268,
+      "step": 6111
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.676535087719299e-06,
+      "loss": 0.2042,
+      "step": 6112
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.675986842105263e-06,
+      "loss": 0.3452,
+      "step": 6113
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.675438596491229e-06,
+      "loss": 0.8006,
+      "step": 6114
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.6748903508771934e-06,
+      "loss": 0.5456,
+      "step": 6115
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.674342105263158e-06,
+      "loss": 0.3243,
+      "step": 6116
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.673793859649124e-06,
+      "loss": 0.7088,
+      "step": 6117
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.673245614035089e-06,
+      "loss": 0.4122,
+      "step": 6118
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.672697368421053e-06,
+      "loss": 0.6894,
+      "step": 6119
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.6721491228070176e-06,
+      "loss": 0.4541,
+      "step": 6120
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.671600877192983e-06,
+      "loss": 0.3679,
+      "step": 6121
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.671052631578948e-06,
+      "loss": 0.8802,
+      "step": 6122
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.670504385964913e-06,
+      "loss": 0.4624,
+      "step": 6123
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.669956140350877e-06,
+      "loss": 0.7459,
+      "step": 6124
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.6694078947368425e-06,
+      "loss": 0.4485,
+      "step": 6125
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.668859649122807e-06,
+      "loss": 0.6536,
+      "step": 6126
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.668311403508772e-06,
+      "loss": 0.6231,
+      "step": 6127
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.667763157894738e-06,
+      "loss": 1.1404,
+      "step": 6128
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.667214912280703e-06,
+      "loss": 0.2958,
+      "step": 6129
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 0.9392,
+      "step": 6130
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.666118421052632e-06,
+      "loss": 0.9028,
+      "step": 6131
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.665570175438597e-06,
+      "loss": 1.4162,
+      "step": 6132
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.665021929824562e-06,
+      "loss": 1.5933,
+      "step": 6133
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.664473684210528e-06,
+      "loss": 0.7365,
+      "step": 6134
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.663925438596492e-06,
+      "loss": 1.358,
+      "step": 6135
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.6633771929824564e-06,
+      "loss": 0.9465,
+      "step": 6136
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.662828947368421e-06,
+      "loss": 1.3673,
+      "step": 6137
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.662280701754387e-06,
+      "loss": 0.5122,
+      "step": 6138
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.661732456140352e-06,
+      "loss": 1.8387,
+      "step": 6139
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.661184210526316e-06,
+      "loss": 0.2762,
+      "step": 6140
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.6606359649122806e-06,
+      "loss": 0.3747,
+      "step": 6141
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.660087719298246e-06,
+      "loss": 1.2973,
+      "step": 6142
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.659539473684211e-06,
+      "loss": 0.3416,
+      "step": 6143
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.658991228070176e-06,
+      "loss": 0.292,
+      "step": 6144
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.6584429824561416e-06,
+      "loss": 0.7259,
+      "step": 6145
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.6578947368421055e-06,
+      "loss": 0.3403,
+      "step": 6146
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.65734649122807e-06,
+      "loss": 0.8615,
+      "step": 6147
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.656798245614036e-06,
+      "loss": 1.8543,
+      "step": 6148
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.656250000000001e-06,
+      "loss": 1.1688,
+      "step": 6149
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.655701754385966e-06,
+      "loss": 1.3215,
+      "step": 6150
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.65515350877193e-06,
+      "loss": 0.5367,
+      "step": 6151
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.654605263157895e-06,
+      "loss": 1.3569,
+      "step": 6152
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.65405701754386e-06,
+      "loss": 0.4805,
+      "step": 6153
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.653508771929825e-06,
+      "loss": 0.7337,
+      "step": 6154
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.652960526315791e-06,
+      "loss": 0.3604,
+      "step": 6155
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.652412280701755e-06,
+      "loss": 0.8736,
+      "step": 6156
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.6518640350877195e-06,
+      "loss": 0.4592,
+      "step": 6157
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.651315789473684e-06,
+      "loss": 0.7272,
+      "step": 6158
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.65076754385965e-06,
+      "loss": 1.171,
+      "step": 6159
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.650219298245615e-06,
+      "loss": 0.4113,
+      "step": 6160
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.64967105263158e-06,
+      "loss": 1.3443,
+      "step": 6161
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.649122807017544e-06,
+      "loss": 0.3672,
+      "step": 6162
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.648574561403509e-06,
+      "loss": 0.7417,
+      "step": 6163
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.648026315789474e-06,
+      "loss": 0.3547,
+      "step": 6164
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.64747807017544e-06,
+      "loss": 0.3871,
+      "step": 6165
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.6469298245614046e-06,
+      "loss": 0.3844,
+      "step": 6166
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.6463815789473685e-06,
+      "loss": 1.4475,
+      "step": 6167
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.645833333333333e-06,
+      "loss": 1.0502,
+      "step": 6168
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.645285087719299e-06,
+      "loss": 0.3536,
+      "step": 6169
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.644736842105264e-06,
+      "loss": 0.6291,
+      "step": 6170
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.644188596491229e-06,
+      "loss": 0.5391,
+      "step": 6171
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.643640350877194e-06,
+      "loss": 1.3391,
+      "step": 6172
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.643092105263158e-06,
+      "loss": 0.7112,
+      "step": 6173
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.642543859649123e-06,
+      "loss": 0.7346,
+      "step": 6174
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.641995614035088e-06,
+      "loss": 1.5773,
+      "step": 6175
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.641447368421054e-06,
+      "loss": 0.6308,
+      "step": 6176
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.6408991228070185e-06,
+      "loss": 0.8695,
+      "step": 6177
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.6403508771929825e-06,
+      "loss": 2.0575,
+      "step": 6178
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.639802631578947e-06,
+      "loss": 0.3774,
+      "step": 6179
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.639254385964913e-06,
+      "loss": 0.7595,
+      "step": 6180
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.638706140350878e-06,
+      "loss": 1.2372,
+      "step": 6181
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.638157894736843e-06,
+      "loss": 0.8642,
+      "step": 6182
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.6376096491228074e-06,
+      "loss": 0.8753,
+      "step": 6183
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.637061403508772e-06,
+      "loss": 0.8455,
+      "step": 6184
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.636513157894737e-06,
+      "loss": 1.2224,
+      "step": 6185
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.635964912280703e-06,
+      "loss": 0.9794,
+      "step": 6186
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.635416666666668e-06,
+      "loss": 1.3684,
+      "step": 6187
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.634868421052632e-06,
+      "loss": 0.4304,
+      "step": 6188
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.634320175438596e-06,
+      "loss": 0.9917,
+      "step": 6189
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.633771929824562e-06,
+      "loss": 0.5369,
+      "step": 6190
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.633223684210527e-06,
+      "loss": 0.5902,
+      "step": 6191
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.632675438596492e-06,
+      "loss": 1.339,
+      "step": 6192
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.632127192982457e-06,
+      "loss": 0.8154,
+      "step": 6193
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.631578947368421e-06,
+      "loss": 0.5931,
+      "step": 6194
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.631030701754386e-06,
+      "loss": 0.6413,
+      "step": 6195
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.630482456140351e-06,
+      "loss": 0.7226,
+      "step": 6196
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.629934210526317e-06,
+      "loss": 0.8426,
+      "step": 6197
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.6293859649122815e-06,
+      "loss": 1.0546,
+      "step": 6198
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.628837719298246e-06,
+      "loss": 0.6424,
+      "step": 6199
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.62828947368421e-06,
+      "loss": 0.3375,
+      "step": 6200
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.627741228070176e-06,
+      "loss": 0.7027,
+      "step": 6201
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.627192982456141e-06,
+      "loss": 0.9589,
+      "step": 6202
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.6266447368421065e-06,
+      "loss": 0.503,
+      "step": 6203
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.626096491228071e-06,
+      "loss": 0.3482,
+      "step": 6204
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.625548245614035e-06,
+      "loss": 0.5384,
+      "step": 6205
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.625e-06,
+      "loss": 1.2224,
+      "step": 6206
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.624451754385966e-06,
+      "loss": 0.6048,
+      "step": 6207
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.623903508771931e-06,
+      "loss": 1.0688,
+      "step": 6208
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.623355263157895e-06,
+      "loss": 1.1112,
+      "step": 6209
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.622807017543859e-06,
+      "loss": 1.32,
+      "step": 6210
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.622258771929825e-06,
+      "loss": 1.0981,
+      "step": 6211
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.62171052631579e-06,
+      "loss": 1.2036,
+      "step": 6212
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.621162280701755e-06,
+      "loss": 0.6031,
+      "step": 6213
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.62061403508772e-06,
+      "loss": 1.2794,
+      "step": 6214
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.620065789473685e-06,
+      "loss": 1.1108,
+      "step": 6215
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.619517543859649e-06,
+      "loss": 0.5903,
+      "step": 6216
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.618969298245614e-06,
+      "loss": 0.7176,
+      "step": 6217
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.61842105263158e-06,
+      "loss": 0.6999,
+      "step": 6218
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.6178728070175445e-06,
+      "loss": 0.7995,
+      "step": 6219
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.61732456140351e-06,
+      "loss": 0.676,
+      "step": 6220
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.616776315789474e-06,
+      "loss": 0.6926,
+      "step": 6221
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.616228070175439e-06,
+      "loss": 1.123,
+      "step": 6222
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.615679824561404e-06,
+      "loss": 1.2259,
+      "step": 6223
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.6151315789473695e-06,
+      "loss": 0.7544,
+      "step": 6224
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.614583333333334e-06,
+      "loss": 0.4938,
+      "step": 6225
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.614035087719298e-06,
+      "loss": 0.8173,
+      "step": 6226
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.613486842105263e-06,
+      "loss": 1.017,
+      "step": 6227
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.612938596491229e-06,
+      "loss": 0.9554,
+      "step": 6228
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.612390350877194e-06,
+      "loss": 0.6106,
+      "step": 6229
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.6118421052631584e-06,
+      "loss": 1.0171,
+      "step": 6230
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.611293859649124e-06,
+      "loss": 0.5165,
+      "step": 6231
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.610745614035088e-06,
+      "loss": 1.2967,
+      "step": 6232
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.610197368421053e-06,
+      "loss": 0.1441,
+      "step": 6233
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.609649122807018e-06,
+      "loss": 1.0448,
+      "step": 6234
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.609100877192983e-06,
+      "loss": 1.1681,
+      "step": 6235
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.608552631578948e-06,
+      "loss": 0.5601,
+      "step": 6236
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.608004385964912e-06,
+      "loss": 1.6644,
+      "step": 6237
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.607456140350878e-06,
+      "loss": 0.9657,
+      "step": 6238
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.606907894736843e-06,
+      "loss": 0.586,
+      "step": 6239
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.6063596491228075e-06,
+      "loss": 0.6878,
+      "step": 6240
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.605811403508773e-06,
+      "loss": 0.7443,
+      "step": 6241
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.605263157894738e-06,
+      "loss": 0.4632,
+      "step": 6242
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.604714912280702e-06,
+      "loss": 1.1334,
+      "step": 6243
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.604166666666667e-06,
+      "loss": 0.6849,
+      "step": 6244
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.6036184210526325e-06,
+      "loss": 0.5088,
+      "step": 6245
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.603070175438597e-06,
+      "loss": 0.3074,
+      "step": 6246
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.602521929824562e-06,
+      "loss": 0.2816,
+      "step": 6247
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.601973684210526e-06,
+      "loss": 0.6209,
+      "step": 6248
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.601425438596492e-06,
+      "loss": 0.6165,
+      "step": 6249
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.600877192982457e-06,
+      "loss": 0.8678,
+      "step": 6250
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.6003289473684214e-06,
+      "loss": 1.0742,
+      "step": 6251
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.599780701754387e-06,
+      "loss": 0.7374,
+      "step": 6252
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.599232456140351e-06,
+      "loss": 0.3798,
+      "step": 6253
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.598684210526316e-06,
+      "loss": 0.7394,
+      "step": 6254
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.598135964912281e-06,
+      "loss": 0.7978,
+      "step": 6255
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.597587719298246e-06,
+      "loss": 0.5914,
+      "step": 6256
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.597039473684211e-06,
+      "loss": 1.1439,
+      "step": 6257
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.596491228070177e-06,
+      "loss": 0.7715,
+      "step": 6258
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.595942982456141e-06,
+      "loss": 0.5736,
+      "step": 6259
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.595394736842106e-06,
+      "loss": 1.0529,
+      "step": 6260
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.5948464912280705e-06,
+      "loss": 0.4733,
+      "step": 6261
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.594298245614036e-06,
+      "loss": 0.6802,
+      "step": 6262
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.593750000000001e-06,
+      "loss": 1.3228,
+      "step": 6263
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.593201754385965e-06,
+      "loss": 0.4066,
+      "step": 6264
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.59265350877193e-06,
+      "loss": 1.5553,
+      "step": 6265
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.5921052631578955e-06,
+      "loss": 0.58,
+      "step": 6266
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.59155701754386e-06,
+      "loss": 1.2622,
+      "step": 6267
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.591008771929825e-06,
+      "loss": 0.5642,
+      "step": 6268
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.590460526315789e-06,
+      "loss": 1.0744,
+      "step": 6269
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.589912280701755e-06,
+      "loss": 0.7737,
+      "step": 6270
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.58936403508772e-06,
+      "loss": 0.4637,
+      "step": 6271
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.5888157894736845e-06,
+      "loss": 1.2118,
+      "step": 6272
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.58826754385965e-06,
+      "loss": 1.345,
+      "step": 6273
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.587719298245615e-06,
+      "loss": 0.7359,
+      "step": 6274
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.587171052631579e-06,
+      "loss": 0.6096,
+      "step": 6275
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.586622807017545e-06,
+      "loss": 0.3361,
+      "step": 6276
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.5860745614035094e-06,
+      "loss": 0.7524,
+      "step": 6277
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.585526315789474e-06,
+      "loss": 1.2544,
+      "step": 6278
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.58497807017544e-06,
+      "loss": 0.8385,
+      "step": 6279
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.584429824561404e-06,
+      "loss": 1.5011,
+      "step": 6280
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.583881578947369e-06,
+      "loss": 0.401,
+      "step": 6281
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.5833333333333335e-06,
+      "loss": 0.502,
+      "step": 6282
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.582785087719299e-06,
+      "loss": 0.7845,
+      "step": 6283
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.582236842105264e-06,
+      "loss": 1.1135,
+      "step": 6284
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.581688596491229e-06,
+      "loss": 0.4965,
+      "step": 6285
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.581140350877193e-06,
+      "loss": 1.6904,
+      "step": 6286
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.5805921052631585e-06,
+      "loss": 0.8893,
+      "step": 6287
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.580043859649123e-06,
+      "loss": 0.9646,
+      "step": 6288
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.579495614035088e-06,
+      "loss": 0.3384,
+      "step": 6289
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.578947368421054e-06,
+      "loss": 1.019,
+      "step": 6290
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.578399122807018e-06,
+      "loss": 1.3942,
+      "step": 6291
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.577850877192983e-06,
+      "loss": 0.865,
+      "step": 6292
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.5773026315789475e-06,
+      "loss": 0.3342,
+      "step": 6293
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.576754385964913e-06,
+      "loss": 0.7557,
+      "step": 6294
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.576206140350878e-06,
+      "loss": 0.8839,
+      "step": 6295
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.575657894736842e-06,
+      "loss": 0.7516,
+      "step": 6296
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.575109649122808e-06,
+      "loss": 0.6488,
+      "step": 6297
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.5745614035087724e-06,
+      "loss": 1.054,
+      "step": 6298
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.574013157894737e-06,
+      "loss": 0.722,
+      "step": 6299
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.573464912280703e-06,
+      "loss": 1.0256,
+      "step": 6300
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.572916666666668e-06,
+      "loss": 0.929,
+      "step": 6301
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.572368421052632e-06,
+      "loss": 0.6507,
+      "step": 6302
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.5718201754385966e-06,
+      "loss": 0.4338,
+      "step": 6303
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.571271929824562e-06,
+      "loss": 0.3175,
+      "step": 6304
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.570723684210527e-06,
+      "loss": 0.4519,
+      "step": 6305
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.570175438596492e-06,
+      "loss": 1.2134,
+      "step": 6306
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.569627192982456e-06,
+      "loss": 0.6151,
+      "step": 6307
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.5690789473684215e-06,
+      "loss": 0.5241,
+      "step": 6308
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.568530701754386e-06,
+      "loss": 0.9341,
+      "step": 6309
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.567982456140351e-06,
+      "loss": 1.1293,
+      "step": 6310
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.567434210526317e-06,
+      "loss": 0.9068,
+      "step": 6311
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.566885964912281e-06,
+      "loss": 0.6983,
+      "step": 6312
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.566337719298246e-06,
+      "loss": 0.9063,
+      "step": 6313
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.565789473684211e-06,
+      "loss": 0.573,
+      "step": 6314
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.565241228070176e-06,
+      "loss": 0.7259,
+      "step": 6315
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.564692982456141e-06,
+      "loss": 0.4999,
+      "step": 6316
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.564144736842107e-06,
+      "loss": 0.7722,
+      "step": 6317
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.563596491228071e-06,
+      "loss": 0.9807,
+      "step": 6318
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.5630482456140354e-06,
+      "loss": 0.9993,
+      "step": 6319
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.5625e-06,
+      "loss": 0.799,
+      "step": 6320
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.561951754385966e-06,
+      "loss": 0.7972,
+      "step": 6321
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.561403508771931e-06,
+      "loss": 1.3264,
+      "step": 6322
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.560855263157895e-06,
+      "loss": 0.4888,
+      "step": 6323
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.5603070175438596e-06,
+      "loss": 0.8822,
+      "step": 6324
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.559758771929825e-06,
+      "loss": 1.4521,
+      "step": 6325
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.55921052631579e-06,
+      "loss": 1.6745,
+      "step": 6326
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.558662280701755e-06,
+      "loss": 1.648,
+      "step": 6327
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.5581140350877206e-06,
+      "loss": 1.2695,
+      "step": 6328
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.5575657894736845e-06,
+      "loss": 0.857,
+      "step": 6329
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.557017543859649e-06,
+      "loss": 0.6396,
+      "step": 6330
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.556469298245615e-06,
+      "loss": 1.0942,
+      "step": 6331
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.55592105263158e-06,
+      "loss": 0.4822,
+      "step": 6332
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.555372807017545e-06,
+      "loss": 1.3272,
+      "step": 6333
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.554824561403509e-06,
+      "loss": 1.1734,
+      "step": 6334
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.554276315789474e-06,
+      "loss": 1.4155,
+      "step": 6335
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.553728070175439e-06,
+      "loss": 0.6791,
+      "step": 6336
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.553179824561404e-06,
+      "loss": 1.1928,
+      "step": 6337
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.55263157894737e-06,
+      "loss": 0.773,
+      "step": 6338
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.552083333333334e-06,
+      "loss": 0.4468,
+      "step": 6339
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.5515350877192985e-06,
+      "loss": 0.6585,
+      "step": 6340
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.550986842105263e-06,
+      "loss": 0.9304,
+      "step": 6341
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.550438596491229e-06,
+      "loss": 0.566,
+      "step": 6342
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.549890350877194e-06,
+      "loss": 0.7675,
+      "step": 6343
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.549342105263159e-06,
+      "loss": 0.4945,
+      "step": 6344
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.548793859649123e-06,
+      "loss": 0.4896,
+      "step": 6345
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.548245614035088e-06,
+      "loss": 0.788,
+      "step": 6346
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.547697368421053e-06,
+      "loss": 0.4576,
+      "step": 6347
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.547149122807018e-06,
+      "loss": 0.6533,
+      "step": 6348
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.5466008771929836e-06,
+      "loss": 0.8633,
+      "step": 6349
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.5460526315789476e-06,
+      "loss": 0.5546,
+      "step": 6350
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.545504385964912e-06,
+      "loss": 1.4738,
+      "step": 6351
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.544956140350878e-06,
+      "loss": 0.6549,
+      "step": 6352
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.544407894736843e-06,
+      "loss": 0.6376,
+      "step": 6353
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.543859649122808e-06,
+      "loss": 0.9739,
+      "step": 6354
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.543311403508772e-06,
+      "loss": 0.5769,
+      "step": 6355
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.542763157894737e-06,
+      "loss": 0.3332,
+      "step": 6356
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.542214912280702e-06,
+      "loss": 0.5241,
+      "step": 6357
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.541666666666667e-06,
+      "loss": 0.5565,
+      "step": 6358
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.541118421052633e-06,
+      "loss": 1.0406,
+      "step": 6359
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.5405701754385975e-06,
+      "loss": 0.6922,
+      "step": 6360
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.5400219298245615e-06,
+      "loss": 0.6443,
+      "step": 6361
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.539473684210526e-06,
+      "loss": 0.5569,
+      "step": 6362
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.538925438596492e-06,
+      "loss": 0.4021,
+      "step": 6363
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.538377192982457e-06,
+      "loss": 0.4032,
+      "step": 6364
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.537828947368422e-06,
+      "loss": 1.2675,
+      "step": 6365
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.537280701754386e-06,
+      "loss": 0.9382,
+      "step": 6366
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.536732456140351e-06,
+      "loss": 0.8379,
+      "step": 6367
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.536184210526316e-06,
+      "loss": 1.5949,
+      "step": 6368
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.535635964912282e-06,
+      "loss": 1.1173,
+      "step": 6369
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.535087719298247e-06,
+      "loss": 0.8783,
+      "step": 6370
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.534539473684211e-06,
+      "loss": 0.2859,
+      "step": 6371
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.533991228070175e-06,
+      "loss": 0.9951,
+      "step": 6372
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.533442982456141e-06,
+      "loss": 0.299,
+      "step": 6373
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.532894736842106e-06,
+      "loss": 0.5832,
+      "step": 6374
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.532346491228071e-06,
+      "loss": 0.6138,
+      "step": 6375
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.531798245614036e-06,
+      "loss": 1.2072,
+      "step": 6376
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.53125e-06,
+      "loss": 0.4805,
+      "step": 6377
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.530701754385965e-06,
+      "loss": 0.6592,
+      "step": 6378
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.53015350877193e-06,
+      "loss": 0.3396,
+      "step": 6379
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.529605263157896e-06,
+      "loss": 0.8049,
+      "step": 6380
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.5290570175438605e-06,
+      "loss": 0.8075,
+      "step": 6381
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.5285087719298245e-06,
+      "loss": 0.3249,
+      "step": 6382
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.527960526315789e-06,
+      "loss": 0.7027,
+      "step": 6383
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.527412280701755e-06,
+      "loss": 0.3693,
+      "step": 6384
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.52686403508772e-06,
+      "loss": 0.6541,
+      "step": 6385
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.526315789473685e-06,
+      "loss": 0.4724,
+      "step": 6386
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.52576754385965e-06,
+      "loss": 1.132,
+      "step": 6387
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.525219298245614e-06,
+      "loss": 0.8536,
+      "step": 6388
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.524671052631579e-06,
+      "loss": 0.5935,
+      "step": 6389
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.524122807017545e-06,
+      "loss": 0.6827,
+      "step": 6390
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.52357456140351e-06,
+      "loss": 0.4353,
+      "step": 6391
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.523026315789474e-06,
+      "loss": 0.5792,
+      "step": 6392
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.522478070175438e-06,
+      "loss": 0.3194,
+      "step": 6393
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.521929824561404e-06,
+      "loss": 1.2671,
+      "step": 6394
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.521381578947369e-06,
+      "loss": 0.5574,
+      "step": 6395
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.520833333333334e-06,
+      "loss": 0.5379,
+      "step": 6396
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.520285087719299e-06,
+      "loss": 1.0808,
+      "step": 6397
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.519736842105263e-06,
+      "loss": 0.4504,
+      "step": 6398
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.519188596491228e-06,
+      "loss": 0.7691,
+      "step": 6399
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.518640350877193e-06,
+      "loss": 0.7219,
+      "step": 6400
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.518092105263159e-06,
+      "loss": 2.5401,
+      "step": 6401
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.5175438596491235e-06,
+      "loss": 0.3785,
+      "step": 6402
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.516995614035088e-06,
+      "loss": 0.7098,
+      "step": 6403
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.516447368421053e-06,
+      "loss": 0.5598,
+      "step": 6404
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.515899122807018e-06,
+      "loss": 0.4527,
+      "step": 6405
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.515350877192983e-06,
+      "loss": 0.8279,
+      "step": 6406
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.5148026315789485e-06,
+      "loss": 0.9551,
+      "step": 6407
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.514254385964913e-06,
+      "loss": 1.8555,
+      "step": 6408
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.513706140350877e-06,
+      "loss": 0.805,
+      "step": 6409
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.513157894736842e-06,
+      "loss": 0.3816,
+      "step": 6410
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.512609649122808e-06,
+      "loss": 1.4035,
+      "step": 6411
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.512061403508773e-06,
+      "loss": 0.3041,
+      "step": 6412
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.5115131578947374e-06,
+      "loss": 0.9421,
+      "step": 6413
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.510964912280703e-06,
+      "loss": 1.1607,
+      "step": 6414
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.510416666666667e-06,
+      "loss": 1.0064,
+      "step": 6415
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.509868421052632e-06,
+      "loss": 1.0443,
+      "step": 6416
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.509320175438597e-06,
+      "loss": 0.8751,
+      "step": 6417
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.508771929824562e-06,
+      "loss": 0.498,
+      "step": 6418
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.508223684210527e-06,
+      "loss": 0.7498,
+      "step": 6419
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.507675438596491e-06,
+      "loss": 1.27,
+      "step": 6420
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.507127192982456e-06,
+      "loss": 0.6414,
+      "step": 6421
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.506578947368422e-06,
+      "loss": 1.1176,
+      "step": 6422
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.5060307017543865e-06,
+      "loss": 1.2811,
+      "step": 6423
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.505482456140352e-06,
+      "loss": 0.6172,
+      "step": 6424
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.504934210526316e-06,
+      "loss": 1.2362,
+      "step": 6425
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.504385964912281e-06,
+      "loss": 1.3446,
+      "step": 6426
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.503837719298246e-06,
+      "loss": 0.4782,
+      "step": 6427
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.5032894736842115e-06,
+      "loss": 1.1161,
+      "step": 6428
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.502741228070176e-06,
+      "loss": 1.0193,
+      "step": 6429
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.502192982456141e-06,
+      "loss": 0.8275,
+      "step": 6430
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.501644736842105e-06,
+      "loss": 1.0375,
+      "step": 6431
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.501096491228071e-06,
+      "loss": 0.5422,
+      "step": 6432
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.500548245614036e-06,
+      "loss": 0.3918,
+      "step": 6433
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.5000000000000004e-06,
+      "loss": 0.9175,
+      "step": 6434
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.499451754385966e-06,
+      "loss": 0.5362,
+      "step": 6435
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.49890350877193e-06,
+      "loss": 1.0582,
+      "step": 6436
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.498355263157895e-06,
+      "loss": 0.7736,
+      "step": 6437
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.49780701754386e-06,
+      "loss": 1.0357,
+      "step": 6438
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.497258771929825e-06,
+      "loss": 0.6766,
+      "step": 6439
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.49671052631579e-06,
+      "loss": 0.7401,
+      "step": 6440
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.496162280701754e-06,
+      "loss": 0.7666,
+      "step": 6441
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.49561403508772e-06,
+      "loss": 0.595,
+      "step": 6442
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.495065789473685e-06,
+      "loss": 0.9226,
+      "step": 6443
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.4945175438596495e-06,
+      "loss": 1.1197,
+      "step": 6444
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.493969298245615e-06,
+      "loss": 0.4762,
+      "step": 6445
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.49342105263158e-06,
+      "loss": 0.9774,
+      "step": 6446
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.492872807017544e-06,
+      "loss": 0.916,
+      "step": 6447
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.492324561403509e-06,
+      "loss": 1.1259,
+      "step": 6448
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.4917763157894745e-06,
+      "loss": 0.9523,
+      "step": 6449
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.491228070175439e-06,
+      "loss": 0.7229,
+      "step": 6450
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.490679824561404e-06,
+      "loss": 0.3472,
+      "step": 6451
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.490131578947368e-06,
+      "loss": 1.3841,
+      "step": 6452
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.489583333333334e-06,
+      "loss": 0.6303,
+      "step": 6453
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.489035087719299e-06,
+      "loss": 0.7004,
+      "step": 6454
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.4884868421052635e-06,
+      "loss": 1.3324,
+      "step": 6455
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.487938596491229e-06,
+      "loss": 0.5479,
+      "step": 6456
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.487390350877194e-06,
+      "loss": 0.4899,
+      "step": 6457
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.486842105263158e-06,
+      "loss": 0.5323,
+      "step": 6458
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.486293859649123e-06,
+      "loss": 1.0746,
+      "step": 6459
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.4857456140350884e-06,
+      "loss": 0.5655,
+      "step": 6460
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.485197368421053e-06,
+      "loss": 0.4853,
+      "step": 6461
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.484649122807019e-06,
+      "loss": 0.9192,
+      "step": 6462
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.484100877192983e-06,
+      "loss": 0.6466,
+      "step": 6463
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.483552631578948e-06,
+      "loss": 0.5843,
+      "step": 6464
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.4830043859649126e-06,
+      "loss": 1.0014,
+      "step": 6465
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.482456140350878e-06,
+      "loss": 0.5061,
+      "step": 6466
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.481907894736843e-06,
+      "loss": 0.6635,
+      "step": 6467
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.481359649122807e-06,
+      "loss": 1.1568,
+      "step": 6468
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.480811403508772e-06,
+      "loss": 0.4964,
+      "step": 6469
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.4802631578947375e-06,
+      "loss": 0.4859,
+      "step": 6470
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.479714912280702e-06,
+      "loss": 0.7181,
+      "step": 6471
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.479166666666667e-06,
+      "loss": 0.6701,
+      "step": 6472
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.478618421052633e-06,
+      "loss": 0.6946,
+      "step": 6473
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.478070175438597e-06,
+      "loss": 0.4075,
+      "step": 6474
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.477521929824562e-06,
+      "loss": 1.2106,
+      "step": 6475
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.4769736842105265e-06,
+      "loss": 1.2996,
+      "step": 6476
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.476425438596492e-06,
+      "loss": 0.4795,
+      "step": 6477
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.475877192982457e-06,
+      "loss": 0.3347,
+      "step": 6478
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.475328947368421e-06,
+      "loss": 0.8599,
+      "step": 6479
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.474780701754387e-06,
+      "loss": 0.5532,
+      "step": 6480
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.4742324561403514e-06,
+      "loss": 1.3863,
+      "step": 6481
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.473684210526316e-06,
+      "loss": 1.0025,
+      "step": 6482
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.473135964912282e-06,
+      "loss": 0.576,
+      "step": 6483
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.472587719298247e-06,
+      "loss": 0.7439,
+      "step": 6484
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.472039473684211e-06,
+      "loss": 0.4929,
+      "step": 6485
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.4714912280701756e-06,
+      "loss": 0.6366,
+      "step": 6486
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.470942982456141e-06,
+      "loss": 1.1478,
+      "step": 6487
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.470394736842106e-06,
+      "loss": 0.532,
+      "step": 6488
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.469846491228071e-06,
+      "loss": 0.9671,
+      "step": 6489
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.469298245614035e-06,
+      "loss": 0.6636,
+      "step": 6490
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.4687500000000005e-06,
+      "loss": 0.9512,
+      "step": 6491
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.468201754385965e-06,
+      "loss": 1.3846,
+      "step": 6492
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.46765350877193e-06,
+      "loss": 1.3279,
+      "step": 6493
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.467105263157896e-06,
+      "loss": 0.7906,
+      "step": 6494
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.46655701754386e-06,
+      "loss": 0.5615,
+      "step": 6495
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.466008771929825e-06,
+      "loss": 1.1653,
+      "step": 6496
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.46546052631579e-06,
+      "loss": 0.6291,
+      "step": 6497
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.464912280701755e-06,
+      "loss": 0.2474,
+      "step": 6498
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.46436403508772e-06,
+      "loss": 1.4869,
+      "step": 6499
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.463815789473686e-06,
+      "loss": 1.4496,
+      "step": 6500
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.46326754385965e-06,
+      "loss": 0.4063,
+      "step": 6501
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.4627192982456145e-06,
+      "loss": 0.7324,
+      "step": 6502
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.462171052631579e-06,
+      "loss": 1.1296,
+      "step": 6503
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.461622807017545e-06,
+      "loss": 1.1702,
+      "step": 6504
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.46107456140351e-06,
+      "loss": 0.6624,
+      "step": 6505
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.460526315789474e-06,
+      "loss": 1.132,
+      "step": 6506
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.4599780701754386e-06,
+      "loss": 0.4042,
+      "step": 6507
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.459429824561404e-06,
+      "loss": 0.538,
+      "step": 6508
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.458881578947369e-06,
+      "loss": 0.6598,
+      "step": 6509
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.458333333333334e-06,
+      "loss": 0.7283,
+      "step": 6510
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.457785087719298e-06,
+      "loss": 1.2147,
+      "step": 6511
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.4572368421052635e-06,
+      "loss": 1.2492,
+      "step": 6512
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.456688596491228e-06,
+      "loss": 1.0462,
+      "step": 6513
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.456140350877193e-06,
+      "loss": 0.5944,
+      "step": 6514
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.455592105263159e-06,
+      "loss": 1.0082,
+      "step": 6515
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.455043859649124e-06,
+      "loss": 0.9509,
+      "step": 6516
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.454495614035088e-06,
+      "loss": 0.3041,
+      "step": 6517
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.453947368421053e-06,
+      "loss": 1.183,
+      "step": 6518
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.453399122807018e-06,
+      "loss": 0.6811,
+      "step": 6519
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.452850877192983e-06,
+      "loss": 0.1722,
+      "step": 6520
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.452302631578949e-06,
+      "loss": 0.7997,
+      "step": 6521
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.451754385964913e-06,
+      "loss": 0.7198,
+      "step": 6522
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.4512061403508775e-06,
+      "loss": 1.2785,
+      "step": 6523
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.450657894736842e-06,
+      "loss": 0.8622,
+      "step": 6524
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.450109649122808e-06,
+      "loss": 0.6503,
+      "step": 6525
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.449561403508773e-06,
+      "loss": 1.4036,
+      "step": 6526
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.449013157894738e-06,
+      "loss": 0.922,
+      "step": 6527
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.448464912280702e-06,
+      "loss": 0.4459,
+      "step": 6528
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.447916666666667e-06,
+      "loss": 0.5264,
+      "step": 6529
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.447368421052632e-06,
+      "loss": 0.4056,
+      "step": 6530
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.446820175438597e-06,
+      "loss": 0.6991,
+      "step": 6531
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.446271929824563e-06,
+      "loss": 0.2584,
+      "step": 6532
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.4457236842105266e-06,
+      "loss": 0.6105,
+      "step": 6533
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.445175438596491e-06,
+      "loss": 0.4622,
+      "step": 6534
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.444627192982457e-06,
+      "loss": 0.5181,
+      "step": 6535
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.444078947368422e-06,
+      "loss": 0.698,
+      "step": 6536
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.443530701754387e-06,
+      "loss": 0.4967,
+      "step": 6537
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.442982456140351e-06,
+      "loss": 0.7716,
+      "step": 6538
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.442434210526316e-06,
+      "loss": 0.8948,
+      "step": 6539
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.441885964912281e-06,
+      "loss": 1.2478,
+      "step": 6540
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.441337719298246e-06,
+      "loss": 0.5404,
+      "step": 6541
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.440789473684212e-06,
+      "loss": 1.1236,
+      "step": 6542
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.4402412280701765e-06,
+      "loss": 0.4661,
+      "step": 6543
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.4396929824561405e-06,
+      "loss": 0.4928,
+      "step": 6544
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.439144736842105e-06,
+      "loss": 1.5538,
+      "step": 6545
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.438596491228071e-06,
+      "loss": 0.9734,
+      "step": 6546
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.438048245614036e-06,
+      "loss": 0.4984,
+      "step": 6547
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.437500000000001e-06,
+      "loss": 0.5332,
+      "step": 6548
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.436951754385965e-06,
+      "loss": 0.9489,
+      "step": 6549
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.43640350877193e-06,
+      "loss": 0.9539,
+      "step": 6550
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.435855263157895e-06,
+      "loss": 0.6086,
+      "step": 6551
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.43530701754386e-06,
+      "loss": 0.5215,
+      "step": 6552
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.434758771929826e-06,
+      "loss": 0.8667,
+      "step": 6553
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.4342105263157896e-06,
+      "loss": 0.4169,
+      "step": 6554
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.433662280701754e-06,
+      "loss": 0.576,
+      "step": 6555
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.43311403508772e-06,
+      "loss": 1.1906,
+      "step": 6556
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.432565789473685e-06,
+      "loss": 0.9142,
+      "step": 6557
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.43201754385965e-06,
+      "loss": 0.8555,
+      "step": 6558
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.431469298245615e-06,
+      "loss": 1.0642,
+      "step": 6559
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.430921052631579e-06,
+      "loss": 0.7738,
+      "step": 6560
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.430372807017544e-06,
+      "loss": 0.6404,
+      "step": 6561
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.429824561403509e-06,
+      "loss": 1.0155,
+      "step": 6562
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.429276315789475e-06,
+      "loss": 0.2389,
+      "step": 6563
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.4287280701754395e-06,
+      "loss": 0.4409,
+      "step": 6564
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.4281798245614035e-06,
+      "loss": 0.3698,
+      "step": 6565
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.427631578947368e-06,
+      "loss": 0.4281,
+      "step": 6566
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.427083333333334e-06,
+      "loss": 1.0596,
+      "step": 6567
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.426535087719299e-06,
+      "loss": 1.0608,
+      "step": 6568
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.425986842105264e-06,
+      "loss": 1.4032,
+      "step": 6569
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.425438596491229e-06,
+      "loss": 1.1635,
+      "step": 6570
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.424890350877193e-06,
+      "loss": 1.2351,
+      "step": 6571
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.424342105263158e-06,
+      "loss": 0.9432,
+      "step": 6572
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.423793859649124e-06,
+      "loss": 0.497,
+      "step": 6573
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.423245614035089e-06,
+      "loss": 0.6613,
+      "step": 6574
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.4226973684210534e-06,
+      "loss": 0.7373,
+      "step": 6575
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.422149122807017e-06,
+      "loss": 0.4149,
+      "step": 6576
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.421600877192983e-06,
+      "loss": 1.3549,
+      "step": 6577
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.421052631578948e-06,
+      "loss": 0.9774,
+      "step": 6578
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.420504385964913e-06,
+      "loss": 0.7717,
+      "step": 6579
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.419956140350878e-06,
+      "loss": 0.2654,
+      "step": 6580
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.419407894736842e-06,
+      "loss": 1.0602,
+      "step": 6581
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.418859649122807e-06,
+      "loss": 0.8449,
+      "step": 6582
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.418311403508772e-06,
+      "loss": 1.1967,
+      "step": 6583
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.417763157894738e-06,
+      "loss": 0.6282,
+      "step": 6584
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.4172149122807025e-06,
+      "loss": 0.6116,
+      "step": 6585
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.416666666666667e-06,
+      "loss": 0.6548,
+      "step": 6586
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.416118421052631e-06,
+      "loss": 0.6411,
+      "step": 6587
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.415570175438597e-06,
+      "loss": 0.6847,
+      "step": 6588
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.415021929824562e-06,
+      "loss": 1.5417,
+      "step": 6589
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.4144736842105275e-06,
+      "loss": 0.9434,
+      "step": 6590
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.413925438596492e-06,
+      "loss": 0.6081,
+      "step": 6591
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.413377192982456e-06,
+      "loss": 0.8349,
+      "step": 6592
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.412828947368421e-06,
+      "loss": 0.5392,
+      "step": 6593
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.412280701754387e-06,
+      "loss": 1.1644,
+      "step": 6594
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.411732456140352e-06,
+      "loss": 0.9358,
+      "step": 6595
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.4111842105263164e-06,
+      "loss": 1.0758,
+      "step": 6596
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.41063596491228e-06,
+      "loss": 0.6007,
+      "step": 6597
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.410087719298246e-06,
+      "loss": 0.6097,
+      "step": 6598
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.409539473684211e-06,
+      "loss": 0.8327,
+      "step": 6599
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.408991228070176e-06,
+      "loss": 0.6765,
+      "step": 6600
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.408442982456141e-06,
+      "loss": 0.7467,
+      "step": 6601
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.407894736842106e-06,
+      "loss": 0.9834,
+      "step": 6602
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.40734649122807e-06,
+      "loss": 1.1,
+      "step": 6603
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.406798245614035e-06,
+      "loss": 1.7884,
+      "step": 6604
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.406250000000001e-06,
+      "loss": 0.7225,
+      "step": 6605
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.4057017543859655e-06,
+      "loss": 1.045,
+      "step": 6606
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.40515350877193e-06,
+      "loss": 0.603,
+      "step": 6607
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.404605263157895e-06,
+      "loss": 0.843,
+      "step": 6608
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.40405701754386e-06,
+      "loss": 0.84,
+      "step": 6609
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.403508771929825e-06,
+      "loss": 0.6267,
+      "step": 6610
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.4029605263157905e-06,
+      "loss": 0.5061,
+      "step": 6611
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.402412280701755e-06,
+      "loss": 0.3247,
+      "step": 6612
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.40186403508772e-06,
+      "loss": 0.5729,
+      "step": 6613
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.401315789473684e-06,
+      "loss": 0.432,
+      "step": 6614
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.40076754385965e-06,
+      "loss": 0.7783,
+      "step": 6615
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.400219298245615e-06,
+      "loss": 0.8998,
+      "step": 6616
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.3996710526315795e-06,
+      "loss": 1.2773,
+      "step": 6617
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.399122807017545e-06,
+      "loss": 0.5839,
+      "step": 6618
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.398574561403509e-06,
+      "loss": 0.2664,
+      "step": 6619
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.398026315789474e-06,
+      "loss": 0.8026,
+      "step": 6620
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.397478070175439e-06,
+      "loss": 1.3044,
+      "step": 6621
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.396929824561404e-06,
+      "loss": 1.1819,
+      "step": 6622
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.396381578947369e-06,
+      "loss": 0.9006,
+      "step": 6623
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.395833333333333e-06,
+      "loss": 1.4325,
+      "step": 6624
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.395285087719298e-06,
+      "loss": 0.4006,
+      "step": 6625
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.394736842105264e-06,
+      "loss": 0.8379,
+      "step": 6626
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.3941885964912285e-06,
+      "loss": 0.5186,
+      "step": 6627
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.393640350877194e-06,
+      "loss": 1.2148,
+      "step": 6628
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.393092105263159e-06,
+      "loss": 0.952,
+      "step": 6629
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.392543859649123e-06,
+      "loss": 0.8005,
+      "step": 6630
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.391995614035088e-06,
+      "loss": 0.8314,
+      "step": 6631
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.3914473684210535e-06,
+      "loss": 0.5672,
+      "step": 6632
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.390899122807018e-06,
+      "loss": 1.139,
+      "step": 6633
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.390350877192983e-06,
+      "loss": 0.726,
+      "step": 6634
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.389802631578947e-06,
+      "loss": 0.6355,
+      "step": 6635
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.389254385964913e-06,
+      "loss": 1.0421,
+      "step": 6636
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.388706140350878e-06,
+      "loss": 0.783,
+      "step": 6637
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.3881578947368425e-06,
+      "loss": 1.181,
+      "step": 6638
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.387609649122808e-06,
+      "loss": 0.6283,
+      "step": 6639
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.387061403508772e-06,
+      "loss": 0.7846,
+      "step": 6640
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.386513157894737e-06,
+      "loss": 0.9409,
+      "step": 6641
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.385964912280702e-06,
+      "loss": 1.4279,
+      "step": 6642
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.3854166666666674e-06,
+      "loss": 0.8356,
+      "step": 6643
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.384868421052632e-06,
+      "loss": 0.3845,
+      "step": 6644
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.384320175438597e-06,
+      "loss": 1.1485,
+      "step": 6645
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.383771929824562e-06,
+      "loss": 0.7247,
+      "step": 6646
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.383223684210527e-06,
+      "loss": 0.4946,
+      "step": 6647
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.3826754385964916e-06,
+      "loss": 0.8565,
+      "step": 6648
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.382127192982457e-06,
+      "loss": 0.8634,
+      "step": 6649
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.381578947368422e-06,
+      "loss": 0.6865,
+      "step": 6650
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.381030701754386e-06,
+      "loss": 1.2721,
+      "step": 6651
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.380482456140351e-06,
+      "loss": 1.0278,
+      "step": 6652
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.3799342105263165e-06,
+      "loss": 1.0988,
+      "step": 6653
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.379385964912281e-06,
+      "loss": 0.8737,
+      "step": 6654
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.378837719298246e-06,
+      "loss": 0.7303,
+      "step": 6655
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.378289473684212e-06,
+      "loss": 0.3521,
+      "step": 6656
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.377741228070176e-06,
+      "loss": 1.008,
+      "step": 6657
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.377192982456141e-06,
+      "loss": 0.3657,
+      "step": 6658
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.3766447368421055e-06,
+      "loss": 0.9232,
+      "step": 6659
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.376096491228071e-06,
+      "loss": 0.4623,
+      "step": 6660
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.375548245614036e-06,
+      "loss": 0.2628,
+      "step": 6661
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.375e-06,
+      "loss": 0.5488,
+      "step": 6662
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.374451754385965e-06,
+      "loss": 0.4232,
+      "step": 6663
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.3739035087719304e-06,
+      "loss": 0.9146,
+      "step": 6664
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.373355263157895e-06,
+      "loss": 0.5944,
+      "step": 6665
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.372807017543861e-06,
+      "loss": 0.7042,
+      "step": 6666
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.372258771929825e-06,
+      "loss": 0.9954,
+      "step": 6667
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.37171052631579e-06,
+      "loss": 0.5975,
+      "step": 6668
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.3711622807017546e-06,
+      "loss": 0.8226,
+      "step": 6669
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.37061403508772e-06,
+      "loss": 1.1646,
+      "step": 6670
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.370065789473685e-06,
+      "loss": 0.2407,
+      "step": 6671
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.36951754385965e-06,
+      "loss": 0.3372,
+      "step": 6672
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.368969298245614e-06,
+      "loss": 0.4336,
+      "step": 6673
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.3684210526315795e-06,
+      "loss": 0.4898,
+      "step": 6674
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.367872807017544e-06,
+      "loss": 0.6943,
+      "step": 6675
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.367324561403509e-06,
+      "loss": 0.92,
+      "step": 6676
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.366776315789475e-06,
+      "loss": 0.4082,
+      "step": 6677
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.366228070175439e-06,
+      "loss": 0.6568,
+      "step": 6678
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.365679824561404e-06,
+      "loss": 1.0592,
+      "step": 6679
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.3651315789473685e-06,
+      "loss": 0.7215,
+      "step": 6680
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.364583333333334e-06,
+      "loss": 2.0639,
+      "step": 6681
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.364035087719299e-06,
+      "loss": 0.3328,
+      "step": 6682
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.363486842105263e-06,
+      "loss": 1.023,
+      "step": 6683
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.362938596491229e-06,
+      "loss": 0.8856,
+      "step": 6684
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.3623903508771935e-06,
+      "loss": 1.1729,
+      "step": 6685
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.361842105263158e-06,
+      "loss": 0.6554,
+      "step": 6686
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.361293859649124e-06,
+      "loss": 1.1575,
+      "step": 6687
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.360745614035089e-06,
+      "loss": 0.7284,
+      "step": 6688
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.360197368421053e-06,
+      "loss": 0.3306,
+      "step": 6689
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.359649122807018e-06,
+      "loss": 1.3584,
+      "step": 6690
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.359100877192983e-06,
+      "loss": 1.1299,
+      "step": 6691
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.358552631578948e-06,
+      "loss": 0.2994,
+      "step": 6692
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.358004385964913e-06,
+      "loss": 0.674,
+      "step": 6693
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.357456140350877e-06,
+      "loss": 0.6681,
+      "step": 6694
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.3569078947368426e-06,
+      "loss": 0.6449,
+      "step": 6695
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.356359649122807e-06,
+      "loss": 1.4407,
+      "step": 6696
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.355811403508772e-06,
+      "loss": 0.7368,
+      "step": 6697
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.355263157894738e-06,
+      "loss": 0.5339,
+      "step": 6698
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.354714912280703e-06,
+      "loss": 0.517,
+      "step": 6699
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.354166666666667e-06,
+      "loss": 1.0332,
+      "step": 6700
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.353618421052632e-06,
+      "loss": 0.6759,
+      "step": 6701
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.353070175438597e-06,
+      "loss": 0.4164,
+      "step": 6702
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.352521929824562e-06,
+      "loss": 0.3911,
+      "step": 6703
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.351973684210528e-06,
+      "loss": 0.4195,
+      "step": 6704
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.351425438596492e-06,
+      "loss": 1.7792,
+      "step": 6705
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.3508771929824565e-06,
+      "loss": 0.9717,
+      "step": 6706
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.350328947368421e-06,
+      "loss": 0.6206,
+      "step": 6707
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.349780701754387e-06,
+      "loss": 0.7943,
+      "step": 6708
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.349232456140352e-06,
+      "loss": 0.7674,
+      "step": 6709
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.348684210526316e-06,
+      "loss": 0.7575,
+      "step": 6710
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.348135964912281e-06,
+      "loss": 0.7558,
+      "step": 6711
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.347587719298246e-06,
+      "loss": 0.3105,
+      "step": 6712
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.347039473684211e-06,
+      "loss": 0.6117,
+      "step": 6713
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.346491228070176e-06,
+      "loss": 0.866,
+      "step": 6714
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.345942982456142e-06,
+      "loss": 0.5179,
+      "step": 6715
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.3453947368421056e-06,
+      "loss": 1.0875,
+      "step": 6716
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.34484649122807e-06,
+      "loss": 0.5233,
+      "step": 6717
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.344298245614035e-06,
+      "loss": 0.7256,
+      "step": 6718
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.343750000000001e-06,
+      "loss": 0.4276,
+      "step": 6719
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.343201754385966e-06,
+      "loss": 0.3042,
+      "step": 6720
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.34265350877193e-06,
+      "loss": 0.5504,
+      "step": 6721
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.342105263157895e-06,
+      "loss": 0.3624,
+      "step": 6722
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.34155701754386e-06,
+      "loss": 0.5152,
+      "step": 6723
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.341008771929825e-06,
+      "loss": 1.6324,
+      "step": 6724
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.340460526315791e-06,
+      "loss": 0.9646,
+      "step": 6725
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.339912280701755e-06,
+      "loss": 1.3211,
+      "step": 6726
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.3393640350877195e-06,
+      "loss": 0.3916,
+      "step": 6727
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.338815789473684e-06,
+      "loss": 0.3685,
+      "step": 6728
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.33826754385965e-06,
+      "loss": 1.2922,
+      "step": 6729
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.337719298245615e-06,
+      "loss": 0.4855,
+      "step": 6730
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.33717105263158e-06,
+      "loss": 0.2274,
+      "step": 6731
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.336622807017544e-06,
+      "loss": 0.7113,
+      "step": 6732
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.336074561403509e-06,
+      "loss": 1.1175,
+      "step": 6733
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.335526315789474e-06,
+      "loss": 0.8632,
+      "step": 6734
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.334978070175439e-06,
+      "loss": 1.1217,
+      "step": 6735
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.334429824561405e-06,
+      "loss": 0.2338,
+      "step": 6736
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.3338815789473686e-06,
+      "loss": 0.2408,
+      "step": 6737
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.333333333333333e-06,
+      "loss": 0.6211,
+      "step": 6738
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.332785087719299e-06,
+      "loss": 0.8869,
+      "step": 6739
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.332236842105264e-06,
+      "loss": 0.138,
+      "step": 6740
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.331688596491229e-06,
+      "loss": 0.7751,
+      "step": 6741
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.331140350877194e-06,
+      "loss": 0.5847,
+      "step": 6742
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.330592105263158e-06,
+      "loss": 0.5577,
+      "step": 6743
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.330043859649123e-06,
+      "loss": 1.3783,
+      "step": 6744
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.329495614035088e-06,
+      "loss": 0.9046,
+      "step": 6745
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.328947368421054e-06,
+      "loss": 0.7746,
+      "step": 6746
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.3283991228070185e-06,
+      "loss": 0.3718,
+      "step": 6747
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.3278508771929825e-06,
+      "loss": 0.4592,
+      "step": 6748
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.327302631578947e-06,
+      "loss": 1.2757,
+      "step": 6749
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.326754385964913e-06,
+      "loss": 1.0133,
+      "step": 6750
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.326206140350878e-06,
+      "loss": 1.0154,
+      "step": 6751
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.325657894736843e-06,
+      "loss": 1.3619,
+      "step": 6752
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.325109649122807e-06,
+      "loss": 1.4541,
+      "step": 6753
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.324561403508772e-06,
+      "loss": 1.0722,
+      "step": 6754
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.324013157894737e-06,
+      "loss": 0.642,
+      "step": 6755
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.323464912280703e-06,
+      "loss": 1.0462,
+      "step": 6756
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.322916666666668e-06,
+      "loss": 0.9822,
+      "step": 6757
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.3223684210526324e-06,
+      "loss": 0.3,
+      "step": 6758
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.321820175438596e-06,
+      "loss": 0.5441,
+      "step": 6759
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.321271929824562e-06,
+      "loss": 0.5251,
+      "step": 6760
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.320723684210527e-06,
+      "loss": 0.7536,
+      "step": 6761
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.320175438596492e-06,
+      "loss": 0.2448,
+      "step": 6762
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.319627192982457e-06,
+      "loss": 0.8453,
+      "step": 6763
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.319078947368421e-06,
+      "loss": 1.2666,
+      "step": 6764
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.318530701754386e-06,
+      "loss": 0.7419,
+      "step": 6765
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.317982456140351e-06,
+      "loss": 1.6699,
+      "step": 6766
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.317434210526317e-06,
+      "loss": 1.0032,
+      "step": 6767
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.3168859649122815e-06,
+      "loss": 1.2506,
+      "step": 6768
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.316337719298246e-06,
+      "loss": 0.8003,
+      "step": 6769
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.31578947368421e-06,
+      "loss": 1.1314,
+      "step": 6770
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.315241228070176e-06,
+      "loss": 0.6188,
+      "step": 6771
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.314692982456141e-06,
+      "loss": 0.834,
+      "step": 6772
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.314144736842106e-06,
+      "loss": 1.0181,
+      "step": 6773
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.313596491228071e-06,
+      "loss": 0.9324,
+      "step": 6774
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.313048245614035e-06,
+      "loss": 0.9737,
+      "step": 6775
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.3125e-06,
+      "loss": 0.671,
+      "step": 6776
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.311951754385966e-06,
+      "loss": 0.8765,
+      "step": 6777
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.311403508771931e-06,
+      "loss": 1.4162,
+      "step": 6778
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.3108552631578954e-06,
+      "loss": 0.2591,
+      "step": 6779
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.310307017543859e-06,
+      "loss": 1.0106,
+      "step": 6780
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.309758771929825e-06,
+      "loss": 0.7522,
+      "step": 6781
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.30921052631579e-06,
+      "loss": 0.9374,
+      "step": 6782
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.308662280701755e-06,
+      "loss": 0.7545,
+      "step": 6783
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.30811403508772e-06,
+      "loss": 0.6289,
+      "step": 6784
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.307565789473685e-06,
+      "loss": 0.4813,
+      "step": 6785
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.307017543859649e-06,
+      "loss": 0.9642,
+      "step": 6786
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.306469298245614e-06,
+      "loss": 0.9009,
+      "step": 6787
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.30592105263158e-06,
+      "loss": 1.9494,
+      "step": 6788
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.3053728070175445e-06,
+      "loss": 1.4491,
+      "step": 6789
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.304824561403509e-06,
+      "loss": 1.2865,
+      "step": 6790
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.304276315789473e-06,
+      "loss": 0.8127,
+      "step": 6791
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.303728070175439e-06,
+      "loss": 0.4957,
+      "step": 6792
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.303179824561404e-06,
+      "loss": 0.3971,
+      "step": 6793
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.3026315789473695e-06,
+      "loss": 0.5025,
+      "step": 6794
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.302083333333334e-06,
+      "loss": 0.2558,
+      "step": 6795
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.301535087719298e-06,
+      "loss": 0.9252,
+      "step": 6796
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.300986842105263e-06,
+      "loss": 0.6563,
+      "step": 6797
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.300438596491229e-06,
+      "loss": 0.3891,
+      "step": 6798
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.299890350877194e-06,
+      "loss": 0.8447,
+      "step": 6799
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.2993421052631585e-06,
+      "loss": 0.6242,
+      "step": 6800
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.298793859649124e-06,
+      "loss": 0.7104,
+      "step": 6801
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.298245614035088e-06,
+      "loss": 1.02,
+      "step": 6802
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.297697368421053e-06,
+      "loss": 1.1624,
+      "step": 6803
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.297149122807018e-06,
+      "loss": 0.9361,
+      "step": 6804
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.2966008771929834e-06,
+      "loss": 0.7797,
+      "step": 6805
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.296052631578948e-06,
+      "loss": 0.8318,
+      "step": 6806
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.295504385964912e-06,
+      "loss": 0.9588,
+      "step": 6807
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.294956140350877e-06,
+      "loss": 0.5254,
+      "step": 6808
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.294407894736843e-06,
+      "loss": 0.6862,
+      "step": 6809
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.2938596491228076e-06,
+      "loss": 0.7117,
+      "step": 6810
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.293311403508772e-06,
+      "loss": 0.6647,
+      "step": 6811
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.292763157894738e-06,
+      "loss": 0.253,
+      "step": 6812
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.292214912280702e-06,
+      "loss": 0.7085,
+      "step": 6813
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.291666666666667e-06,
+      "loss": 0.5618,
+      "step": 6814
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.2911184210526325e-06,
+      "loss": 0.897,
+      "step": 6815
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.290570175438597e-06,
+      "loss": 1.1493,
+      "step": 6816
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.290021929824562e-06,
+      "loss": 1.1132,
+      "step": 6817
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.289473684210526e-06,
+      "loss": 0.8516,
+      "step": 6818
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.288925438596492e-06,
+      "loss": 0.4866,
+      "step": 6819
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.288377192982457e-06,
+      "loss": 0.6969,
+      "step": 6820
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.2878289473684215e-06,
+      "loss": 0.8252,
+      "step": 6821
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.287280701754387e-06,
+      "loss": 0.9271,
+      "step": 6822
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.286732456140351e-06,
+      "loss": 0.5623,
+      "step": 6823
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.286184210526316e-06,
+      "loss": 0.5817,
+      "step": 6824
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.285635964912281e-06,
+      "loss": 1.5333,
+      "step": 6825
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.2850877192982464e-06,
+      "loss": 1.0554,
+      "step": 6826
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.284539473684211e-06,
+      "loss": 0.9437,
+      "step": 6827
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.283991228070176e-06,
+      "loss": 0.6405,
+      "step": 6828
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.28344298245614e-06,
+      "loss": 1.1663,
+      "step": 6829
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.282894736842106e-06,
+      "loss": 1.2786,
+      "step": 6830
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.2823464912280706e-06,
+      "loss": 0.2616,
+      "step": 6831
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.281798245614036e-06,
+      "loss": 0.4401,
+      "step": 6832
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.281250000000001e-06,
+      "loss": 0.7883,
+      "step": 6833
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.280701754385965e-06,
+      "loss": 0.7611,
+      "step": 6834
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.28015350877193e-06,
+      "loss": 0.6222,
+      "step": 6835
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.2796052631578955e-06,
+      "loss": 0.5063,
+      "step": 6836
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.27905701754386e-06,
+      "loss": 1.1951,
+      "step": 6837
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.278508771929825e-06,
+      "loss": 0.4789,
+      "step": 6838
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.277960526315789e-06,
+      "loss": 1.9851,
+      "step": 6839
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.277412280701755e-06,
+      "loss": 1.1015,
+      "step": 6840
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.27686403508772e-06,
+      "loss": 0.6253,
+      "step": 6841
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.2763157894736845e-06,
+      "loss": 0.7586,
+      "step": 6842
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.27576754385965e-06,
+      "loss": 1.1986,
+      "step": 6843
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.275219298245615e-06,
+      "loss": 1.1432,
+      "step": 6844
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.274671052631579e-06,
+      "loss": 0.5828,
+      "step": 6845
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.274122807017544e-06,
+      "loss": 0.8343,
+      "step": 6846
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.2735745614035095e-06,
+      "loss": 1.4397,
+      "step": 6847
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.273026315789474e-06,
+      "loss": 0.5753,
+      "step": 6848
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.27247807017544e-06,
+      "loss": 1.1567,
+      "step": 6849
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.271929824561404e-06,
+      "loss": 0.5837,
+      "step": 6850
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.271381578947369e-06,
+      "loss": 0.5682,
+      "step": 6851
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.2708333333333336e-06,
+      "loss": 0.4894,
+      "step": 6852
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.270285087719299e-06,
+      "loss": 0.3231,
+      "step": 6853
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.269736842105264e-06,
+      "loss": 0.8783,
+      "step": 6854
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.269188596491229e-06,
+      "loss": 1.0935,
+      "step": 6855
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.268640350877193e-06,
+      "loss": 0.5416,
+      "step": 6856
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.2680921052631585e-06,
+      "loss": 0.1976,
+      "step": 6857
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.267543859649123e-06,
+      "loss": 1.8172,
+      "step": 6858
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.266995614035088e-06,
+      "loss": 0.2913,
+      "step": 6859
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.266447368421054e-06,
+      "loss": 0.2277,
+      "step": 6860
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.265899122807018e-06,
+      "loss": 0.9763,
+      "step": 6861
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.265350877192983e-06,
+      "loss": 1.3148,
+      "step": 6862
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.2648026315789475e-06,
+      "loss": 0.9054,
+      "step": 6863
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.264254385964913e-06,
+      "loss": 0.1268,
+      "step": 6864
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.263706140350878e-06,
+      "loss": 0.9595,
+      "step": 6865
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.263157894736842e-06,
+      "loss": 0.328,
+      "step": 6866
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.262609649122808e-06,
+      "loss": 0.9214,
+      "step": 6867
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.2620614035087725e-06,
+      "loss": 0.5037,
+      "step": 6868
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.261513157894737e-06,
+      "loss": 0.5505,
+      "step": 6869
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.260964912280703e-06,
+      "loss": 1.1794,
+      "step": 6870
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.260416666666668e-06,
+      "loss": 0.5397,
+      "step": 6871
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.259868421052632e-06,
+      "loss": 1.3011,
+      "step": 6872
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.259320175438597e-06,
+      "loss": 0.5579,
+      "step": 6873
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.258771929824562e-06,
+      "loss": 0.9685,
+      "step": 6874
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.258223684210527e-06,
+      "loss": 1.4773,
+      "step": 6875
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.257675438596492e-06,
+      "loss": 0.8984,
+      "step": 6876
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.257127192982456e-06,
+      "loss": 0.4628,
+      "step": 6877
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.2565789473684216e-06,
+      "loss": 1.1364,
+      "step": 6878
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.256030701754386e-06,
+      "loss": 0.9702,
+      "step": 6879
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.255482456140351e-06,
+      "loss": 0.8506,
+      "step": 6880
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.254934210526317e-06,
+      "loss": 0.5801,
+      "step": 6881
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.254385964912281e-06,
+      "loss": 0.428,
+      "step": 6882
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.253837719298246e-06,
+      "loss": 1.227,
+      "step": 6883
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.2532894736842105e-06,
+      "loss": 1.4261,
+      "step": 6884
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.252741228070176e-06,
+      "loss": 1.0423,
+      "step": 6885
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.252192982456141e-06,
+      "loss": 0.8563,
+      "step": 6886
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.251644736842107e-06,
+      "loss": 1.1786,
+      "step": 6887
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.251096491228071e-06,
+      "loss": 1.3431,
+      "step": 6888
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.2505482456140355e-06,
+      "loss": 0.936,
+      "step": 6889
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.25e-06,
+      "loss": 0.3025,
+      "step": 6890
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.249451754385966e-06,
+      "loss": 0.4218,
+      "step": 6891
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.248903508771931e-06,
+      "loss": 0.8211,
+      "step": 6892
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.248355263157895e-06,
+      "loss": 0.1551,
+      "step": 6893
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.24780701754386e-06,
+      "loss": 0.5853,
+      "step": 6894
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.247258771929825e-06,
+      "loss": 0.4856,
+      "step": 6895
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.24671052631579e-06,
+      "loss": 0.9528,
+      "step": 6896
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.246162280701755e-06,
+      "loss": 0.2019,
+      "step": 6897
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.245614035087721e-06,
+      "loss": 0.4582,
+      "step": 6898
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.2450657894736846e-06,
+      "loss": 0.7546,
+      "step": 6899
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.244517543859649e-06,
+      "loss": 1.079,
+      "step": 6900
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.243969298245614e-06,
+      "loss": 0.5678,
+      "step": 6901
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.24342105263158e-06,
+      "loss": 0.3065,
+      "step": 6902
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.242872807017545e-06,
+      "loss": 0.3825,
+      "step": 6903
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.242324561403509e-06,
+      "loss": 1.1513,
+      "step": 6904
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.241776315789474e-06,
+      "loss": 0.7566,
+      "step": 6905
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.241228070175439e-06,
+      "loss": 0.905,
+      "step": 6906
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.240679824561404e-06,
+      "loss": 0.5894,
+      "step": 6907
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.24013157894737e-06,
+      "loss": 0.495,
+      "step": 6908
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.239583333333334e-06,
+      "loss": 1.564,
+      "step": 6909
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.2390350877192985e-06,
+      "loss": 0.399,
+      "step": 6910
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.238486842105263e-06,
+      "loss": 1.6112,
+      "step": 6911
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.237938596491229e-06,
+      "loss": 1.3746,
+      "step": 6912
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.237390350877194e-06,
+      "loss": 0.7262,
+      "step": 6913
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.236842105263159e-06,
+      "loss": 1.2439,
+      "step": 6914
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.236293859649123e-06,
+      "loss": 0.8593,
+      "step": 6915
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.235745614035088e-06,
+      "loss": 0.3004,
+      "step": 6916
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.235197368421053e-06,
+      "loss": 0.8615,
+      "step": 6917
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.234649122807018e-06,
+      "loss": 0.7685,
+      "step": 6918
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.234100877192984e-06,
+      "loss": 0.6667,
+      "step": 6919
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.233552631578948e-06,
+      "loss": 1.1933,
+      "step": 6920
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.233004385964912e-06,
+      "loss": 0.4943,
+      "step": 6921
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.232456140350877e-06,
+      "loss": 0.5487,
+      "step": 6922
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.231907894736843e-06,
+      "loss": 0.4666,
+      "step": 6923
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.231359649122808e-06,
+      "loss": 0.54,
+      "step": 6924
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.230811403508772e-06,
+      "loss": 0.9833,
+      "step": 6925
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.230263157894737e-06,
+      "loss": 0.4679,
+      "step": 6926
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.229714912280702e-06,
+      "loss": 0.8268,
+      "step": 6927
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.229166666666667e-06,
+      "loss": 0.6345,
+      "step": 6928
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.228618421052633e-06,
+      "loss": 0.1373,
+      "step": 6929
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.2280701754385975e-06,
+      "loss": 0.3361,
+      "step": 6930
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.2275219298245615e-06,
+      "loss": 0.3888,
+      "step": 6931
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.226973684210526e-06,
+      "loss": 0.3735,
+      "step": 6932
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.226425438596492e-06,
+      "loss": 0.4107,
+      "step": 6933
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.225877192982457e-06,
+      "loss": 0.4285,
+      "step": 6934
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.225328947368422e-06,
+      "loss": 0.9266,
+      "step": 6935
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.224780701754386e-06,
+      "loss": 0.963,
+      "step": 6936
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.224232456140351e-06,
+      "loss": 0.976,
+      "step": 6937
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.223684210526316e-06,
+      "loss": 0.9327,
+      "step": 6938
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.223135964912281e-06,
+      "loss": 0.2002,
+      "step": 6939
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.222587719298247e-06,
+      "loss": 0.5945,
+      "step": 6940
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.2220394736842114e-06,
+      "loss": 0.2214,
+      "step": 6941
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.221491228070175e-06,
+      "loss": 0.9611,
+      "step": 6942
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.220942982456141e-06,
+      "loss": 2.0525,
+      "step": 6943
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.220394736842106e-06,
+      "loss": 0.8982,
+      "step": 6944
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.219846491228071e-06,
+      "loss": 0.2752,
+      "step": 6945
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.219298245614036e-06,
+      "loss": 0.4609,
+      "step": 6946
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.21875e-06,
+      "loss": 1.3453,
+      "step": 6947
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.218201754385965e-06,
+      "loss": 0.3805,
+      "step": 6948
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.21765350877193e-06,
+      "loss": 0.2438,
+      "step": 6949
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.217105263157896e-06,
+      "loss": 0.2712,
+      "step": 6950
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.2165570175438605e-06,
+      "loss": 0.7473,
+      "step": 6951
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.2160087719298245e-06,
+      "loss": 1.5909,
+      "step": 6952
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.215460526315789e-06,
+      "loss": 1.0288,
+      "step": 6953
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.214912280701755e-06,
+      "loss": 0.1623,
+      "step": 6954
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.21436403508772e-06,
+      "loss": 0.2445,
+      "step": 6955
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.213815789473685e-06,
+      "loss": 0.2882,
+      "step": 6956
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.21326754385965e-06,
+      "loss": 0.7165,
+      "step": 6957
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.212719298245614e-06,
+      "loss": 1.6654,
+      "step": 6958
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.212171052631579e-06,
+      "loss": 0.3953,
+      "step": 6959
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.211622807017545e-06,
+      "loss": 1.1171,
+      "step": 6960
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.21107456140351e-06,
+      "loss": 0.1186,
+      "step": 6961
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.2105263157894745e-06,
+      "loss": 1.396,
+      "step": 6962
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.2099780701754384e-06,
+      "loss": 1.0344,
+      "step": 6963
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.209429824561404e-06,
+      "loss": 0.4678,
+      "step": 6964
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.208881578947369e-06,
+      "loss": 0.8602,
+      "step": 6965
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.208333333333334e-06,
+      "loss": 1.1155,
+      "step": 6966
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.207785087719299e-06,
+      "loss": 1.0752,
+      "step": 6967
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.207236842105263e-06,
+      "loss": 0.8577,
+      "step": 6968
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.206688596491228e-06,
+      "loss": 0.7392,
+      "step": 6969
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.206140350877193e-06,
+      "loss": 0.9191,
+      "step": 6970
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.205592105263159e-06,
+      "loss": 0.874,
+      "step": 6971
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.2050438596491235e-06,
+      "loss": 1.467,
+      "step": 6972
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.204495614035088e-06,
+      "loss": 0.8956,
+      "step": 6973
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.203947368421052e-06,
+      "loss": 0.6954,
+      "step": 6974
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.203399122807018e-06,
+      "loss": 1.0962,
+      "step": 6975
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.202850877192983e-06,
+      "loss": 0.4732,
+      "step": 6976
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.202302631578948e-06,
+      "loss": 1.311,
+      "step": 6977
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.201754385964913e-06,
+      "loss": 0.8045,
+      "step": 6978
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.201206140350877e-06,
+      "loss": 0.6092,
+      "step": 6979
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.200657894736842e-06,
+      "loss": 0.6606,
+      "step": 6980
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.200109649122808e-06,
+      "loss": 1.6534,
+      "step": 6981
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.199561403508773e-06,
+      "loss": 0.5642,
+      "step": 6982
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.1990131578947375e-06,
+      "loss": 0.4537,
+      "step": 6983
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.198464912280703e-06,
+      "loss": 1.1665,
+      "step": 6984
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.197916666666667e-06,
+      "loss": 0.5603,
+      "step": 6985
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.197368421052632e-06,
+      "loss": 0.3479,
+      "step": 6986
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.196820175438597e-06,
+      "loss": 1.0337,
+      "step": 6987
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.1962719298245624e-06,
+      "loss": 0.9694,
+      "step": 6988
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.195723684210527e-06,
+      "loss": 0.9618,
+      "step": 6989
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.195175438596491e-06,
+      "loss": 0.6963,
+      "step": 6990
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.194627192982456e-06,
+      "loss": 0.4157,
+      "step": 6991
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.194078947368422e-06,
+      "loss": 0.5823,
+      "step": 6992
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.1935307017543866e-06,
+      "loss": 0.6919,
+      "step": 6993
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.192982456140351e-06,
+      "loss": 0.8481,
+      "step": 6994
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.192434210526315e-06,
+      "loss": 0.4407,
+      "step": 6995
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.191885964912281e-06,
+      "loss": 1.6161,
+      "step": 6996
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.191337719298246e-06,
+      "loss": 0.4567,
+      "step": 6997
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.1907894736842115e-06,
+      "loss": 0.4958,
+      "step": 6998
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.190241228070176e-06,
+      "loss": 0.4822,
+      "step": 6999
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.189692982456141e-06,
+      "loss": 1.1859,
+      "step": 7000
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.189144736842105e-06,
+      "loss": 1.4042,
+      "step": 7001
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.188596491228071e-06,
+      "loss": 1.1613,
+      "step": 7002
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.188048245614036e-06,
+      "loss": 1.1904,
+      "step": 7003
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.1875000000000005e-06,
+      "loss": 1.4578,
+      "step": 7004
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.186951754385966e-06,
+      "loss": 0.5371,
+      "step": 7005
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.18640350877193e-06,
+      "loss": 0.9075,
+      "step": 7006
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.185855263157895e-06,
+      "loss": 0.5619,
+      "step": 7007
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.18530701754386e-06,
+      "loss": 1.2907,
+      "step": 7008
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.1847587719298254e-06,
+      "loss": 0.6087,
+      "step": 7009
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.18421052631579e-06,
+      "loss": 0.6641,
+      "step": 7010
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.183662280701754e-06,
+      "loss": 0.8461,
+      "step": 7011
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.183114035087719e-06,
+      "loss": 0.5904,
+      "step": 7012
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.182565789473685e-06,
+      "loss": 0.6136,
+      "step": 7013
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.1820175438596496e-06,
+      "loss": 0.4604,
+      "step": 7014
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.181469298245614e-06,
+      "loss": 0.4257,
+      "step": 7015
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.18092105263158e-06,
+      "loss": 0.2568,
+      "step": 7016
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.180372807017544e-06,
+      "loss": 0.3994,
+      "step": 7017
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.179824561403509e-06,
+      "loss": 1.3156,
+      "step": 7018
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.1792763157894745e-06,
+      "loss": 0.8299,
+      "step": 7019
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.178728070175439e-06,
+      "loss": 0.8657,
+      "step": 7020
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.178179824561404e-06,
+      "loss": 0.4795,
+      "step": 7021
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.177631578947368e-06,
+      "loss": 1.1748,
+      "step": 7022
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.177083333333334e-06,
+      "loss": 1.0561,
+      "step": 7023
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.176535087719299e-06,
+      "loss": 0.3556,
+      "step": 7024
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.1759868421052635e-06,
+      "loss": 0.539,
+      "step": 7025
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.175438596491229e-06,
+      "loss": 1.3978,
+      "step": 7026
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.174890350877194e-06,
+      "loss": 1.039,
+      "step": 7027
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.174342105263158e-06,
+      "loss": 0.3561,
+      "step": 7028
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.173793859649123e-06,
+      "loss": 1.5025,
+      "step": 7029
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.1732456140350885e-06,
+      "loss": 0.4578,
+      "step": 7030
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.172697368421053e-06,
+      "loss": 0.8731,
+      "step": 7031
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.172149122807018e-06,
+      "loss": 0.6624,
+      "step": 7032
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.171600877192983e-06,
+      "loss": 1.0729,
+      "step": 7033
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.171052631578948e-06,
+      "loss": 0.5007,
+      "step": 7034
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.170504385964913e-06,
+      "loss": 0.7259,
+      "step": 7035
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.169956140350878e-06,
+      "loss": 0.3205,
+      "step": 7036
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.169407894736843e-06,
+      "loss": 0.3435,
+      "step": 7037
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.168859649122807e-06,
+      "loss": 1.2911,
+      "step": 7038
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.168311403508772e-06,
+      "loss": 1.1042,
+      "step": 7039
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.1677631578947376e-06,
+      "loss": 1.2355,
+      "step": 7040
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.167214912280702e-06,
+      "loss": 0.3727,
+      "step": 7041
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.166666666666667e-06,
+      "loss": 0.6155,
+      "step": 7042
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.166118421052633e-06,
+      "loss": 1.2292,
+      "step": 7043
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.165570175438597e-06,
+      "loss": 0.5471,
+      "step": 7044
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.165021929824562e-06,
+      "loss": 0.3903,
+      "step": 7045
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.1644736842105265e-06,
+      "loss": 1.1596,
+      "step": 7046
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.163925438596492e-06,
+      "loss": 0.538,
+      "step": 7047
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.163377192982457e-06,
+      "loss": 0.6727,
+      "step": 7048
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.162828947368421e-06,
+      "loss": 0.3683,
+      "step": 7049
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.162280701754386e-06,
+      "loss": 0.4689,
+      "step": 7050
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.1617324561403515e-06,
+      "loss": 1.1312,
+      "step": 7051
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.161184210526316e-06,
+      "loss": 1.0894,
+      "step": 7052
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.160635964912282e-06,
+      "loss": 0.4869,
+      "step": 7053
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.160087719298247e-06,
+      "loss": 0.7739,
+      "step": 7054
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.159539473684211e-06,
+      "loss": 0.6001,
+      "step": 7055
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.158991228070176e-06,
+      "loss": 0.5882,
+      "step": 7056
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.158442982456141e-06,
+      "loss": 0.9242,
+      "step": 7057
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.157894736842106e-06,
+      "loss": 1.2725,
+      "step": 7058
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.157346491228071e-06,
+      "loss": 0.484,
+      "step": 7059
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.156798245614035e-06,
+      "loss": 0.492,
+      "step": 7060
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.1562500000000006e-06,
+      "loss": 0.7358,
+      "step": 7061
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.155701754385965e-06,
+      "loss": 0.767,
+      "step": 7062
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.15515350877193e-06,
+      "loss": 0.574,
+      "step": 7063
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.154605263157896e-06,
+      "loss": 0.5228,
+      "step": 7064
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.15405701754386e-06,
+      "loss": 0.6101,
+      "step": 7065
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.153508771929825e-06,
+      "loss": 1.125,
+      "step": 7066
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.1529605263157895e-06,
+      "loss": 0.8733,
+      "step": 7067
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.152412280701755e-06,
+      "loss": 0.6387,
+      "step": 7068
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.15186403508772e-06,
+      "loss": 0.6574,
+      "step": 7069
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.151315789473685e-06,
+      "loss": 0.3691,
+      "step": 7070
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.15076754385965e-06,
+      "loss": 0.3172,
+      "step": 7071
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.1502192982456145e-06,
+      "loss": 0.7945,
+      "step": 7072
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.149671052631579e-06,
+      "loss": 0.2543,
+      "step": 7073
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.149122807017545e-06,
+      "loss": 0.6617,
+      "step": 7074
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.14857456140351e-06,
+      "loss": 0.7058,
+      "step": 7075
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.148026315789474e-06,
+      "loss": 0.4037,
+      "step": 7076
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.147478070175439e-06,
+      "loss": 0.6301,
+      "step": 7077
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.146929824561404e-06,
+      "loss": 1.4325,
+      "step": 7078
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.146381578947369e-06,
+      "loss": 1.1742,
+      "step": 7079
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.145833333333334e-06,
+      "loss": 1.1717,
+      "step": 7080
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.145285087719298e-06,
+      "loss": 0.4293,
+      "step": 7081
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.1447368421052636e-06,
+      "loss": 0.6962,
+      "step": 7082
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.144188596491228e-06,
+      "loss": 0.7249,
+      "step": 7083
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.143640350877193e-06,
+      "loss": 0.3023,
+      "step": 7084
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.143092105263159e-06,
+      "loss": 0.7272,
+      "step": 7085
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.142543859649124e-06,
+      "loss": 1.6055,
+      "step": 7086
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.141995614035088e-06,
+      "loss": 0.4379,
+      "step": 7087
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.1414473684210525e-06,
+      "loss": 0.672,
+      "step": 7088
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.140899122807018e-06,
+      "loss": 1.1981,
+      "step": 7089
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.140350877192983e-06,
+      "loss": 0.3243,
+      "step": 7090
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.139802631578949e-06,
+      "loss": 0.877,
+      "step": 7091
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.139254385964913e-06,
+      "loss": 1.044,
+      "step": 7092
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.1387061403508775e-06,
+      "loss": 0.3239,
+      "step": 7093
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.138157894736842e-06,
+      "loss": 1.0239,
+      "step": 7094
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.137609649122808e-06,
+      "loss": 0.6645,
+      "step": 7095
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.137061403508773e-06,
+      "loss": 1.008,
+      "step": 7096
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.136513157894738e-06,
+      "loss": 0.5843,
+      "step": 7097
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.135964912280702e-06,
+      "loss": 0.9966,
+      "step": 7098
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.135416666666667e-06,
+      "loss": 0.2451,
+      "step": 7099
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.134868421052632e-06,
+      "loss": 0.6959,
+      "step": 7100
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.134320175438597e-06,
+      "loss": 0.4683,
+      "step": 7101
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.133771929824563e-06,
+      "loss": 0.4203,
+      "step": 7102
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.133223684210527e-06,
+      "loss": 0.9526,
+      "step": 7103
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.132675438596491e-06,
+      "loss": 0.9168,
+      "step": 7104
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.132127192982456e-06,
+      "loss": 0.5417,
+      "step": 7105
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.131578947368422e-06,
+      "loss": 0.6509,
+      "step": 7106
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.131030701754387e-06,
+      "loss": 0.327,
+      "step": 7107
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.130482456140351e-06,
+      "loss": 1.0623,
+      "step": 7108
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.129934210526316e-06,
+      "loss": 0.6981,
+      "step": 7109
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.129385964912281e-06,
+      "loss": 0.3881,
+      "step": 7110
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.128837719298246e-06,
+      "loss": 0.592,
+      "step": 7111
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.128289473684212e-06,
+      "loss": 0.9686,
+      "step": 7112
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.1277412280701765e-06,
+      "loss": 0.6874,
+      "step": 7113
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.1271929824561405e-06,
+      "loss": 0.9027,
+      "step": 7114
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.126644736842105e-06,
+      "loss": 0.2771,
+      "step": 7115
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.126096491228071e-06,
+      "loss": 0.51,
+      "step": 7116
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.125548245614036e-06,
+      "loss": 0.8255,
+      "step": 7117
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.125000000000001e-06,
+      "loss": 0.5299,
+      "step": 7118
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.124451754385965e-06,
+      "loss": 1.1677,
+      "step": 7119
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.12390350877193e-06,
+      "loss": 0.5334,
+      "step": 7120
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.123355263157895e-06,
+      "loss": 0.6295,
+      "step": 7121
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.12280701754386e-06,
+      "loss": 1.4214,
+      "step": 7122
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.122258771929826e-06,
+      "loss": 0.4557,
+      "step": 7123
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.12171052631579e-06,
+      "loss": 0.7484,
+      "step": 7124
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.121162280701754e-06,
+      "loss": 0.4645,
+      "step": 7125
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.12061403508772e-06,
+      "loss": 1.1596,
+      "step": 7126
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.120065789473685e-06,
+      "loss": 0.8065,
+      "step": 7127
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.11951754385965e-06,
+      "loss": 0.1867,
+      "step": 7128
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.118969298245615e-06,
+      "loss": 0.6297,
+      "step": 7129
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.118421052631579e-06,
+      "loss": 0.3066,
+      "step": 7130
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.117872807017544e-06,
+      "loss": 1.0974,
+      "step": 7131
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.117324561403509e-06,
+      "loss": 1.0012,
+      "step": 7132
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.116776315789475e-06,
+      "loss": 0.5605,
+      "step": 7133
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.1162280701754395e-06,
+      "loss": 0.8288,
+      "step": 7134
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.1156798245614035e-06,
+      "loss": 1.3162,
+      "step": 7135
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.115131578947368e-06,
+      "loss": 0.6343,
+      "step": 7136
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.114583333333334e-06,
+      "loss": 0.8242,
+      "step": 7137
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.114035087719299e-06,
+      "loss": 0.4213,
+      "step": 7138
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.113486842105264e-06,
+      "loss": 0.3748,
+      "step": 7139
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.112938596491229e-06,
+      "loss": 1.0279,
+      "step": 7140
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.112390350877193e-06,
+      "loss": 0.787,
+      "step": 7141
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.111842105263158e-06,
+      "loss": 0.6954,
+      "step": 7142
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.111293859649123e-06,
+      "loss": 0.4353,
+      "step": 7143
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.110745614035089e-06,
+      "loss": 0.8424,
+      "step": 7144
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.1101973684210535e-06,
+      "loss": 1.0397,
+      "step": 7145
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.1096491228070174e-06,
+      "loss": 1.0316,
+      "step": 7146
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.109100877192983e-06,
+      "loss": 0.4351,
+      "step": 7147
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.108552631578948e-06,
+      "loss": 0.8576,
+      "step": 7148
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.108004385964913e-06,
+      "loss": 0.7054,
+      "step": 7149
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.1074561403508784e-06,
+      "loss": 0.8777,
+      "step": 7150
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.106907894736842e-06,
+      "loss": 0.4736,
+      "step": 7151
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.106359649122807e-06,
+      "loss": 1.4901,
+      "step": 7152
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.105811403508772e-06,
+      "loss": 1.3952,
+      "step": 7153
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.105263157894738e-06,
+      "loss": 1.3671,
+      "step": 7154
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.1047149122807025e-06,
+      "loss": 0.2584,
+      "step": 7155
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.104166666666667e-06,
+      "loss": 0.3421,
+      "step": 7156
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.103618421052631e-06,
+      "loss": 0.6176,
+      "step": 7157
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.103070175438597e-06,
+      "loss": 1.2056,
+      "step": 7158
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.102521929824562e-06,
+      "loss": 0.4691,
+      "step": 7159
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.101973684210527e-06,
+      "loss": 1.4323,
+      "step": 7160
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.101425438596492e-06,
+      "loss": 1.2484,
+      "step": 7161
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.100877192982456e-06,
+      "loss": 0.5334,
+      "step": 7162
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.100328947368421e-06,
+      "loss": 0.9948,
+      "step": 7163
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.099780701754387e-06,
+      "loss": 0.6697,
+      "step": 7164
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.099232456140352e-06,
+      "loss": 0.6729,
+      "step": 7165
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.0986842105263165e-06,
+      "loss": 0.6288,
+      "step": 7166
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.0981359649122804e-06,
+      "loss": 1.0132,
+      "step": 7167
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.097587719298246e-06,
+      "loss": 0.3388,
+      "step": 7168
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.097039473684211e-06,
+      "loss": 1.0093,
+      "step": 7169
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.096491228070176e-06,
+      "loss": 0.9612,
+      "step": 7170
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.0959429824561414e-06,
+      "loss": 1.8927,
+      "step": 7171
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.095394736842106e-06,
+      "loss": 0.7063,
+      "step": 7172
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.09484649122807e-06,
+      "loss": 0.7228,
+      "step": 7173
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.094298245614035e-06,
+      "loss": 1.0558,
+      "step": 7174
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.093750000000001e-06,
+      "loss": 1.0185,
+      "step": 7175
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.0932017543859656e-06,
+      "loss": 0.5213,
+      "step": 7176
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.09265350877193e-06,
+      "loss": 0.5206,
+      "step": 7177
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.092105263157894e-06,
+      "loss": 1.4713,
+      "step": 7178
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.09155701754386e-06,
+      "loss": 0.9573,
+      "step": 7179
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.091008771929825e-06,
+      "loss": 0.3617,
+      "step": 7180
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.09046052631579e-06,
+      "loss": 1.2493,
+      "step": 7181
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.089912280701755e-06,
+      "loss": 0.7601,
+      "step": 7182
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.08936403508772e-06,
+      "loss": 0.4854,
+      "step": 7183
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.088815789473684e-06,
+      "loss": 0.4553,
+      "step": 7184
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.08826754385965e-06,
+      "loss": 1.1401,
+      "step": 7185
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.087719298245615e-06,
+      "loss": 0.4352,
+      "step": 7186
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.0871710526315795e-06,
+      "loss": 0.5957,
+      "step": 7187
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.086622807017545e-06,
+      "loss": 1.9316,
+      "step": 7188
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.086074561403509e-06,
+      "loss": 0.4021,
+      "step": 7189
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.085526315789474e-06,
+      "loss": 1.0936,
+      "step": 7190
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.084978070175439e-06,
+      "loss": 0.852,
+      "step": 7191
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.0844298245614044e-06,
+      "loss": 0.7737,
+      "step": 7192
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.083881578947369e-06,
+      "loss": 1.0988,
+      "step": 7193
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.083333333333333e-06,
+      "loss": 0.3695,
+      "step": 7194
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.082785087719298e-06,
+      "loss": 1.0979,
+      "step": 7195
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.082236842105264e-06,
+      "loss": 0.5429,
+      "step": 7196
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.0816885964912286e-06,
+      "loss": 1.3183,
+      "step": 7197
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.081140350877193e-06,
+      "loss": 0.361,
+      "step": 7198
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.080592105263159e-06,
+      "loss": 1.0216,
+      "step": 7199
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.080043859649123e-06,
+      "loss": 0.6076,
+      "step": 7200
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.079495614035088e-06,
+      "loss": 0.7493,
+      "step": 7201
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.0789473684210535e-06,
+      "loss": 0.4491,
+      "step": 7202
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.078399122807018e-06,
+      "loss": 0.8065,
+      "step": 7203
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.077850877192983e-06,
+      "loss": 0.9111,
+      "step": 7204
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.077302631578947e-06,
+      "loss": 0.9052,
+      "step": 7205
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.076754385964913e-06,
+      "loss": 0.4279,
+      "step": 7206
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.076206140350878e-06,
+      "loss": 0.5653,
+      "step": 7207
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.0756578947368425e-06,
+      "loss": 0.738,
+      "step": 7208
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.075109649122808e-06,
+      "loss": 0.5059,
+      "step": 7209
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.074561403508772e-06,
+      "loss": 0.7069,
+      "step": 7210
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.074013157894737e-06,
+      "loss": 0.6864,
+      "step": 7211
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.073464912280702e-06,
+      "loss": 1.4621,
+      "step": 7212
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.0729166666666675e-06,
+      "loss": 1.1817,
+      "step": 7213
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.072368421052632e-06,
+      "loss": 0.9455,
+      "step": 7214
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.071820175438597e-06,
+      "loss": 0.6423,
+      "step": 7215
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.071271929824561e-06,
+      "loss": 0.8111,
+      "step": 7216
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.070723684210527e-06,
+      "loss": 0.5701,
+      "step": 7217
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.070175438596492e-06,
+      "loss": 0.6919,
+      "step": 7218
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.069627192982457e-06,
+      "loss": 0.4583,
+      "step": 7219
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.069078947368422e-06,
+      "loss": 0.4578,
+      "step": 7220
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.068530701754386e-06,
+      "loss": 1.0818,
+      "step": 7221
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.067982456140351e-06,
+      "loss": 0.3311,
+      "step": 7222
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.0674342105263166e-06,
+      "loss": 1.1513,
+      "step": 7223
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.066885964912281e-06,
+      "loss": 0.3779,
+      "step": 7224
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.066337719298246e-06,
+      "loss": 1.035,
+      "step": 7225
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.065789473684212e-06,
+      "loss": 0.6231,
+      "step": 7226
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.065241228070176e-06,
+      "loss": 0.676,
+      "step": 7227
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.064692982456141e-06,
+      "loss": 0.6062,
+      "step": 7228
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.0641447368421055e-06,
+      "loss": 0.282,
+      "step": 7229
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.063596491228071e-06,
+      "loss": 1.0451,
+      "step": 7230
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.063048245614036e-06,
+      "loss": 0.3401,
+      "step": 7231
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.0625e-06,
+      "loss": 1.2709,
+      "step": 7232
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.061951754385965e-06,
+      "loss": 0.393,
+      "step": 7233
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.0614035087719305e-06,
+      "loss": 0.6662,
+      "step": 7234
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.060855263157895e-06,
+      "loss": 0.2416,
+      "step": 7235
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.06030701754386e-06,
+      "loss": 0.5879,
+      "step": 7236
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.059758771929825e-06,
+      "loss": 0.5283,
+      "step": 7237
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.05921052631579e-06,
+      "loss": 1.1069,
+      "step": 7238
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.058662280701755e-06,
+      "loss": 0.6536,
+      "step": 7239
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.05811403508772e-06,
+      "loss": 0.6994,
+      "step": 7240
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.057565789473685e-06,
+      "loss": 0.7912,
+      "step": 7241
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.05701754385965e-06,
+      "loss": 0.3381,
+      "step": 7242
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.056469298245614e-06,
+      "loss": 0.4266,
+      "step": 7243
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.0559210526315796e-06,
+      "loss": 0.2331,
+      "step": 7244
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.055372807017544e-06,
+      "loss": 0.555,
+      "step": 7245
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.054824561403509e-06,
+      "loss": 0.6956,
+      "step": 7246
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.054276315789475e-06,
+      "loss": 0.3179,
+      "step": 7247
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.053728070175439e-06,
+      "loss": 1.0232,
+      "step": 7248
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.053179824561404e-06,
+      "loss": 0.4579,
+      "step": 7249
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.0526315789473685e-06,
+      "loss": 0.4326,
+      "step": 7250
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.052083333333334e-06,
+      "loss": 1.1889,
+      "step": 7251
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.051535087719299e-06,
+      "loss": 1.2742,
+      "step": 7252
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.050986842105263e-06,
+      "loss": 0.4581,
+      "step": 7253
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.050438596491228e-06,
+      "loss": 0.5188,
+      "step": 7254
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.0498903508771935e-06,
+      "loss": 0.6454,
+      "step": 7255
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.049342105263158e-06,
+      "loss": 0.9644,
+      "step": 7256
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.048793859649124e-06,
+      "loss": 0.8754,
+      "step": 7257
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.048245614035089e-06,
+      "loss": 1.1978,
+      "step": 7258
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.047697368421053e-06,
+      "loss": 0.7082,
+      "step": 7259
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.047149122807018e-06,
+      "loss": 0.5112,
+      "step": 7260
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.046600877192983e-06,
+      "loss": 1.0194,
+      "step": 7261
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.046052631578948e-06,
+      "loss": 0.5781,
+      "step": 7262
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.045504385964913e-06,
+      "loss": 0.8286,
+      "step": 7263
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.044956140350877e-06,
+      "loss": 0.4849,
+      "step": 7264
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.044407894736843e-06,
+      "loss": 1.0307,
+      "step": 7265
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.043859649122807e-06,
+      "loss": 0.3164,
+      "step": 7266
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.043311403508772e-06,
+      "loss": 1.0609,
+      "step": 7267
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.042763157894738e-06,
+      "loss": 1.2179,
+      "step": 7268
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.042214912280703e-06,
+      "loss": 0.2559,
+      "step": 7269
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.041666666666667e-06,
+      "loss": 1.1629,
+      "step": 7270
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.0411184210526315e-06,
+      "loss": 0.5774,
+      "step": 7271
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.040570175438597e-06,
+      "loss": 0.8599,
+      "step": 7272
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.040021929824562e-06,
+      "loss": 1.6322,
+      "step": 7273
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.039473684210527e-06,
+      "loss": 0.5773,
+      "step": 7274
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.038925438596492e-06,
+      "loss": 0.7898,
+      "step": 7275
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.0383771929824565e-06,
+      "loss": 0.9861,
+      "step": 7276
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.037828947368421e-06,
+      "loss": 0.5765,
+      "step": 7277
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.037280701754387e-06,
+      "loss": 0.7242,
+      "step": 7278
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.036732456140352e-06,
+      "loss": 0.1667,
+      "step": 7279
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.036184210526316e-06,
+      "loss": 0.9201,
+      "step": 7280
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.035635964912281e-06,
+      "loss": 1.1279,
+      "step": 7281
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.035087719298246e-06,
+      "loss": 1.2135,
+      "step": 7282
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.034539473684211e-06,
+      "loss": 0.854,
+      "step": 7283
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.033991228070176e-06,
+      "loss": 1.3493,
+      "step": 7284
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.033442982456142e-06,
+      "loss": 1.5669,
+      "step": 7285
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.032894736842106e-06,
+      "loss": 0.3606,
+      "step": 7286
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.03234649122807e-06,
+      "loss": 0.6792,
+      "step": 7287
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.031798245614035e-06,
+      "loss": 0.8436,
+      "step": 7288
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.031250000000001e-06,
+      "loss": 0.6738,
+      "step": 7289
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.030701754385966e-06,
+      "loss": 1.3714,
+      "step": 7290
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.03015350877193e-06,
+      "loss": 1.0456,
+      "step": 7291
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.0296052631578945e-06,
+      "loss": 0.5817,
+      "step": 7292
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.02905701754386e-06,
+      "loss": 1.0944,
+      "step": 7293
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.028508771929825e-06,
+      "loss": 1.2396,
+      "step": 7294
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.027960526315791e-06,
+      "loss": 0.6618,
+      "step": 7295
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.027412280701755e-06,
+      "loss": 0.1807,
+      "step": 7296
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.0268640350877195e-06,
+      "loss": 0.9276,
+      "step": 7297
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.026315789473684e-06,
+      "loss": 0.487,
+      "step": 7298
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.02576754385965e-06,
+      "loss": 0.6858,
+      "step": 7299
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.025219298245615e-06,
+      "loss": 0.9918,
+      "step": 7300
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.02467105263158e-06,
+      "loss": 0.672,
+      "step": 7301
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.024122807017544e-06,
+      "loss": 0.8312,
+      "step": 7302
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.023574561403509e-06,
+      "loss": 0.6189,
+      "step": 7303
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.023026315789474e-06,
+      "loss": 1.0705,
+      "step": 7304
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.022478070175439e-06,
+      "loss": 1.102,
+      "step": 7305
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.021929824561405e-06,
+      "loss": 0.6129,
+      "step": 7306
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.021381578947369e-06,
+      "loss": 0.6271,
+      "step": 7307
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.0208333333333334e-06,
+      "loss": 1.3392,
+      "step": 7308
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.020285087719298e-06,
+      "loss": 0.6575,
+      "step": 7309
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.019736842105264e-06,
+      "loss": 1.1006,
+      "step": 7310
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.019188596491229e-06,
+      "loss": 0.6477,
+      "step": 7311
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.018640350877194e-06,
+      "loss": 0.2861,
+      "step": 7312
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.018092105263158e-06,
+      "loss": 1.1567,
+      "step": 7313
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.017543859649123e-06,
+      "loss": 1.0421,
+      "step": 7314
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.016995614035088e-06,
+      "loss": 0.1876,
+      "step": 7315
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.016447368421054e-06,
+      "loss": 0.7349,
+      "step": 7316
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.6994053721550133,
+      "eval_loss": 0.7773284912109375,
+      "eval_roc_auc": 0.9203856506795203,
+      "eval_runtime": 181.6268,
+      "eval_samples_per_second": 26.852,
+      "eval_steps_per_second": 3.359,
+      "step": 7316
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.0158991228070185e-06,
+      "loss": 0.4075,
+      "step": 7317
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.0153508771929825e-06,
+      "loss": 0.3048,
+      "step": 7318
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.014802631578947e-06,
+      "loss": 0.6133,
+      "step": 7319
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.014254385964913e-06,
+      "loss": 0.6327,
+      "step": 7320
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.013706140350878e-06,
+      "loss": 0.5558,
+      "step": 7321
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.013157894736843e-06,
+      "loss": 0.4863,
+      "step": 7322
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.012609649122807e-06,
+      "loss": 1.1396,
+      "step": 7323
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.012061403508772e-06,
+      "loss": 0.4168,
+      "step": 7324
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.011513157894737e-06,
+      "loss": 0.5604,
+      "step": 7325
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 6.010964912280702e-06,
+      "loss": 1.3294,
+      "step": 7326
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 6.010416666666668e-06,
+      "loss": 0.277,
+      "step": 7327
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 6.0098684210526325e-06,
+      "loss": 0.4295,
+      "step": 7328
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 6.0093201754385964e-06,
+      "loss": 0.7328,
+      "step": 7329
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 6.008771929824562e-06,
+      "loss": 0.8191,
+      "step": 7330
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 6.008223684210527e-06,
+      "loss": 0.9289,
+      "step": 7331
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 6.007675438596492e-06,
+      "loss": 1.007,
+      "step": 7332
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 6.0071271929824574e-06,
+      "loss": 0.4901,
+      "step": 7333
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 6.006578947368421e-06,
+      "loss": 0.317,
+      "step": 7334
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 6.006030701754386e-06,
+      "loss": 0.5371,
+      "step": 7335
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 6.005482456140351e-06,
+      "loss": 0.6782,
+      "step": 7336
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 6.004934210526317e-06,
+      "loss": 0.8256,
+      "step": 7337
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 6.0043859649122816e-06,
+      "loss": 0.4066,
+      "step": 7338
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 6.003837719298246e-06,
+      "loss": 0.4794,
+      "step": 7339
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 6.00328947368421e-06,
+      "loss": 0.1061,
+      "step": 7340
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 6.002741228070176e-06,
+      "loss": 0.9413,
+      "step": 7341
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 6.002192982456141e-06,
+      "loss": 0.9008,
+      "step": 7342
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 6.001644736842106e-06,
+      "loss": 0.9964,
+      "step": 7343
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 6.001096491228071e-06,
+      "loss": 0.8353,
+      "step": 7344
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 6.000548245614035e-06,
+      "loss": 0.251,
+      "step": 7345
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 6e-06,
+      "loss": 1.0821,
+      "step": 7346
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.999451754385965e-06,
+      "loss": 0.2032,
+      "step": 7347
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.998903508771931e-06,
+      "loss": 0.9287,
+      "step": 7348
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.9983552631578955e-06,
+      "loss": 0.2829,
+      "step": 7349
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.9978070175438595e-06,
+      "loss": 0.8685,
+      "step": 7350
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.997258771929825e-06,
+      "loss": 0.4379,
+      "step": 7351
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.99671052631579e-06,
+      "loss": 0.408,
+      "step": 7352
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.996162280701755e-06,
+      "loss": 1.0502,
+      "step": 7353
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.9956140350877204e-06,
+      "loss": 0.2341,
+      "step": 7354
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.995065789473685e-06,
+      "loss": 0.1941,
+      "step": 7355
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.994517543859649e-06,
+      "loss": 1.0117,
+      "step": 7356
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.993969298245614e-06,
+      "loss": 0.5689,
+      "step": 7357
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.99342105263158e-06,
+      "loss": 1.1088,
+      "step": 7358
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.9928728070175446e-06,
+      "loss": 0.6313,
+      "step": 7359
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.992324561403509e-06,
+      "loss": 0.7765,
+      "step": 7360
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.991776315789473e-06,
+      "loss": 0.1941,
+      "step": 7361
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.991228070175439e-06,
+      "loss": 0.8361,
+      "step": 7362
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.990679824561404e-06,
+      "loss": 1.4258,
+      "step": 7363
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.990131578947369e-06,
+      "loss": 0.7883,
+      "step": 7364
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.989583333333334e-06,
+      "loss": 0.618,
+      "step": 7365
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.989035087719298e-06,
+      "loss": 0.9754,
+      "step": 7366
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.988486842105263e-06,
+      "loss": 0.4093,
+      "step": 7367
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.987938596491229e-06,
+      "loss": 0.5998,
+      "step": 7368
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.987390350877194e-06,
+      "loss": 0.8351,
+      "step": 7369
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.9868421052631585e-06,
+      "loss": 0.5423,
+      "step": 7370
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.986293859649124e-06,
+      "loss": 0.3275,
+      "step": 7371
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.985745614035088e-06,
+      "loss": 0.9596,
+      "step": 7372
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.985197368421053e-06,
+      "loss": 0.5978,
+      "step": 7373
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.984649122807018e-06,
+      "loss": 1.2097,
+      "step": 7374
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.9841008771929835e-06,
+      "loss": 0.4023,
+      "step": 7375
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.983552631578948e-06,
+      "loss": 0.9295,
+      "step": 7376
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.983004385964912e-06,
+      "loss": 0.5596,
+      "step": 7377
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.982456140350877e-06,
+      "loss": 0.5461,
+      "step": 7378
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.981907894736843e-06,
+      "loss": 0.4426,
+      "step": 7379
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.981359649122808e-06,
+      "loss": 0.5583,
+      "step": 7380
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.980811403508772e-06,
+      "loss": 0.7428,
+      "step": 7381
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.980263157894738e-06,
+      "loss": 0.1865,
+      "step": 7382
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.979714912280702e-06,
+      "loss": 1.6753,
+      "step": 7383
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.979166666666667e-06,
+      "loss": 1.1207,
+      "step": 7384
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.9786184210526325e-06,
+      "loss": 0.8216,
+      "step": 7385
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.978070175438597e-06,
+      "loss": 0.7914,
+      "step": 7386
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.977521929824562e-06,
+      "loss": 1.2377,
+      "step": 7387
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.976973684210526e-06,
+      "loss": 0.8129,
+      "step": 7388
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.976425438596492e-06,
+      "loss": 1.1509,
+      "step": 7389
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.975877192982457e-06,
+      "loss": 1.2222,
+      "step": 7390
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.9753289473684215e-06,
+      "loss": 1.0403,
+      "step": 7391
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.974780701754387e-06,
+      "loss": 0.5856,
+      "step": 7392
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.974232456140351e-06,
+      "loss": 0.5819,
+      "step": 7393
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.973684210526316e-06,
+      "loss": 0.8807,
+      "step": 7394
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.973135964912281e-06,
+      "loss": 0.552,
+      "step": 7395
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.9725877192982465e-06,
+      "loss": 0.759,
+      "step": 7396
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.972039473684211e-06,
+      "loss": 0.3539,
+      "step": 7397
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.971491228070176e-06,
+      "loss": 0.5872,
+      "step": 7398
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.97094298245614e-06,
+      "loss": 0.4356,
+      "step": 7399
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.970394736842106e-06,
+      "loss": 0.5536,
+      "step": 7400
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.969846491228071e-06,
+      "loss": 0.215,
+      "step": 7401
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.969298245614035e-06,
+      "loss": 0.7503,
+      "step": 7402
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.968750000000001e-06,
+      "loss": 0.6166,
+      "step": 7403
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.968201754385965e-06,
+      "loss": 0.5601,
+      "step": 7404
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.96765350877193e-06,
+      "loss": 0.4409,
+      "step": 7405
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.9671052631578956e-06,
+      "loss": 0.3499,
+      "step": 7406
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.96655701754386e-06,
+      "loss": 0.5907,
+      "step": 7407
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.966008771929825e-06,
+      "loss": 0.2645,
+      "step": 7408
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.965460526315789e-06,
+      "loss": 0.8358,
+      "step": 7409
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.964912280701755e-06,
+      "loss": 0.8731,
+      "step": 7410
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.96436403508772e-06,
+      "loss": 0.5458,
+      "step": 7411
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.9638157894736845e-06,
+      "loss": 1.1074,
+      "step": 7412
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.96326754385965e-06,
+      "loss": 0.4789,
+      "step": 7413
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.962719298245615e-06,
+      "loss": 0.5074,
+      "step": 7414
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.962171052631579e-06,
+      "loss": 0.7208,
+      "step": 7415
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.961622807017544e-06,
+      "loss": 0.4263,
+      "step": 7416
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.9610745614035095e-06,
+      "loss": 0.8208,
+      "step": 7417
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.960526315789474e-06,
+      "loss": 0.6174,
+      "step": 7418
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.959978070175439e-06,
+      "loss": 0.6037,
+      "step": 7419
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.959429824561403e-06,
+      "loss": 1.4237,
+      "step": 7420
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.958881578947369e-06,
+      "loss": 0.6528,
+      "step": 7421
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.958333333333334e-06,
+      "loss": 0.7103,
+      "step": 7422
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.957785087719299e-06,
+      "loss": 1.1502,
+      "step": 7423
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.957236842105264e-06,
+      "loss": 0.5787,
+      "step": 7424
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.956688596491229e-06,
+      "loss": 1.214,
+      "step": 7425
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.956140350877193e-06,
+      "loss": 0.7358,
+      "step": 7426
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.9555921052631586e-06,
+      "loss": 0.632,
+      "step": 7427
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.955043859649123e-06,
+      "loss": 1.2959,
+      "step": 7428
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.954495614035088e-06,
+      "loss": 1.0589,
+      "step": 7429
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.953947368421054e-06,
+      "loss": 0.959,
+      "step": 7430
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.953399122807018e-06,
+      "loss": 1.0273,
+      "step": 7431
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.952850877192983e-06,
+      "loss": 0.2674,
+      "step": 7432
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.9523026315789475e-06,
+      "loss": 0.4203,
+      "step": 7433
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.951754385964913e-06,
+      "loss": 0.8476,
+      "step": 7434
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.951206140350878e-06,
+      "loss": 0.7328,
+      "step": 7435
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.950657894736842e-06,
+      "loss": 0.4345,
+      "step": 7436
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.950109649122807e-06,
+      "loss": 0.8515,
+      "step": 7437
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.9495614035087725e-06,
+      "loss": 0.7776,
+      "step": 7438
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.949013157894737e-06,
+      "loss": 0.5158,
+      "step": 7439
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.948464912280702e-06,
+      "loss": 1.0834,
+      "step": 7440
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.947916666666668e-06,
+      "loss": 0.1345,
+      "step": 7441
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.947368421052632e-06,
+      "loss": 0.795,
+      "step": 7442
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.946820175438597e-06,
+      "loss": 0.6512,
+      "step": 7443
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.946271929824562e-06,
+      "loss": 0.4314,
+      "step": 7444
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.945723684210527e-06,
+      "loss": 0.1558,
+      "step": 7445
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.945175438596492e-06,
+      "loss": 0.3875,
+      "step": 7446
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.944627192982456e-06,
+      "loss": 0.9119,
+      "step": 7447
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.944078947368422e-06,
+      "loss": 0.9774,
+      "step": 7448
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.943530701754386e-06,
+      "loss": 0.3783,
+      "step": 7449
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.942982456140351e-06,
+      "loss": 0.5203,
+      "step": 7450
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.942434210526317e-06,
+      "loss": 1.2067,
+      "step": 7451
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.941885964912281e-06,
+      "loss": 1.1627,
+      "step": 7452
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.941337719298246e-06,
+      "loss": 1.0031,
+      "step": 7453
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.9407894736842105e-06,
+      "loss": 0.2055,
+      "step": 7454
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.940241228070176e-06,
+      "loss": 0.8428,
+      "step": 7455
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.939692982456141e-06,
+      "loss": 0.5157,
+      "step": 7456
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.939144736842106e-06,
+      "loss": 0.7169,
+      "step": 7457
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.93859649122807e-06,
+      "loss": 0.2979,
+      "step": 7458
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.9380482456140355e-06,
+      "loss": 0.4038,
+      "step": 7459
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.9375e-06,
+      "loss": 0.6965,
+      "step": 7460
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.936951754385966e-06,
+      "loss": 0.3586,
+      "step": 7461
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.936403508771931e-06,
+      "loss": 0.6788,
+      "step": 7462
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.935855263157895e-06,
+      "loss": 0.7761,
+      "step": 7463
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.93530701754386e-06,
+      "loss": 0.7714,
+      "step": 7464
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.934758771929825e-06,
+      "loss": 0.3525,
+      "step": 7465
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.93421052631579e-06,
+      "loss": 0.3996,
+      "step": 7466
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.933662280701755e-06,
+      "loss": 1.0859,
+      "step": 7467
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.933114035087721e-06,
+      "loss": 0.7193,
+      "step": 7468
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.932565789473685e-06,
+      "loss": 0.5177,
+      "step": 7469
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.932017543859649e-06,
+      "loss": 0.3889,
+      "step": 7470
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.931469298245614e-06,
+      "loss": 1.3432,
+      "step": 7471
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.93092105263158e-06,
+      "loss": 1.0914,
+      "step": 7472
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.930372807017545e-06,
+      "loss": 0.4953,
+      "step": 7473
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.929824561403509e-06,
+      "loss": 0.997,
+      "step": 7474
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.9292763157894735e-06,
+      "loss": 0.8679,
+      "step": 7475
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.928728070175439e-06,
+      "loss": 0.4594,
+      "step": 7476
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.928179824561404e-06,
+      "loss": 1.1199,
+      "step": 7477
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.92763157894737e-06,
+      "loss": 0.3399,
+      "step": 7478
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.927083333333334e-06,
+      "loss": 0.5827,
+      "step": 7479
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.9265350877192985e-06,
+      "loss": 0.5959,
+      "step": 7480
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.925986842105263e-06,
+      "loss": 0.7118,
+      "step": 7481
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.925438596491229e-06,
+      "loss": 0.5812,
+      "step": 7482
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.924890350877194e-06,
+      "loss": 0.6677,
+      "step": 7483
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.924342105263159e-06,
+      "loss": 0.8106,
+      "step": 7484
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.923793859649123e-06,
+      "loss": 0.53,
+      "step": 7485
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.923245614035088e-06,
+      "loss": 0.6663,
+      "step": 7486
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.922697368421053e-06,
+      "loss": 0.6544,
+      "step": 7487
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.922149122807018e-06,
+      "loss": 1.0479,
+      "step": 7488
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.921600877192984e-06,
+      "loss": 0.2779,
+      "step": 7489
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.921052631578948e-06,
+      "loss": 0.1964,
+      "step": 7490
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.9205043859649124e-06,
+      "loss": 0.8516,
+      "step": 7491
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.919956140350877e-06,
+      "loss": 0.3484,
+      "step": 7492
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.919407894736843e-06,
+      "loss": 0.6571,
+      "step": 7493
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.918859649122808e-06,
+      "loss": 0.5949,
+      "step": 7494
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.918311403508772e-06,
+      "loss": 0.6313,
+      "step": 7495
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.917763157894737e-06,
+      "loss": 0.4342,
+      "step": 7496
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.917214912280702e-06,
+      "loss": 0.4797,
+      "step": 7497
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.916666666666667e-06,
+      "loss": 0.8765,
+      "step": 7498
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.916118421052633e-06,
+      "loss": 0.631,
+      "step": 7499
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.9155701754385975e-06,
+      "loss": 0.2679,
+      "step": 7500
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.9150219298245615e-06,
+      "loss": 0.9668,
+      "step": 7501
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.914473684210526e-06,
+      "loss": 0.7535,
+      "step": 7502
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.913925438596492e-06,
+      "loss": 0.48,
+      "step": 7503
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.913377192982457e-06,
+      "loss": 0.5965,
+      "step": 7504
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.912828947368422e-06,
+      "loss": 0.8082,
+      "step": 7505
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.912280701754386e-06,
+      "loss": 1.8578,
+      "step": 7506
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.911732456140351e-06,
+      "loss": 0.7559,
+      "step": 7507
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.911184210526316e-06,
+      "loss": 0.7088,
+      "step": 7508
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.910635964912281e-06,
+      "loss": 0.7603,
+      "step": 7509
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.910087719298247e-06,
+      "loss": 0.772,
+      "step": 7510
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.9095394736842115e-06,
+      "loss": 0.5121,
+      "step": 7511
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.9089912280701754e-06,
+      "loss": 0.7917,
+      "step": 7512
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.90844298245614e-06,
+      "loss": 1.1111,
+      "step": 7513
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.907894736842106e-06,
+      "loss": 0.7364,
+      "step": 7514
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.907346491228071e-06,
+      "loss": 0.6437,
+      "step": 7515
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.9067982456140364e-06,
+      "loss": 1.5828,
+      "step": 7516
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.90625e-06,
+      "loss": 0.4797,
+      "step": 7517
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.905701754385965e-06,
+      "loss": 1.2034,
+      "step": 7518
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.90515350877193e-06,
+      "loss": 0.4917,
+      "step": 7519
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.904605263157896e-06,
+      "loss": 0.308,
+      "step": 7520
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.9040570175438606e-06,
+      "loss": 0.5831,
+      "step": 7521
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.9035087719298245e-06,
+      "loss": 1.0459,
+      "step": 7522
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.902960526315789e-06,
+      "loss": 0.6442,
+      "step": 7523
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.902412280701755e-06,
+      "loss": 0.8867,
+      "step": 7524
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.90186403508772e-06,
+      "loss": 1.3977,
+      "step": 7525
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.901315789473685e-06,
+      "loss": 0.3163,
+      "step": 7526
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.90076754385965e-06,
+      "loss": 0.42,
+      "step": 7527
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.900219298245614e-06,
+      "loss": 0.6001,
+      "step": 7528
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.899671052631579e-06,
+      "loss": 0.5765,
+      "step": 7529
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.899122807017544e-06,
+      "loss": 0.9293,
+      "step": 7530
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.89857456140351e-06,
+      "loss": 0.3594,
+      "step": 7531
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.8980263157894745e-06,
+      "loss": 0.2092,
+      "step": 7532
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.8974780701754385e-06,
+      "loss": 0.4501,
+      "step": 7533
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.896929824561404e-06,
+      "loss": 0.6434,
+      "step": 7534
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.896381578947369e-06,
+      "loss": 0.9575,
+      "step": 7535
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.895833333333334e-06,
+      "loss": 0.3833,
+      "step": 7536
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.8952850877192994e-06,
+      "loss": 0.4385,
+      "step": 7537
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.8947368421052634e-06,
+      "loss": 0.5766,
+      "step": 7538
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.894188596491228e-06,
+      "loss": 0.6247,
+      "step": 7539
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.893640350877193e-06,
+      "loss": 0.3965,
+      "step": 7540
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.893092105263159e-06,
+      "loss": 0.4899,
+      "step": 7541
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.8925438596491236e-06,
+      "loss": 0.6383,
+      "step": 7542
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.891995614035088e-06,
+      "loss": 0.5627,
+      "step": 7543
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.891447368421052e-06,
+      "loss": 0.5772,
+      "step": 7544
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.890899122807018e-06,
+      "loss": 0.2957,
+      "step": 7545
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.890350877192983e-06,
+      "loss": 0.3584,
+      "step": 7546
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.889802631578948e-06,
+      "loss": 0.5508,
+      "step": 7547
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.889254385964913e-06,
+      "loss": 0.5975,
+      "step": 7548
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.888706140350877e-06,
+      "loss": 0.6309,
+      "step": 7549
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.888157894736842e-06,
+      "loss": 0.4855,
+      "step": 7550
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.887609649122807e-06,
+      "loss": 0.5734,
+      "step": 7551
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.887061403508773e-06,
+      "loss": 1.174,
+      "step": 7552
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.8865131578947375e-06,
+      "loss": 1.2594,
+      "step": 7553
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.885964912280703e-06,
+      "loss": 0.5617,
+      "step": 7554
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.885416666666667e-06,
+      "loss": 0.7836,
+      "step": 7555
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.884868421052632e-06,
+      "loss": 0.6524,
+      "step": 7556
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.884320175438597e-06,
+      "loss": 0.4971,
+      "step": 7557
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.8837719298245625e-06,
+      "loss": 0.4343,
+      "step": 7558
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.883223684210527e-06,
+      "loss": 0.8374,
+      "step": 7559
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.882675438596491e-06,
+      "loss": 0.4347,
+      "step": 7560
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.882127192982456e-06,
+      "loss": 0.2958,
+      "step": 7561
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.881578947368422e-06,
+      "loss": 0.764,
+      "step": 7562
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.881030701754387e-06,
+      "loss": 0.8216,
+      "step": 7563
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.880482456140351e-06,
+      "loss": 0.7562,
+      "step": 7564
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.879934210526315e-06,
+      "loss": 1.068,
+      "step": 7565
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.879385964912281e-06,
+      "loss": 0.961,
+      "step": 7566
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.878837719298246e-06,
+      "loss": 1.5514,
+      "step": 7567
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.878289473684211e-06,
+      "loss": 0.4136,
+      "step": 7568
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.877741228070176e-06,
+      "loss": 1.0373,
+      "step": 7569
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.877192982456141e-06,
+      "loss": 0.6175,
+      "step": 7570
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.876644736842105e-06,
+      "loss": 0.6748,
+      "step": 7571
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.876096491228071e-06,
+      "loss": 0.7728,
+      "step": 7572
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.875548245614036e-06,
+      "loss": 0.8749,
+      "step": 7573
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.8750000000000005e-06,
+      "loss": 0.4823,
+      "step": 7574
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.874451754385966e-06,
+      "loss": 1.0432,
+      "step": 7575
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.87390350877193e-06,
+      "loss": 0.7715,
+      "step": 7576
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.873355263157895e-06,
+      "loss": 0.7733,
+      "step": 7577
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.87280701754386e-06,
+      "loss": 1.0565,
+      "step": 7578
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.8722587719298255e-06,
+      "loss": 0.7384,
+      "step": 7579
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.87171052631579e-06,
+      "loss": 1.1635,
+      "step": 7580
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.871162280701754e-06,
+      "loss": 0.8925,
+      "step": 7581
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.870614035087719e-06,
+      "loss": 0.7182,
+      "step": 7582
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.870065789473685e-06,
+      "loss": 0.773,
+      "step": 7583
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.86951754385965e-06,
+      "loss": 0.4971,
+      "step": 7584
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.868969298245614e-06,
+      "loss": 1.0585,
+      "step": 7585
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.86842105263158e-06,
+      "loss": 0.6062,
+      "step": 7586
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.867872807017544e-06,
+      "loss": 0.8174,
+      "step": 7587
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.867324561403509e-06,
+      "loss": 0.7151,
+      "step": 7588
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.8667763157894746e-06,
+      "loss": 1.0167,
+      "step": 7589
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.866228070175439e-06,
+      "loss": 0.3791,
+      "step": 7590
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.865679824561404e-06,
+      "loss": 0.7776,
+      "step": 7591
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.865131578947368e-06,
+      "loss": 0.4778,
+      "step": 7592
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.864583333333334e-06,
+      "loss": 1.0427,
+      "step": 7593
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.864035087719299e-06,
+      "loss": 0.8524,
+      "step": 7594
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.8634868421052635e-06,
+      "loss": 1.391,
+      "step": 7595
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.862938596491229e-06,
+      "loss": 0.8882,
+      "step": 7596
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.862390350877194e-06,
+      "loss": 0.3396,
+      "step": 7597
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.861842105263158e-06,
+      "loss": 0.6756,
+      "step": 7598
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.861293859649123e-06,
+      "loss": 1.2568,
+      "step": 7599
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.8607456140350885e-06,
+      "loss": 1.4166,
+      "step": 7600
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.860197368421053e-06,
+      "loss": 0.2841,
+      "step": 7601
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.859649122807018e-06,
+      "loss": 1.1339,
+      "step": 7602
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.859100877192982e-06,
+      "loss": 0.3859,
+      "step": 7603
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.858552631578948e-06,
+      "loss": 0.8586,
+      "step": 7604
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.858004385964913e-06,
+      "loss": 0.6455,
+      "step": 7605
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.8574561403508774e-06,
+      "loss": 0.5353,
+      "step": 7606
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.856907894736843e-06,
+      "loss": 0.2869,
+      "step": 7607
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.856359649122807e-06,
+      "loss": 0.4608,
+      "step": 7608
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.855811403508772e-06,
+      "loss": 0.4833,
+      "step": 7609
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.855263157894738e-06,
+      "loss": 0.8945,
+      "step": 7610
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.854714912280702e-06,
+      "loss": 0.7789,
+      "step": 7611
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.854166666666667e-06,
+      "loss": 0.7034,
+      "step": 7612
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.853618421052633e-06,
+      "loss": 1.1181,
+      "step": 7613
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.853070175438597e-06,
+      "loss": 0.409,
+      "step": 7614
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.852521929824562e-06,
+      "loss": 0.6552,
+      "step": 7615
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.8519736842105265e-06,
+      "loss": 1.5271,
+      "step": 7616
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.851425438596492e-06,
+      "loss": 0.1473,
+      "step": 7617
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.850877192982457e-06,
+      "loss": 0.9886,
+      "step": 7618
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.850328947368421e-06,
+      "loss": 0.4386,
+      "step": 7619
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.849780701754386e-06,
+      "loss": 0.596,
+      "step": 7620
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.8492324561403515e-06,
+      "loss": 0.6809,
+      "step": 7621
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.848684210526316e-06,
+      "loss": 0.536,
+      "step": 7622
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.848135964912281e-06,
+      "loss": 0.9311,
+      "step": 7623
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.847587719298247e-06,
+      "loss": 0.572,
+      "step": 7624
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.847039473684211e-06,
+      "loss": 0.4985,
+      "step": 7625
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.846491228070176e-06,
+      "loss": 0.8969,
+      "step": 7626
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.845942982456141e-06,
+      "loss": 0.5532,
+      "step": 7627
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.845394736842106e-06,
+      "loss": 0.3838,
+      "step": 7628
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.844846491228071e-06,
+      "loss": 0.7006,
+      "step": 7629
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.844298245614035e-06,
+      "loss": 0.9688,
+      "step": 7630
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.843750000000001e-06,
+      "loss": 1.059,
+      "step": 7631
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.843201754385965e-06,
+      "loss": 0.8316,
+      "step": 7632
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.84265350877193e-06,
+      "loss": 1.4024,
+      "step": 7633
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.842105263157896e-06,
+      "loss": 1.6123,
+      "step": 7634
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.84155701754386e-06,
+      "loss": 0.8437,
+      "step": 7635
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.841008771929825e-06,
+      "loss": 1.228,
+      "step": 7636
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.8404605263157895e-06,
+      "loss": 0.4097,
+      "step": 7637
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.839912280701755e-06,
+      "loss": 1.1048,
+      "step": 7638
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.83936403508772e-06,
+      "loss": 0.8948,
+      "step": 7639
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.838815789473685e-06,
+      "loss": 0.5938,
+      "step": 7640
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.838267543859649e-06,
+      "loss": 1.3173,
+      "step": 7641
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.8377192982456145e-06,
+      "loss": 0.254,
+      "step": 7642
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.837171052631579e-06,
+      "loss": 0.3349,
+      "step": 7643
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.836622807017544e-06,
+      "loss": 0.4498,
+      "step": 7644
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.83607456140351e-06,
+      "loss": 0.9134,
+      "step": 7645
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.835526315789474e-06,
+      "loss": 0.7094,
+      "step": 7646
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.834978070175439e-06,
+      "loss": 0.7797,
+      "step": 7647
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.834429824561404e-06,
+      "loss": 0.42,
+      "step": 7648
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.833881578947369e-06,
+      "loss": 0.6408,
+      "step": 7649
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.833333333333334e-06,
+      "loss": 0.9129,
+      "step": 7650
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.832785087719298e-06,
+      "loss": 0.4505,
+      "step": 7651
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.832236842105264e-06,
+      "loss": 0.7775,
+      "step": 7652
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.8316885964912284e-06,
+      "loss": 0.1065,
+      "step": 7653
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.831140350877193e-06,
+      "loss": 0.5058,
+      "step": 7654
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.830592105263159e-06,
+      "loss": 0.3815,
+      "step": 7655
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.830043859649124e-06,
+      "loss": 0.977,
+      "step": 7656
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.829495614035088e-06,
+      "loss": 1.0518,
+      "step": 7657
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.8289473684210525e-06,
+      "loss": 0.3939,
+      "step": 7658
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.828399122807018e-06,
+      "loss": 0.3666,
+      "step": 7659
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.827850877192983e-06,
+      "loss": 0.908,
+      "step": 7660
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.827302631578948e-06,
+      "loss": 0.6952,
+      "step": 7661
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.826754385964913e-06,
+      "loss": 0.8699,
+      "step": 7662
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.8262061403508775e-06,
+      "loss": 0.2695,
+      "step": 7663
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.825657894736842e-06,
+      "loss": 0.5879,
+      "step": 7664
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.825109649122808e-06,
+      "loss": 0.5951,
+      "step": 7665
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.824561403508773e-06,
+      "loss": 0.6768,
+      "step": 7666
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.824013157894738e-06,
+      "loss": 1.0959,
+      "step": 7667
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.823464912280702e-06,
+      "loss": 0.505,
+      "step": 7668
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.822916666666667e-06,
+      "loss": 0.8208,
+      "step": 7669
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.822368421052632e-06,
+      "loss": 0.7672,
+      "step": 7670
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.821820175438597e-06,
+      "loss": 0.3779,
+      "step": 7671
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.821271929824563e-06,
+      "loss": 0.5482,
+      "step": 7672
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.820723684210527e-06,
+      "loss": 0.9255,
+      "step": 7673
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.8201754385964914e-06,
+      "loss": 0.9443,
+      "step": 7674
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.819627192982456e-06,
+      "loss": 0.6629,
+      "step": 7675
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.819078947368422e-06,
+      "loss": 0.7475,
+      "step": 7676
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.818530701754387e-06,
+      "loss": 0.2126,
+      "step": 7677
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.817982456140351e-06,
+      "loss": 0.7718,
+      "step": 7678
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.8174342105263156e-06,
+      "loss": 1.1435,
+      "step": 7679
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.816885964912281e-06,
+      "loss": 0.7272,
+      "step": 7680
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.816337719298246e-06,
+      "loss": 0.5543,
+      "step": 7681
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.815789473684212e-06,
+      "loss": 0.5729,
+      "step": 7682
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.8152412280701766e-06,
+      "loss": 0.9853,
+      "step": 7683
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.8146929824561405e-06,
+      "loss": 0.7586,
+      "step": 7684
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.814144736842105e-06,
+      "loss": 0.3105,
+      "step": 7685
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.813596491228071e-06,
+      "loss": 0.2452,
+      "step": 7686
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.813048245614036e-06,
+      "loss": 1.5025,
+      "step": 7687
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.812500000000001e-06,
+      "loss": 0.4628,
+      "step": 7688
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.811951754385965e-06,
+      "loss": 0.4611,
+      "step": 7689
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.81140350877193e-06,
+      "loss": 0.3048,
+      "step": 7690
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.810855263157895e-06,
+      "loss": 0.9671,
+      "step": 7691
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.81030701754386e-06,
+      "loss": 0.6712,
+      "step": 7692
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.809758771929826e-06,
+      "loss": 0.8239,
+      "step": 7693
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.80921052631579e-06,
+      "loss": 0.5597,
+      "step": 7694
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.8086622807017544e-06,
+      "loss": 0.4471,
+      "step": 7695
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.808114035087719e-06,
+      "loss": 0.5211,
+      "step": 7696
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.807565789473685e-06,
+      "loss": 1.1183,
+      "step": 7697
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.80701754385965e-06,
+      "loss": 0.4826,
+      "step": 7698
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.806469298245615e-06,
+      "loss": 0.8227,
+      "step": 7699
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.805921052631579e-06,
+      "loss": 0.6534,
+      "step": 7700
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.805372807017544e-06,
+      "loss": 1.5298,
+      "step": 7701
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.804824561403509e-06,
+      "loss": 0.8891,
+      "step": 7702
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.804276315789475e-06,
+      "loss": 0.7899,
+      "step": 7703
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.8037280701754396e-06,
+      "loss": 0.3235,
+      "step": 7704
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.8031798245614035e-06,
+      "loss": 0.4668,
+      "step": 7705
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.802631578947368e-06,
+      "loss": 0.3734,
+      "step": 7706
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.802083333333334e-06,
+      "loss": 0.2184,
+      "step": 7707
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.801535087719299e-06,
+      "loss": 0.7317,
+      "step": 7708
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.800986842105264e-06,
+      "loss": 0.2456,
+      "step": 7709
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.800438596491229e-06,
+      "loss": 0.8776,
+      "step": 7710
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.799890350877193e-06,
+      "loss": 0.6745,
+      "step": 7711
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.799342105263158e-06,
+      "loss": 0.9175,
+      "step": 7712
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.798793859649123e-06,
+      "loss": 0.5431,
+      "step": 7713
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.798245614035089e-06,
+      "loss": 0.8446,
+      "step": 7714
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.7976973684210535e-06,
+      "loss": 0.8921,
+      "step": 7715
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.7971491228070175e-06,
+      "loss": 0.4037,
+      "step": 7716
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.796600877192982e-06,
+      "loss": 0.4472,
+      "step": 7717
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.796052631578948e-06,
+      "loss": 0.2345,
+      "step": 7718
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.795504385964913e-06,
+      "loss": 1.2468,
+      "step": 7719
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.7949561403508785e-06,
+      "loss": 0.7775,
+      "step": 7720
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.7944078947368424e-06,
+      "loss": 0.6726,
+      "step": 7721
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.793859649122807e-06,
+      "loss": 0.5347,
+      "step": 7722
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.793311403508772e-06,
+      "loss": 0.1782,
+      "step": 7723
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.792763157894738e-06,
+      "loss": 1.0006,
+      "step": 7724
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.792214912280703e-06,
+      "loss": 0.6784,
+      "step": 7725
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.791666666666667e-06,
+      "loss": 0.441,
+      "step": 7726
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.791118421052631e-06,
+      "loss": 1.0421,
+      "step": 7727
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.790570175438597e-06,
+      "loss": 0.5205,
+      "step": 7728
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.790021929824562e-06,
+      "loss": 0.5459,
+      "step": 7729
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.789473684210527e-06,
+      "loss": 0.5375,
+      "step": 7730
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.788925438596492e-06,
+      "loss": 0.4849,
+      "step": 7731
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.788377192982456e-06,
+      "loss": 1.2425,
+      "step": 7732
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.787828947368421e-06,
+      "loss": 0.8525,
+      "step": 7733
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.787280701754386e-06,
+      "loss": 0.2837,
+      "step": 7734
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.786732456140352e-06,
+      "loss": 0.7217,
+      "step": 7735
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.7861842105263165e-06,
+      "loss": 0.7773,
+      "step": 7736
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.7856359649122805e-06,
+      "loss": 0.5544,
+      "step": 7737
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.785087719298246e-06,
+      "loss": 0.9362,
+      "step": 7738
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.784539473684211e-06,
+      "loss": 0.4053,
+      "step": 7739
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.783991228070176e-06,
+      "loss": 0.6717,
+      "step": 7740
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.7834429824561415e-06,
+      "loss": 1.1041,
+      "step": 7741
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.782894736842106e-06,
+      "loss": 0.6849,
+      "step": 7742
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.78234649122807e-06,
+      "loss": 0.311,
+      "step": 7743
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.781798245614035e-06,
+      "loss": 1.014,
+      "step": 7744
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.781250000000001e-06,
+      "loss": 0.4679,
+      "step": 7745
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.780701754385966e-06,
+      "loss": 0.729,
+      "step": 7746
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.78015350877193e-06,
+      "loss": 0.3852,
+      "step": 7747
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.779605263157894e-06,
+      "loss": 0.7312,
+      "step": 7748
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.77905701754386e-06,
+      "loss": 1.6346,
+      "step": 7749
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.778508771929825e-06,
+      "loss": 0.4844,
+      "step": 7750
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.77796052631579e-06,
+      "loss": 0.9933,
+      "step": 7751
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.777412280701755e-06,
+      "loss": 0.8657,
+      "step": 7752
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.77686403508772e-06,
+      "loss": 1.2497,
+      "step": 7753
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.776315789473684e-06,
+      "loss": 0.7401,
+      "step": 7754
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.77576754385965e-06,
+      "loss": 0.2405,
+      "step": 7755
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.775219298245615e-06,
+      "loss": 1.1355,
+      "step": 7756
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.7746710526315795e-06,
+      "loss": 0.8683,
+      "step": 7757
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.774122807017545e-06,
+      "loss": 1.2079,
+      "step": 7758
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.773574561403509e-06,
+      "loss": 1.1159,
+      "step": 7759
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.773026315789474e-06,
+      "loss": 0.8726,
+      "step": 7760
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.772478070175439e-06,
+      "loss": 0.662,
+      "step": 7761
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.7719298245614045e-06,
+      "loss": 1.2246,
+      "step": 7762
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.771381578947369e-06,
+      "loss": 0.4022,
+      "step": 7763
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.770833333333333e-06,
+      "loss": 1.5396,
+      "step": 7764
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.770285087719298e-06,
+      "loss": 0.5696,
+      "step": 7765
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.769736842105264e-06,
+      "loss": 0.4759,
+      "step": 7766
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.769188596491229e-06,
+      "loss": 0.7909,
+      "step": 7767
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.7686403508771934e-06,
+      "loss": 0.6015,
+      "step": 7768
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.768092105263159e-06,
+      "loss": 0.5247,
+      "step": 7769
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.767543859649123e-06,
+      "loss": 1.2698,
+      "step": 7770
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.766995614035088e-06,
+      "loss": 0.8106,
+      "step": 7771
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.766447368421053e-06,
+      "loss": 0.8976,
+      "step": 7772
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.765899122807018e-06,
+      "loss": 1.2001,
+      "step": 7773
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.765350877192983e-06,
+      "loss": 0.379,
+      "step": 7774
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.764802631578947e-06,
+      "loss": 0.668,
+      "step": 7775
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.764254385964913e-06,
+      "loss": 1.0227,
+      "step": 7776
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.763706140350878e-06,
+      "loss": 0.3116,
+      "step": 7777
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.7631578947368425e-06,
+      "loss": 0.6225,
+      "step": 7778
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.762609649122808e-06,
+      "loss": 0.674,
+      "step": 7779
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.762061403508772e-06,
+      "loss": 0.9609,
+      "step": 7780
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.761513157894737e-06,
+      "loss": 0.5701,
+      "step": 7781
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.760964912280702e-06,
+      "loss": 0.8947,
+      "step": 7782
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.7604166666666675e-06,
+      "loss": 1.0136,
+      "step": 7783
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.759868421052632e-06,
+      "loss": 0.2409,
+      "step": 7784
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.759320175438597e-06,
+      "loss": 0.1925,
+      "step": 7785
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.758771929824561e-06,
+      "loss": 0.2798,
+      "step": 7786
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.758223684210527e-06,
+      "loss": 0.221,
+      "step": 7787
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.757675438596492e-06,
+      "loss": 0.4376,
+      "step": 7788
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.7571271929824564e-06,
+      "loss": 1.3123,
+      "step": 7789
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.756578947368422e-06,
+      "loss": 1.2749,
+      "step": 7790
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.756030701754386e-06,
+      "loss": 0.3897,
+      "step": 7791
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.755482456140351e-06,
+      "loss": 0.923,
+      "step": 7792
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.754934210526317e-06,
+      "loss": 0.5859,
+      "step": 7793
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.754385964912281e-06,
+      "loss": 0.9442,
+      "step": 7794
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.753837719298246e-06,
+      "loss": 0.8873,
+      "step": 7795
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.753289473684212e-06,
+      "loss": 0.7232,
+      "step": 7796
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.752741228070176e-06,
+      "loss": 0.164,
+      "step": 7797
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.752192982456141e-06,
+      "loss": 0.4273,
+      "step": 7798
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.7516447368421055e-06,
+      "loss": 0.7704,
+      "step": 7799
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.751096491228071e-06,
+      "loss": 0.8454,
+      "step": 7800
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.750548245614036e-06,
+      "loss": 0.7595,
+      "step": 7801
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.75e-06,
+      "loss": 0.156,
+      "step": 7802
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.749451754385965e-06,
+      "loss": 0.5533,
+      "step": 7803
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.7489035087719305e-06,
+      "loss": 0.2456,
+      "step": 7804
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.748355263157895e-06,
+      "loss": 0.4987,
+      "step": 7805
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.74780701754386e-06,
+      "loss": 0.5304,
+      "step": 7806
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.747258771929824e-06,
+      "loss": 0.4329,
+      "step": 7807
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.74671052631579e-06,
+      "loss": 0.7048,
+      "step": 7808
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.746162280701755e-06,
+      "loss": 0.6054,
+      "step": 7809
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.7456140350877194e-06,
+      "loss": 1.7506,
+      "step": 7810
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.745065789473685e-06,
+      "loss": 0.8418,
+      "step": 7811
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.74451754385965e-06,
+      "loss": 0.5047,
+      "step": 7812
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.743969298245614e-06,
+      "loss": 0.2064,
+      "step": 7813
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.74342105263158e-06,
+      "loss": 1.0213,
+      "step": 7814
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.742872807017544e-06,
+      "loss": 0.4464,
+      "step": 7815
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.742324561403509e-06,
+      "loss": 0.4889,
+      "step": 7816
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.741776315789475e-06,
+      "loss": 0.7549,
+      "step": 7817
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.741228070175439e-06,
+      "loss": 0.9449,
+      "step": 7818
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.740679824561404e-06,
+      "loss": 0.2424,
+      "step": 7819
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.7401315789473685e-06,
+      "loss": 1.05,
+      "step": 7820
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.739583333333334e-06,
+      "loss": 0.4015,
+      "step": 7821
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.739035087719299e-06,
+      "loss": 0.8633,
+      "step": 7822
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.738486842105263e-06,
+      "loss": 0.8717,
+      "step": 7823
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.737938596491228e-06,
+      "loss": 1.5039,
+      "step": 7824
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.7373903508771935e-06,
+      "loss": 0.1342,
+      "step": 7825
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.736842105263158e-06,
+      "loss": 0.3786,
+      "step": 7826
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.736293859649123e-06,
+      "loss": 0.6758,
+      "step": 7827
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.735745614035089e-06,
+      "loss": 0.8668,
+      "step": 7828
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.735197368421053e-06,
+      "loss": 0.7598,
+      "step": 7829
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.734649122807018e-06,
+      "loss": 0.547,
+      "step": 7830
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.734100877192983e-06,
+      "loss": 0.8388,
+      "step": 7831
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.733552631578948e-06,
+      "loss": 0.7021,
+      "step": 7832
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.733004385964913e-06,
+      "loss": 0.7319,
+      "step": 7833
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.732456140350877e-06,
+      "loss": 0.8806,
+      "step": 7834
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.731907894736843e-06,
+      "loss": 0.8244,
+      "step": 7835
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.7313596491228074e-06,
+      "loss": 0.5323,
+      "step": 7836
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.730811403508772e-06,
+      "loss": 0.2485,
+      "step": 7837
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.730263157894738e-06,
+      "loss": 1.0535,
+      "step": 7838
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.729714912280703e-06,
+      "loss": 0.2141,
+      "step": 7839
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.729166666666667e-06,
+      "loss": 1.6011,
+      "step": 7840
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.7286184210526316e-06,
+      "loss": 0.3178,
+      "step": 7841
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.728070175438597e-06,
+      "loss": 0.9352,
+      "step": 7842
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.727521929824562e-06,
+      "loss": 0.3816,
+      "step": 7843
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.726973684210527e-06,
+      "loss": 1.4203,
+      "step": 7844
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.726425438596491e-06,
+      "loss": 0.4623,
+      "step": 7845
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.7258771929824565e-06,
+      "loss": 0.681,
+      "step": 7846
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.725328947368421e-06,
+      "loss": 0.5364,
+      "step": 7847
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.724780701754387e-06,
+      "loss": 0.9077,
+      "step": 7848
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.724232456140352e-06,
+      "loss": 0.8065,
+      "step": 7849
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.723684210526316e-06,
+      "loss": 1.0986,
+      "step": 7850
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.723135964912281e-06,
+      "loss": 0.4495,
+      "step": 7851
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.722587719298246e-06,
+      "loss": 0.3526,
+      "step": 7852
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.722039473684211e-06,
+      "loss": 0.3485,
+      "step": 7853
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.721491228070176e-06,
+      "loss": 0.6437,
+      "step": 7854
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.720942982456142e-06,
+      "loss": 0.7492,
+      "step": 7855
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.720394736842106e-06,
+      "loss": 0.9877,
+      "step": 7856
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.7198464912280704e-06,
+      "loss": 0.7161,
+      "step": 7857
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.719298245614035e-06,
+      "loss": 0.4474,
+      "step": 7858
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.718750000000001e-06,
+      "loss": 0.4949,
+      "step": 7859
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.718201754385966e-06,
+      "loss": 0.7176,
+      "step": 7860
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.71765350877193e-06,
+      "loss": 1.5269,
+      "step": 7861
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.7171052631578946e-06,
+      "loss": 0.5557,
+      "step": 7862
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.71655701754386e-06,
+      "loss": 0.5235,
+      "step": 7863
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.716008771929825e-06,
+      "loss": 0.5501,
+      "step": 7864
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.71546052631579e-06,
+      "loss": 0.3884,
+      "step": 7865
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.714912280701755e-06,
+      "loss": 0.3134,
+      "step": 7866
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.7143640350877195e-06,
+      "loss": 0.6131,
+      "step": 7867
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.713815789473684e-06,
+      "loss": 0.5679,
+      "step": 7868
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.71326754385965e-06,
+      "loss": 0.4334,
+      "step": 7869
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.712719298245615e-06,
+      "loss": 0.6801,
+      "step": 7870
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.71217105263158e-06,
+      "loss": 1.0838,
+      "step": 7871
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.711622807017544e-06,
+      "loss": 1.5386,
+      "step": 7872
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.711074561403509e-06,
+      "loss": 0.3465,
+      "step": 7873
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.710526315789474e-06,
+      "loss": 0.7139,
+      "step": 7874
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.709978070175439e-06,
+      "loss": 2.0228,
+      "step": 7875
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.709429824561405e-06,
+      "loss": 0.9459,
+      "step": 7876
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.708881578947369e-06,
+      "loss": 1.2566,
+      "step": 7877
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.7083333333333335e-06,
+      "loss": 0.3772,
+      "step": 7878
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.707785087719298e-06,
+      "loss": 1.0787,
+      "step": 7879
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.707236842105264e-06,
+      "loss": 0.9452,
+      "step": 7880
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.706688596491229e-06,
+      "loss": 0.6609,
+      "step": 7881
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.706140350877194e-06,
+      "loss": 0.5236,
+      "step": 7882
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.705592105263158e-06,
+      "loss": 0.4555,
+      "step": 7883
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.705043859649123e-06,
+      "loss": 0.6759,
+      "step": 7884
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.704495614035088e-06,
+      "loss": 0.8668,
+      "step": 7885
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.703947368421054e-06,
+      "loss": 1.1156,
+      "step": 7886
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.7033991228070186e-06,
+      "loss": 1.3472,
+      "step": 7887
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.7028508771929825e-06,
+      "loss": 0.4529,
+      "step": 7888
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.702302631578947e-06,
+      "loss": 0.9969,
+      "step": 7889
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.701754385964913e-06,
+      "loss": 0.7771,
+      "step": 7890
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.701206140350878e-06,
+      "loss": 0.9732,
+      "step": 7891
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.700657894736843e-06,
+      "loss": 0.6826,
+      "step": 7892
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.700109649122807e-06,
+      "loss": 1.0346,
+      "step": 7893
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.699561403508772e-06,
+      "loss": 0.4485,
+      "step": 7894
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.699013157894737e-06,
+      "loss": 0.6374,
+      "step": 7895
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.698464912280702e-06,
+      "loss": 0.4582,
+      "step": 7896
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.697916666666668e-06,
+      "loss": 0.6991,
+      "step": 7897
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.6973684210526325e-06,
+      "loss": 0.4973,
+      "step": 7898
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.6968201754385965e-06,
+      "loss": 1.4631,
+      "step": 7899
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.696271929824561e-06,
+      "loss": 1.13,
+      "step": 7900
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.695723684210527e-06,
+      "loss": 0.7495,
+      "step": 7901
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.695175438596492e-06,
+      "loss": 0.3263,
+      "step": 7902
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.694627192982457e-06,
+      "loss": 0.7588,
+      "step": 7903
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.6940789473684214e-06,
+      "loss": 0.5256,
+      "step": 7904
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.693530701754386e-06,
+      "loss": 1.2809,
+      "step": 7905
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.692982456140351e-06,
+      "loss": 0.8874,
+      "step": 7906
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.692434210526317e-06,
+      "loss": 0.9828,
+      "step": 7907
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.691885964912282e-06,
+      "loss": 0.6112,
+      "step": 7908
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.691337719298246e-06,
+      "loss": 0.6808,
+      "step": 7909
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.69078947368421e-06,
+      "loss": 0.3103,
+      "step": 7910
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.690241228070176e-06,
+      "loss": 0.6648,
+      "step": 7911
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.689692982456141e-06,
+      "loss": 0.5734,
+      "step": 7912
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.689144736842106e-06,
+      "loss": 0.6765,
+      "step": 7913
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.688596491228071e-06,
+      "loss": 0.2129,
+      "step": 7914
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.688048245614035e-06,
+      "loss": 0.5214,
+      "step": 7915
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.6875e-06,
+      "loss": 0.3574,
+      "step": 7916
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.686951754385965e-06,
+      "loss": 0.263,
+      "step": 7917
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.686403508771931e-06,
+      "loss": 0.2346,
+      "step": 7918
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.6858552631578955e-06,
+      "loss": 0.9315,
+      "step": 7919
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.6853070175438595e-06,
+      "loss": 0.3686,
+      "step": 7920
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.684758771929825e-06,
+      "loss": 0.936,
+      "step": 7921
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.68421052631579e-06,
+      "loss": 1.022,
+      "step": 7922
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.683662280701755e-06,
+      "loss": 0.7465,
+      "step": 7923
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.6831140350877205e-06,
+      "loss": 1.2637,
+      "step": 7924
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.682565789473685e-06,
+      "loss": 0.4548,
+      "step": 7925
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.682017543859649e-06,
+      "loss": 0.7674,
+      "step": 7926
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.681469298245614e-06,
+      "loss": 0.6311,
+      "step": 7927
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.68092105263158e-06,
+      "loss": 0.3487,
+      "step": 7928
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.680372807017545e-06,
+      "loss": 1.4296,
+      "step": 7929
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.679824561403509e-06,
+      "loss": 0.6537,
+      "step": 7930
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.679276315789473e-06,
+      "loss": 0.4116,
+      "step": 7931
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.678728070175439e-06,
+      "loss": 1.1852,
+      "step": 7932
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.678179824561404e-06,
+      "loss": 0.4521,
+      "step": 7933
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.677631578947369e-06,
+      "loss": 0.3493,
+      "step": 7934
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.677083333333334e-06,
+      "loss": 0.5755,
+      "step": 7935
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.676535087719298e-06,
+      "loss": 0.6869,
+      "step": 7936
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.675986842105263e-06,
+      "loss": 0.5248,
+      "step": 7937
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.675438596491228e-06,
+      "loss": 0.8371,
+      "step": 7938
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.674890350877194e-06,
+      "loss": 0.1678,
+      "step": 7939
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.6743421052631585e-06,
+      "loss": 0.2497,
+      "step": 7940
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.673793859649124e-06,
+      "loss": 0.4411,
+      "step": 7941
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.673245614035088e-06,
+      "loss": 0.4505,
+      "step": 7942
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.672697368421053e-06,
+      "loss": 0.6628,
+      "step": 7943
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.672149122807018e-06,
+      "loss": 0.546,
+      "step": 7944
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.6716008771929835e-06,
+      "loss": 0.617,
+      "step": 7945
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.671052631578948e-06,
+      "loss": 1.621,
+      "step": 7946
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.670504385964912e-06,
+      "loss": 0.9239,
+      "step": 7947
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.669956140350877e-06,
+      "loss": 0.526,
+      "step": 7948
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.669407894736843e-06,
+      "loss": 0.2294,
+      "step": 7949
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.668859649122808e-06,
+      "loss": 1.3304,
+      "step": 7950
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.6683114035087724e-06,
+      "loss": 0.2695,
+      "step": 7951
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.667763157894738e-06,
+      "loss": 0.6585,
+      "step": 7952
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.667214912280702e-06,
+      "loss": 0.3471,
+      "step": 7953
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.666666666666667e-06,
+      "loss": 0.2074,
+      "step": 7954
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.666118421052632e-06,
+      "loss": 0.4411,
+      "step": 7955
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.665570175438597e-06,
+      "loss": 0.6769,
+      "step": 7956
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.665021929824562e-06,
+      "loss": 1.1034,
+      "step": 7957
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.664473684210526e-06,
+      "loss": 1.026,
+      "step": 7958
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.663925438596492e-06,
+      "loss": 0.4323,
+      "step": 7959
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.663377192982457e-06,
+      "loss": 0.6423,
+      "step": 7960
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.6628289473684215e-06,
+      "loss": 0.4141,
+      "step": 7961
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.662280701754387e-06,
+      "loss": 0.9061,
+      "step": 7962
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.661732456140351e-06,
+      "loss": 0.7095,
+      "step": 7963
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.661184210526316e-06,
+      "loss": 0.8373,
+      "step": 7964
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.660635964912281e-06,
+      "loss": 0.8892,
+      "step": 7965
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.6600877192982465e-06,
+      "loss": 0.7798,
+      "step": 7966
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.659539473684211e-06,
+      "loss": 0.7032,
+      "step": 7967
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.658991228070176e-06,
+      "loss": 0.6782,
+      "step": 7968
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.65844298245614e-06,
+      "loss": 1.0119,
+      "step": 7969
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.657894736842106e-06,
+      "loss": 0.7103,
+      "step": 7970
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.657346491228071e-06,
+      "loss": 0.5512,
+      "step": 7971
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.6567982456140354e-06,
+      "loss": 0.6401,
+      "step": 7972
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.656250000000001e-06,
+      "loss": 0.6859,
+      "step": 7973
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.655701754385965e-06,
+      "loss": 0.5209,
+      "step": 7974
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.65515350877193e-06,
+      "loss": 1.2365,
+      "step": 7975
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.654605263157895e-06,
+      "loss": 0.4113,
+      "step": 7976
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.65405701754386e-06,
+      "loss": 0.8771,
+      "step": 7977
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.653508771929825e-06,
+      "loss": 1.0122,
+      "step": 7978
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.652960526315789e-06,
+      "loss": 0.5898,
+      "step": 7979
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.652412280701755e-06,
+      "loss": 0.2693,
+      "step": 7980
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.65186403508772e-06,
+      "loss": 0.4684,
+      "step": 7981
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.6513157894736845e-06,
+      "loss": 0.3468,
+      "step": 7982
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.65076754385965e-06,
+      "loss": 0.7835,
+      "step": 7983
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.650219298245615e-06,
+      "loss": 0.454,
+      "step": 7984
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.649671052631579e-06,
+      "loss": 0.6683,
+      "step": 7985
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.649122807017544e-06,
+      "loss": 1.1203,
+      "step": 7986
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.6485745614035095e-06,
+      "loss": 0.2281,
+      "step": 7987
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.648026315789474e-06,
+      "loss": 0.2299,
+      "step": 7988
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.647478070175439e-06,
+      "loss": 0.8591,
+      "step": 7989
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.646929824561403e-06,
+      "loss": 1.5674,
+      "step": 7990
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.646381578947369e-06,
+      "loss": 1.1585,
+      "step": 7991
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.645833333333334e-06,
+      "loss": 0.3675,
+      "step": 7992
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.6452850877192985e-06,
+      "loss": 0.7783,
+      "step": 7993
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.644736842105264e-06,
+      "loss": 1.34,
+      "step": 7994
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.644188596491229e-06,
+      "loss": 0.8916,
+      "step": 7995
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.643640350877193e-06,
+      "loss": 0.8422,
+      "step": 7996
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.643092105263159e-06,
+      "loss": 0.7493,
+      "step": 7997
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.6425438596491234e-06,
+      "loss": 0.3222,
+      "step": 7998
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.641995614035088e-06,
+      "loss": 0.3897,
+      "step": 7999
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.641447368421054e-06,
+      "loss": 0.8211,
+      "step": 8000
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.640899122807018e-06,
+      "loss": 0.6553,
+      "step": 8001
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.640350877192983e-06,
+      "loss": 0.4806,
+      "step": 8002
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.6398026315789475e-06,
+      "loss": 0.8615,
+      "step": 8003
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.639254385964913e-06,
+      "loss": 0.4603,
+      "step": 8004
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.638706140350878e-06,
+      "loss": 0.5769,
+      "step": 8005
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.638157894736842e-06,
+      "loss": 0.6565,
+      "step": 8006
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.637609649122807e-06,
+      "loss": 1.4364,
+      "step": 8007
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.6370614035087725e-06,
+      "loss": 0.6029,
+      "step": 8008
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.636513157894737e-06,
+      "loss": 0.6455,
+      "step": 8009
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.635964912280702e-06,
+      "loss": 0.9169,
+      "step": 8010
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.635416666666668e-06,
+      "loss": 1.7349,
+      "step": 8011
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.634868421052632e-06,
+      "loss": 0.7538,
+      "step": 8012
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.634320175438597e-06,
+      "loss": 1.2461,
+      "step": 8013
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.633771929824562e-06,
+      "loss": 0.7192,
+      "step": 8014
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.633223684210527e-06,
+      "loss": 0.4276,
+      "step": 8015
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.632675438596492e-06,
+      "loss": 0.7972,
+      "step": 8016
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.632127192982456e-06,
+      "loss": 0.6855,
+      "step": 8017
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.631578947368422e-06,
+      "loss": 0.6451,
+      "step": 8018
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.6310307017543864e-06,
+      "loss": 1.0946,
+      "step": 8019
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.630482456140351e-06,
+      "loss": 0.7706,
+      "step": 8020
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.629934210526317e-06,
+      "loss": 0.6666,
+      "step": 8021
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.629385964912281e-06,
+      "loss": 0.3498,
+      "step": 8022
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.628837719298246e-06,
+      "loss": 0.9411,
+      "step": 8023
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.6282894736842106e-06,
+      "loss": 0.7545,
+      "step": 8024
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.627741228070176e-06,
+      "loss": 1.236,
+      "step": 8025
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.627192982456141e-06,
+      "loss": 0.6602,
+      "step": 8026
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.626644736842106e-06,
+      "loss": 1.1616,
+      "step": 8027
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.62609649122807e-06,
+      "loss": 0.6912,
+      "step": 8028
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.6255482456140355e-06,
+      "loss": 0.528,
+      "step": 8029
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.625e-06,
+      "loss": 1.0139,
+      "step": 8030
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.624451754385965e-06,
+      "loss": 0.5618,
+      "step": 8031
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.623903508771931e-06,
+      "loss": 0.6378,
+      "step": 8032
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.623355263157895e-06,
+      "loss": 0.8153,
+      "step": 8033
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.62280701754386e-06,
+      "loss": 1.6829,
+      "step": 8034
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.622258771929825e-06,
+      "loss": 1.0026,
+      "step": 8035
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.62171052631579e-06,
+      "loss": 0.5896,
+      "step": 8036
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.621162280701755e-06,
+      "loss": 0.2454,
+      "step": 8037
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.620614035087721e-06,
+      "loss": 0.4817,
+      "step": 8038
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.620065789473685e-06,
+      "loss": 0.8845,
+      "step": 8039
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.6195175438596494e-06,
+      "loss": 1.054,
+      "step": 8040
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.618969298245614e-06,
+      "loss": 0.856,
+      "step": 8041
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.61842105263158e-06,
+      "loss": 0.7741,
+      "step": 8042
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.617872807017545e-06,
+      "loss": 0.5799,
+      "step": 8043
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.617324561403509e-06,
+      "loss": 0.9536,
+      "step": 8044
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.6167763157894736e-06,
+      "loss": 0.418,
+      "step": 8045
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.616228070175439e-06,
+      "loss": 0.5271,
+      "step": 8046
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.615679824561404e-06,
+      "loss": 0.78,
+      "step": 8047
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.615131578947369e-06,
+      "loss": 0.9198,
+      "step": 8048
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.614583333333333e-06,
+      "loss": 0.4455,
+      "step": 8049
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.6140350877192985e-06,
+      "loss": 0.7375,
+      "step": 8050
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.613486842105263e-06,
+      "loss": 0.3111,
+      "step": 8051
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.612938596491229e-06,
+      "loss": 0.3802,
+      "step": 8052
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.612390350877194e-06,
+      "loss": 0.3878,
+      "step": 8053
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.611842105263159e-06,
+      "loss": 0.45,
+      "step": 8054
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.611293859649123e-06,
+      "loss": 0.3569,
+      "step": 8055
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.610745614035088e-06,
+      "loss": 0.7051,
+      "step": 8056
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.610197368421053e-06,
+      "loss": 0.5397,
+      "step": 8057
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.609649122807018e-06,
+      "loss": 0.8803,
+      "step": 8058
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.609100877192984e-06,
+      "loss": 0.5106,
+      "step": 8059
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.608552631578948e-06,
+      "loss": 0.9527,
+      "step": 8060
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.6080043859649125e-06,
+      "loss": 0.9188,
+      "step": 8061
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.607456140350877e-06,
+      "loss": 0.7925,
+      "step": 8062
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.606907894736843e-06,
+      "loss": 0.8923,
+      "step": 8063
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.606359649122808e-06,
+      "loss": 1.2403,
+      "step": 8064
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.605811403508772e-06,
+      "loss": 0.562,
+      "step": 8065
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.605263157894737e-06,
+      "loss": 0.4702,
+      "step": 8066
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.604714912280702e-06,
+      "loss": 0.9075,
+      "step": 8067
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.604166666666667e-06,
+      "loss": 0.7805,
+      "step": 8068
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.603618421052632e-06,
+      "loss": 0.6916,
+      "step": 8069
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.6030701754385976e-06,
+      "loss": 0.4105,
+      "step": 8070
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.6025219298245616e-06,
+      "loss": 1.2744,
+      "step": 8071
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.601973684210526e-06,
+      "loss": 0.7001,
+      "step": 8072
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.601425438596492e-06,
+      "loss": 0.0815,
+      "step": 8073
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.600877192982457e-06,
+      "loss": 0.8351,
+      "step": 8074
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.600328947368422e-06,
+      "loss": 1.853,
+      "step": 8075
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.599780701754386e-06,
+      "loss": 0.6415,
+      "step": 8076
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.599232456140351e-06,
+      "loss": 0.6154,
+      "step": 8077
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.598684210526316e-06,
+      "loss": 0.2757,
+      "step": 8078
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.598135964912281e-06,
+      "loss": 0.5337,
+      "step": 8079
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.597587719298247e-06,
+      "loss": 0.0737,
+      "step": 8080
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.5970394736842115e-06,
+      "loss": 0.6883,
+      "step": 8081
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.5964912280701755e-06,
+      "loss": 1.0314,
+      "step": 8082
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.59594298245614e-06,
+      "loss": 0.6164,
+      "step": 8083
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.595394736842106e-06,
+      "loss": 0.7173,
+      "step": 8084
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.594846491228071e-06,
+      "loss": 0.85,
+      "step": 8085
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.594298245614036e-06,
+      "loss": 0.7532,
+      "step": 8086
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.59375e-06,
+      "loss": 0.3905,
+      "step": 8087
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.593201754385965e-06,
+      "loss": 0.634,
+      "step": 8088
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.59265350877193e-06,
+      "loss": 0.3592,
+      "step": 8089
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.592105263157896e-06,
+      "loss": 1.2242,
+      "step": 8090
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.591557017543861e-06,
+      "loss": 0.3017,
+      "step": 8091
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.5910087719298246e-06,
+      "loss": 0.337,
+      "step": 8092
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.590460526315789e-06,
+      "loss": 0.5932,
+      "step": 8093
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.589912280701755e-06,
+      "loss": 0.6954,
+      "step": 8094
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.58936403508772e-06,
+      "loss": 0.457,
+      "step": 8095
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.588815789473685e-06,
+      "loss": 0.2692,
+      "step": 8096
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.58826754385965e-06,
+      "loss": 0.6887,
+      "step": 8097
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.587719298245614e-06,
+      "loss": 1.3502,
+      "step": 8098
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.587171052631579e-06,
+      "loss": 0.8411,
+      "step": 8099
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.586622807017544e-06,
+      "loss": 0.1725,
+      "step": 8100
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.58607456140351e-06,
+      "loss": 0.862,
+      "step": 8101
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.5855263157894745e-06,
+      "loss": 0.3388,
+      "step": 8102
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.5849780701754385e-06,
+      "loss": 0.6876,
+      "step": 8103
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.584429824561403e-06,
+      "loss": 0.4623,
+      "step": 8104
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.583881578947369e-06,
+      "loss": 0.7923,
+      "step": 8105
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.583333333333334e-06,
+      "loss": 0.9183,
+      "step": 8106
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.5827850877192995e-06,
+      "loss": 0.8803,
+      "step": 8107
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.5822368421052635e-06,
+      "loss": 1.1869,
+      "step": 8108
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.581688596491228e-06,
+      "loss": 0.4884,
+      "step": 8109
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.581140350877193e-06,
+      "loss": 0.5575,
+      "step": 8110
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.580592105263159e-06,
+      "loss": 0.7855,
+      "step": 8111
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.580043859649124e-06,
+      "loss": 1.5216,
+      "step": 8112
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.5794956140350884e-06,
+      "loss": 0.7214,
+      "step": 8113
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.578947368421052e-06,
+      "loss": 0.2423,
+      "step": 8114
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.578399122807018e-06,
+      "loss": 1.0474,
+      "step": 8115
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.577850877192983e-06,
+      "loss": 0.7422,
+      "step": 8116
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.577302631578948e-06,
+      "loss": 0.1153,
+      "step": 8117
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.576754385964913e-06,
+      "loss": 0.6306,
+      "step": 8118
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.576206140350877e-06,
+      "loss": 0.8947,
+      "step": 8119
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.575657894736842e-06,
+      "loss": 1.1351,
+      "step": 8120
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.575109649122807e-06,
+      "loss": 0.1823,
+      "step": 8121
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.574561403508773e-06,
+      "loss": 0.1417,
+      "step": 8122
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.5740131578947375e-06,
+      "loss": 0.8523,
+      "step": 8123
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.573464912280702e-06,
+      "loss": 0.8621,
+      "step": 8124
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.572916666666667e-06,
+      "loss": 0.4503,
+      "step": 8125
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.572368421052632e-06,
+      "loss": 1.508,
+      "step": 8126
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.571820175438597e-06,
+      "loss": 0.4181,
+      "step": 8127
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.5712719298245625e-06,
+      "loss": 0.4347,
+      "step": 8128
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.570723684210527e-06,
+      "loss": 0.354,
+      "step": 8129
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.570175438596491e-06,
+      "loss": 0.977,
+      "step": 8130
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.569627192982456e-06,
+      "loss": 0.7724,
+      "step": 8131
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.569078947368422e-06,
+      "loss": 1.4999,
+      "step": 8132
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.568530701754387e-06,
+      "loss": 0.5092,
+      "step": 8133
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.5679824561403514e-06,
+      "loss": 0.3805,
+      "step": 8134
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.567434210526315e-06,
+      "loss": 0.6944,
+      "step": 8135
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.566885964912281e-06,
+      "loss": 0.4316,
+      "step": 8136
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.566337719298246e-06,
+      "loss": 1.7352,
+      "step": 8137
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.565789473684211e-06,
+      "loss": 0.7087,
+      "step": 8138
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.565241228070176e-06,
+      "loss": 0.3508,
+      "step": 8139
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.564692982456141e-06,
+      "loss": 0.4292,
+      "step": 8140
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.564144736842105e-06,
+      "loss": 1.196,
+      "step": 8141
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.56359649122807e-06,
+      "loss": 0.4315,
+      "step": 8142
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.563048245614036e-06,
+      "loss": 0.0883,
+      "step": 8143
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.5625000000000005e-06,
+      "loss": 0.6322,
+      "step": 8144
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.561951754385966e-06,
+      "loss": 0.9242,
+      "step": 8145
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.56140350877193e-06,
+      "loss": 0.8534,
+      "step": 8146
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.560855263157895e-06,
+      "loss": 0.4366,
+      "step": 8147
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.56030701754386e-06,
+      "loss": 0.1161,
+      "step": 8148
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.5597587719298255e-06,
+      "loss": 1.0586,
+      "step": 8149
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.55921052631579e-06,
+      "loss": 0.6639,
+      "step": 8150
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.558662280701754e-06,
+      "loss": 1.287,
+      "step": 8151
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.558114035087719e-06,
+      "loss": 0.4569,
+      "step": 8152
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.557565789473685e-06,
+      "loss": 0.6242,
+      "step": 8153
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.55701754385965e-06,
+      "loss": 0.2164,
+      "step": 8154
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.5564692982456144e-06,
+      "loss": 0.2325,
+      "step": 8155
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.55592105263158e-06,
+      "loss": 0.2286,
+      "step": 8156
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.555372807017544e-06,
+      "loss": 1.1912,
+      "step": 8157
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.554824561403509e-06,
+      "loss": 0.9843,
+      "step": 8158
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.554276315789474e-06,
+      "loss": 0.3617,
+      "step": 8159
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.553728070175439e-06,
+      "loss": 0.663,
+      "step": 8160
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.553179824561404e-06,
+      "loss": 0.2391,
+      "step": 8161
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.552631578947368e-06,
+      "loss": 0.6468,
+      "step": 8162
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.552083333333334e-06,
+      "loss": 0.7725,
+      "step": 8163
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.551535087719299e-06,
+      "loss": 0.4255,
+      "step": 8164
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.5509868421052635e-06,
+      "loss": 0.4982,
+      "step": 8165
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.550438596491229e-06,
+      "loss": 0.7427,
+      "step": 8166
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.549890350877194e-06,
+      "loss": 0.8315,
+      "step": 8167
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.549342105263158e-06,
+      "loss": 0.893,
+      "step": 8168
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.548793859649123e-06,
+      "loss": 0.7949,
+      "step": 8169
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.5482456140350885e-06,
+      "loss": 0.4852,
+      "step": 8170
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.547697368421053e-06,
+      "loss": 0.6345,
+      "step": 8171
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.547149122807018e-06,
+      "loss": 0.3875,
+      "step": 8172
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.546600877192982e-06,
+      "loss": 1.3135,
+      "step": 8173
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.546052631578948e-06,
+      "loss": 0.5413,
+      "step": 8174
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.545504385964913e-06,
+      "loss": 0.4441,
+      "step": 8175
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.5449561403508775e-06,
+      "loss": 0.8426,
+      "step": 8176
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.544407894736843e-06,
+      "loss": 0.5541,
+      "step": 8177
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.543859649122807e-06,
+      "loss": 0.5861,
+      "step": 8178
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.543311403508772e-06,
+      "loss": 0.9977,
+      "step": 8179
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.542763157894737e-06,
+      "loss": 2.2103,
+      "step": 8180
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.5422149122807024e-06,
+      "loss": 0.3377,
+      "step": 8181
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.541666666666667e-06,
+      "loss": 0.9604,
+      "step": 8182
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.541118421052633e-06,
+      "loss": 0.57,
+      "step": 8183
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.540570175438597e-06,
+      "loss": 1.6143,
+      "step": 8184
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.540021929824562e-06,
+      "loss": 0.5167,
+      "step": 8185
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.5394736842105266e-06,
+      "loss": 0.6345,
+      "step": 8186
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.538925438596492e-06,
+      "loss": 0.6977,
+      "step": 8187
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.538377192982457e-06,
+      "loss": 0.5093,
+      "step": 8188
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.537828947368421e-06,
+      "loss": 0.9606,
+      "step": 8189
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.537280701754386e-06,
+      "loss": 0.6442,
+      "step": 8190
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.5367324561403515e-06,
+      "loss": 0.3472,
+      "step": 8191
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.536184210526316e-06,
+      "loss": 0.9739,
+      "step": 8192
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.535635964912281e-06,
+      "loss": 0.204,
+      "step": 8193
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.535087719298247e-06,
+      "loss": 0.9073,
+      "step": 8194
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.534539473684211e-06,
+      "loss": 0.7546,
+      "step": 8195
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.533991228070176e-06,
+      "loss": 0.4678,
+      "step": 8196
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.5334429824561405e-06,
+      "loss": 0.706,
+      "step": 8197
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.532894736842106e-06,
+      "loss": 0.5845,
+      "step": 8198
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.532346491228071e-06,
+      "loss": 0.7273,
+      "step": 8199
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.531798245614035e-06,
+      "loss": 1.1035,
+      "step": 8200
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.531250000000001e-06,
+      "loss": 0.8601,
+      "step": 8201
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.5307017543859654e-06,
+      "loss": 1.0454,
+      "step": 8202
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.53015350877193e-06,
+      "loss": 0.4087,
+      "step": 8203
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.529605263157896e-06,
+      "loss": 0.8369,
+      "step": 8204
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.52905701754386e-06,
+      "loss": 0.0806,
+      "step": 8205
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.528508771929825e-06,
+      "loss": 1.3919,
+      "step": 8206
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.5279605263157896e-06,
+      "loss": 0.4873,
+      "step": 8207
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.527412280701755e-06,
+      "loss": 1.1025,
+      "step": 8208
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.52686403508772e-06,
+      "loss": 1.3406,
+      "step": 8209
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.526315789473685e-06,
+      "loss": 0.8361,
+      "step": 8210
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.525767543859649e-06,
+      "loss": 0.392,
+      "step": 8211
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.5252192982456145e-06,
+      "loss": 1.0308,
+      "step": 8212
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.524671052631579e-06,
+      "loss": 0.4909,
+      "step": 8213
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.524122807017544e-06,
+      "loss": 1.4901,
+      "step": 8214
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.52357456140351e-06,
+      "loss": 0.4882,
+      "step": 8215
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.523026315789474e-06,
+      "loss": 0.5349,
+      "step": 8216
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.522478070175439e-06,
+      "loss": 0.8417,
+      "step": 8217
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.521929824561404e-06,
+      "loss": 0.5449,
+      "step": 8218
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.521381578947369e-06,
+      "loss": 0.0915,
+      "step": 8219
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.520833333333334e-06,
+      "loss": 0.7955,
+      "step": 8220
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.520285087719298e-06,
+      "loss": 1.4413,
+      "step": 8221
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.519736842105264e-06,
+      "loss": 1.2119,
+      "step": 8222
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.5191885964912285e-06,
+      "loss": 0.7818,
+      "step": 8223
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.518640350877193e-06,
+      "loss": 0.7452,
+      "step": 8224
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.518092105263159e-06,
+      "loss": 0.9945,
+      "step": 8225
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.517543859649124e-06,
+      "loss": 0.5335,
+      "step": 8226
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.516995614035088e-06,
+      "loss": 0.5124,
+      "step": 8227
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.5164473684210526e-06,
+      "loss": 0.6589,
+      "step": 8228
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.515899122807018e-06,
+      "loss": 0.4674,
+      "step": 8229
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.515350877192983e-06,
+      "loss": 0.3586,
+      "step": 8230
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.514802631578948e-06,
+      "loss": 0.6625,
+      "step": 8231
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.514254385964912e-06,
+      "loss": 0.3942,
+      "step": 8232
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.5137061403508775e-06,
+      "loss": 0.6659,
+      "step": 8233
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.513157894736842e-06,
+      "loss": 0.3679,
+      "step": 8234
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.512609649122807e-06,
+      "loss": 0.8081,
+      "step": 8235
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.512061403508773e-06,
+      "loss": 0.4552,
+      "step": 8236
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.511513157894738e-06,
+      "loss": 0.6417,
+      "step": 8237
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.510964912280702e-06,
+      "loss": 1.3112,
+      "step": 8238
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.510416666666667e-06,
+      "loss": 0.5854,
+      "step": 8239
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.509868421052632e-06,
+      "loss": 0.4309,
+      "step": 8240
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.509320175438597e-06,
+      "loss": 0.6147,
+      "step": 8241
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.508771929824563e-06,
+      "loss": 0.5534,
+      "step": 8242
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.508223684210527e-06,
+      "loss": 0.3594,
+      "step": 8243
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.5076754385964915e-06,
+      "loss": 0.5447,
+      "step": 8244
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.507127192982456e-06,
+      "loss": 0.4455,
+      "step": 8245
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.506578947368422e-06,
+      "loss": 0.434,
+      "step": 8246
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.506030701754387e-06,
+      "loss": 0.7169,
+      "step": 8247
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.505482456140351e-06,
+      "loss": 0.3597,
+      "step": 8248
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.504934210526316e-06,
+      "loss": 0.4714,
+      "step": 8249
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.504385964912281e-06,
+      "loss": 0.4172,
+      "step": 8250
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.503837719298246e-06,
+      "loss": 0.5843,
+      "step": 8251
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.503289473684211e-06,
+      "loss": 0.2993,
+      "step": 8252
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.502741228070177e-06,
+      "loss": 1.143,
+      "step": 8253
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.5021929824561406e-06,
+      "loss": 0.479,
+      "step": 8254
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.501644736842105e-06,
+      "loss": 0.8778,
+      "step": 8255
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.501096491228071e-06,
+      "loss": 0.7617,
+      "step": 8256
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.500548245614036e-06,
+      "loss": 0.5086,
+      "step": 8257
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.500000000000001e-06,
+      "loss": 0.9791,
+      "step": 8258
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.499451754385965e-06,
+      "loss": 0.5427,
+      "step": 8259
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.49890350877193e-06,
+      "loss": 1.3269,
+      "step": 8260
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.498355263157895e-06,
+      "loss": 1.144,
+      "step": 8261
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.49780701754386e-06,
+      "loss": 0.6521,
+      "step": 8262
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.497258771929826e-06,
+      "loss": 0.3399,
+      "step": 8263
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.49671052631579e-06,
+      "loss": 0.5385,
+      "step": 8264
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.4961622807017545e-06,
+      "loss": 0.634,
+      "step": 8265
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.495614035087719e-06,
+      "loss": 0.3248,
+      "step": 8266
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.495065789473685e-06,
+      "loss": 0.4145,
+      "step": 8267
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.49451754385965e-06,
+      "loss": 0.9547,
+      "step": 8268
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.493969298245615e-06,
+      "loss": 1.249,
+      "step": 8269
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.493421052631579e-06,
+      "loss": 0.9866,
+      "step": 8270
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.492872807017544e-06,
+      "loss": 0.8095,
+      "step": 8271
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.492324561403509e-06,
+      "loss": 0.7221,
+      "step": 8272
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.491776315789474e-06,
+      "loss": 1.2626,
+      "step": 8273
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.49122807017544e-06,
+      "loss": 0.3434,
+      "step": 8274
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.4906798245614036e-06,
+      "loss": 0.4905,
+      "step": 8275
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.490131578947368e-06,
+      "loss": 1.3718,
+      "step": 8276
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.489583333333334e-06,
+      "loss": 0.4703,
+      "step": 8277
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.489035087719299e-06,
+      "loss": 0.6952,
+      "step": 8278
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.488486842105264e-06,
+      "loss": 0.6029,
+      "step": 8279
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.487938596491229e-06,
+      "loss": 0.8095,
+      "step": 8280
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.487390350877193e-06,
+      "loss": 0.6424,
+      "step": 8281
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.486842105263158e-06,
+      "loss": 0.7122,
+      "step": 8282
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.486293859649123e-06,
+      "loss": 0.6105,
+      "step": 8283
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.485745614035089e-06,
+      "loss": 0.7101,
+      "step": 8284
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.4851973684210535e-06,
+      "loss": 0.739,
+      "step": 8285
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.4846491228070175e-06,
+      "loss": 0.5595,
+      "step": 8286
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.484100877192982e-06,
+      "loss": 0.3632,
+      "step": 8287
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.483552631578948e-06,
+      "loss": 0.582,
+      "step": 8288
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.483004385964913e-06,
+      "loss": 0.5259,
+      "step": 8289
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.482456140350878e-06,
+      "loss": 0.6675,
+      "step": 8290
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.4819078947368425e-06,
+      "loss": 0.9832,
+      "step": 8291
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.481359649122807e-06,
+      "loss": 0.4991,
+      "step": 8292
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.480811403508772e-06,
+      "loss": 0.4058,
+      "step": 8293
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.480263157894738e-06,
+      "loss": 0.7196,
+      "step": 8294
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.479714912280703e-06,
+      "loss": 0.3407,
+      "step": 8295
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.4791666666666674e-06,
+      "loss": 0.3447,
+      "step": 8296
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.478618421052631e-06,
+      "loss": 0.8741,
+      "step": 8297
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.478070175438597e-06,
+      "loss": 0.6653,
+      "step": 8298
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.477521929824562e-06,
+      "loss": 0.5663,
+      "step": 8299
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.476973684210527e-06,
+      "loss": 0.5513,
+      "step": 8300
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.476425438596492e-06,
+      "loss": 1.2831,
+      "step": 8301
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.475877192982456e-06,
+      "loss": 0.8459,
+      "step": 8302
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.475328947368421e-06,
+      "loss": 0.308,
+      "step": 8303
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.474780701754386e-06,
+      "loss": 0.6741,
+      "step": 8304
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.474232456140352e-06,
+      "loss": 1.0321,
+      "step": 8305
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.4736842105263165e-06,
+      "loss": 1.0395,
+      "step": 8306
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.4731359649122805e-06,
+      "loss": 0.4967,
+      "step": 8307
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.472587719298245e-06,
+      "loss": 1.0912,
+      "step": 8308
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.472039473684211e-06,
+      "loss": 0.1219,
+      "step": 8309
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.471491228070176e-06,
+      "loss": 0.5861,
+      "step": 8310
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.4709429824561415e-06,
+      "loss": 1.3593,
+      "step": 8311
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.470394736842106e-06,
+      "loss": 0.5082,
+      "step": 8312
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.46984649122807e-06,
+      "loss": 0.7239,
+      "step": 8313
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.469298245614035e-06,
+      "loss": 0.9762,
+      "step": 8314
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.468750000000001e-06,
+      "loss": 0.8812,
+      "step": 8315
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.468201754385966e-06,
+      "loss": 1.4033,
+      "step": 8316
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.4676535087719304e-06,
+      "loss": 0.4588,
+      "step": 8317
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.467105263157894e-06,
+      "loss": 0.1573,
+      "step": 8318
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.46655701754386e-06,
+      "loss": 0.3341,
+      "step": 8319
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.466008771929825e-06,
+      "loss": 0.8576,
+      "step": 8320
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.46546052631579e-06,
+      "loss": 0.4713,
+      "step": 8321
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.464912280701755e-06,
+      "loss": 0.4871,
+      "step": 8322
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.46436403508772e-06,
+      "loss": 0.4368,
+      "step": 8323
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.463815789473684e-06,
+      "loss": 0.8406,
+      "step": 8324
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.463267543859649e-06,
+      "loss": 0.8131,
+      "step": 8325
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.462719298245615e-06,
+      "loss": 0.4347,
+      "step": 8326
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.4621710526315795e-06,
+      "loss": 0.8426,
+      "step": 8327
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.461622807017544e-06,
+      "loss": 0.3924,
+      "step": 8328
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.461074561403509e-06,
+      "loss": 1.4504,
+      "step": 8329
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.460526315789474e-06,
+      "loss": 0.3273,
+      "step": 8330
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.459978070175439e-06,
+      "loss": 0.3853,
+      "step": 8331
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.4594298245614045e-06,
+      "loss": 0.4334,
+      "step": 8332
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.458881578947369e-06,
+      "loss": 1.1676,
+      "step": 8333
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.458333333333333e-06,
+      "loss": 0.5457,
+      "step": 8334
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.457785087719298e-06,
+      "loss": 1.4386,
+      "step": 8335
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.457236842105264e-06,
+      "loss": 0.4857,
+      "step": 8336
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.456688596491229e-06,
+      "loss": 0.9239,
+      "step": 8337
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.4561403508771935e-06,
+      "loss": 0.8455,
+      "step": 8338
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.455592105263159e-06,
+      "loss": 0.4285,
+      "step": 8339
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.455043859649123e-06,
+      "loss": 0.656,
+      "step": 8340
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.454495614035088e-06,
+      "loss": 0.5284,
+      "step": 8341
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.453947368421053e-06,
+      "loss": 0.4752,
+      "step": 8342
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.453399122807018e-06,
+      "loss": 0.6446,
+      "step": 8343
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.452850877192983e-06,
+      "loss": 0.8231,
+      "step": 8344
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.452302631578947e-06,
+      "loss": 0.5545,
+      "step": 8345
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.451754385964912e-06,
+      "loss": 1.1976,
+      "step": 8346
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.451206140350878e-06,
+      "loss": 0.7716,
+      "step": 8347
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.4506578947368425e-06,
+      "loss": 0.5269,
+      "step": 8348
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.450109649122808e-06,
+      "loss": 0.4815,
+      "step": 8349
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.449561403508772e-06,
+      "loss": 0.4241,
+      "step": 8350
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.449013157894737e-06,
+      "loss": 1.0438,
+      "step": 8351
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.448464912280702e-06,
+      "loss": 1.2305,
+      "step": 8352
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.4479166666666675e-06,
+      "loss": 0.3072,
+      "step": 8353
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.447368421052632e-06,
+      "loss": 0.2878,
+      "step": 8354
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.446820175438597e-06,
+      "loss": 0.659,
+      "step": 8355
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.446271929824561e-06,
+      "loss": 0.8726,
+      "step": 8356
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.445723684210527e-06,
+      "loss": 1.4308,
+      "step": 8357
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.445175438596492e-06,
+      "loss": 0.4238,
+      "step": 8358
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.4446271929824565e-06,
+      "loss": 0.6466,
+      "step": 8359
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.444078947368422e-06,
+      "loss": 0.698,
+      "step": 8360
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.443530701754386e-06,
+      "loss": 0.8117,
+      "step": 8361
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.442982456140351e-06,
+      "loss": 0.9301,
+      "step": 8362
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.442434210526316e-06,
+      "loss": 1.0416,
+      "step": 8363
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.4418859649122814e-06,
+      "loss": 0.4978,
+      "step": 8364
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.441337719298246e-06,
+      "loss": 0.4641,
+      "step": 8365
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.440789473684212e-06,
+      "loss": 0.4316,
+      "step": 8366
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.440241228070176e-06,
+      "loss": 0.5415,
+      "step": 8367
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.439692982456141e-06,
+      "loss": 0.831,
+      "step": 8368
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.4391447368421056e-06,
+      "loss": 0.4707,
+      "step": 8369
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.438596491228071e-06,
+      "loss": 0.7915,
+      "step": 8370
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.438048245614036e-06,
+      "loss": 0.9281,
+      "step": 8371
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.4375e-06,
+      "loss": 1.3253,
+      "step": 8372
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.436951754385965e-06,
+      "loss": 1.1993,
+      "step": 8373
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.4364035087719305e-06,
+      "loss": 0.5795,
+      "step": 8374
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.435855263157895e-06,
+      "loss": 0.4425,
+      "step": 8375
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.43530701754386e-06,
+      "loss": 0.7815,
+      "step": 8376
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.434758771929824e-06,
+      "loss": 0.48,
+      "step": 8377
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.43421052631579e-06,
+      "loss": 0.7601,
+      "step": 8378
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.433662280701755e-06,
+      "loss": 0.378,
+      "step": 8379
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.4331140350877195e-06,
+      "loss": 0.4914,
+      "step": 8380
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.432565789473685e-06,
+      "loss": 1.6577,
+      "step": 8381
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.43201754385965e-06,
+      "loss": 0.7407,
+      "step": 8382
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.431469298245614e-06,
+      "loss": 0.3249,
+      "step": 8383
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.43092105263158e-06,
+      "loss": 0.4849,
+      "step": 8384
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.4303728070175444e-06,
+      "loss": 1.0351,
+      "step": 8385
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.429824561403509e-06,
+      "loss": 0.5628,
+      "step": 8386
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.429276315789475e-06,
+      "loss": 0.9122,
+      "step": 8387
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.428728070175439e-06,
+      "loss": 0.7297,
+      "step": 8388
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.428179824561404e-06,
+      "loss": 1.2415,
+      "step": 8389
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.4276315789473686e-06,
+      "loss": 0.6328,
+      "step": 8390
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.427083333333334e-06,
+      "loss": 0.2168,
+      "step": 8391
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.426535087719299e-06,
+      "loss": 0.3924,
+      "step": 8392
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.425986842105263e-06,
+      "loss": 0.6122,
+      "step": 8393
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.425438596491228e-06,
+      "loss": 0.38,
+      "step": 8394
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.4248903508771935e-06,
+      "loss": 0.4323,
+      "step": 8395
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.424342105263158e-06,
+      "loss": 0.6381,
+      "step": 8396
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.423793859649123e-06,
+      "loss": 1.2626,
+      "step": 8397
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.423245614035089e-06,
+      "loss": 0.7802,
+      "step": 8398
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.422697368421053e-06,
+      "loss": 0.6781,
+      "step": 8399
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.422149122807018e-06,
+      "loss": 0.3483,
+      "step": 8400
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.4216008771929825e-06,
+      "loss": 0.2753,
+      "step": 8401
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.421052631578948e-06,
+      "loss": 0.8838,
+      "step": 8402
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.420504385964913e-06,
+      "loss": 1.0227,
+      "step": 8403
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.419956140350877e-06,
+      "loss": 0.1519,
+      "step": 8404
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.419407894736843e-06,
+      "loss": 1.1714,
+      "step": 8405
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.4188596491228075e-06,
+      "loss": 0.1165,
+      "step": 8406
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.418311403508772e-06,
+      "loss": 0.508,
+      "step": 8407
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.417763157894738e-06,
+      "loss": 0.5901,
+      "step": 8408
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.417214912280703e-06,
+      "loss": 0.9471,
+      "step": 8409
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.416666666666667e-06,
+      "loss": 0.3105,
+      "step": 8410
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.416118421052632e-06,
+      "loss": 0.3332,
+      "step": 8411
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.415570175438597e-06,
+      "loss": 0.8651,
+      "step": 8412
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.415021929824562e-06,
+      "loss": 1.3506,
+      "step": 8413
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.414473684210527e-06,
+      "loss": 1.0011,
+      "step": 8414
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.413925438596491e-06,
+      "loss": 1.0941,
+      "step": 8415
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.4133771929824566e-06,
+      "loss": 0.325,
+      "step": 8416
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.412828947368421e-06,
+      "loss": 0.4303,
+      "step": 8417
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.412280701754386e-06,
+      "loss": 1.0232,
+      "step": 8418
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.411732456140352e-06,
+      "loss": 0.3396,
+      "step": 8419
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.411184210526316e-06,
+      "loss": 0.2976,
+      "step": 8420
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.410635964912281e-06,
+      "loss": 0.4458,
+      "step": 8421
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.410087719298246e-06,
+      "loss": 0.3375,
+      "step": 8422
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.409539473684211e-06,
+      "loss": 0.2522,
+      "step": 8423
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.408991228070176e-06,
+      "loss": 0.3612,
+      "step": 8424
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.408442982456142e-06,
+      "loss": 0.5716,
+      "step": 8425
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.407894736842106e-06,
+      "loss": 1.1382,
+      "step": 8426
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.4073464912280705e-06,
+      "loss": 1.0606,
+      "step": 8427
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.406798245614035e-06,
+      "loss": 0.5471,
+      "step": 8428
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.406250000000001e-06,
+      "loss": 1.4464,
+      "step": 8429
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.405701754385966e-06,
+      "loss": 0.667,
+      "step": 8430
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.40515350877193e-06,
+      "loss": 0.4927,
+      "step": 8431
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.404605263157895e-06,
+      "loss": 0.6181,
+      "step": 8432
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.40405701754386e-06,
+      "loss": 1.014,
+      "step": 8433
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.403508771929825e-06,
+      "loss": 1.1264,
+      "step": 8434
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.40296052631579e-06,
+      "loss": 0.974,
+      "step": 8435
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.402412280701754e-06,
+      "loss": 0.6297,
+      "step": 8436
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.4018640350877196e-06,
+      "loss": 1.2819,
+      "step": 8437
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.401315789473684e-06,
+      "loss": 0.8831,
+      "step": 8438
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.400767543859649e-06,
+      "loss": 0.9455,
+      "step": 8439
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.400219298245615e-06,
+      "loss": 0.4617,
+      "step": 8440
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.39967105263158e-06,
+      "loss": 0.3948,
+      "step": 8441
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.399122807017544e-06,
+      "loss": 1.2453,
+      "step": 8442
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.398574561403509e-06,
+      "loss": 0.38,
+      "step": 8443
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.398026315789474e-06,
+      "loss": 0.5933,
+      "step": 8444
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.397478070175439e-06,
+      "loss": 0.8903,
+      "step": 8445
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.396929824561405e-06,
+      "loss": 1.5132,
+      "step": 8446
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.396381578947369e-06,
+      "loss": 0.2882,
+      "step": 8447
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.3958333333333335e-06,
+      "loss": 0.5256,
+      "step": 8448
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.395285087719298e-06,
+      "loss": 0.2684,
+      "step": 8449
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.394736842105264e-06,
+      "loss": 0.5739,
+      "step": 8450
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.394188596491229e-06,
+      "loss": 1.501,
+      "step": 8451
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.393640350877194e-06,
+      "loss": 1.6854,
+      "step": 8452
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.393092105263158e-06,
+      "loss": 0.6813,
+      "step": 8453
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.392543859649123e-06,
+      "loss": 0.462,
+      "step": 8454
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.391995614035088e-06,
+      "loss": 0.6428,
+      "step": 8455
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.391447368421053e-06,
+      "loss": 1.1356,
+      "step": 8456
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.390899122807019e-06,
+      "loss": 0.3591,
+      "step": 8457
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.3903508771929826e-06,
+      "loss": 0.9896,
+      "step": 8458
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.389802631578947e-06,
+      "loss": 0.5714,
+      "step": 8459
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.389254385964913e-06,
+      "loss": 1.0126,
+      "step": 8460
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.388706140350878e-06,
+      "loss": 1.5045,
+      "step": 8461
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.388157894736843e-06,
+      "loss": 0.6194,
+      "step": 8462
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.387609649122807e-06,
+      "loss": 0.7201,
+      "step": 8463
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.387061403508772e-06,
+      "loss": 0.3746,
+      "step": 8464
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.386513157894737e-06,
+      "loss": 0.4659,
+      "step": 8465
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.385964912280702e-06,
+      "loss": 0.4924,
+      "step": 8466
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.385416666666668e-06,
+      "loss": 0.3,
+      "step": 8467
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.3848684210526325e-06,
+      "loss": 1.0334,
+      "step": 8468
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.3843201754385965e-06,
+      "loss": 0.7356,
+      "step": 8469
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.383771929824561e-06,
+      "loss": 1.1233,
+      "step": 8470
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.383223684210527e-06,
+      "loss": 0.9911,
+      "step": 8471
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.382675438596492e-06,
+      "loss": 0.7003,
+      "step": 8472
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.382127192982457e-06,
+      "loss": 0.8994,
+      "step": 8473
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.381578947368421e-06,
+      "loss": 0.795,
+      "step": 8474
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.381030701754386e-06,
+      "loss": 0.7952,
+      "step": 8475
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.380482456140351e-06,
+      "loss": 0.8346,
+      "step": 8476
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.379934210526317e-06,
+      "loss": 0.8759,
+      "step": 8477
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.379385964912282e-06,
+      "loss": 0.9324,
+      "step": 8478
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.3788377192982464e-06,
+      "loss": 1.1105,
+      "step": 8479
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.37828947368421e-06,
+      "loss": 1.1335,
+      "step": 8480
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.377741228070176e-06,
+      "loss": 1.1669,
+      "step": 8481
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.377192982456141e-06,
+      "loss": 0.3594,
+      "step": 8482
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.376644736842106e-06,
+      "loss": 0.906,
+      "step": 8483
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.376096491228071e-06,
+      "loss": 0.5979,
+      "step": 8484
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.375548245614035e-06,
+      "loss": 0.5383,
+      "step": 8485
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.375e-06,
+      "loss": 0.7032,
+      "step": 8486
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.374451754385965e-06,
+      "loss": 0.6565,
+      "step": 8487
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.373903508771931e-06,
+      "loss": 0.5008,
+      "step": 8488
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.3733552631578955e-06,
+      "loss": 1.3757,
+      "step": 8489
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.3728070175438595e-06,
+      "loss": 1.1647,
+      "step": 8490
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.372258771929824e-06,
+      "loss": 0.9087,
+      "step": 8491
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.37171052631579e-06,
+      "loss": 0.7817,
+      "step": 8492
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.371162280701755e-06,
+      "loss": 0.4808,
+      "step": 8493
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.37061403508772e-06,
+      "loss": 0.5076,
+      "step": 8494
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.370065789473685e-06,
+      "loss": 1.7074,
+      "step": 8495
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.369517543859649e-06,
+      "loss": 0.292,
+      "step": 8496
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.368969298245614e-06,
+      "loss": 1.2131,
+      "step": 8497
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.36842105263158e-06,
+      "loss": 0.9781,
+      "step": 8498
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.367872807017545e-06,
+      "loss": 0.7268,
+      "step": 8499
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.3673245614035094e-06,
+      "loss": 0.6017,
+      "step": 8500
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.3667763157894734e-06,
+      "loss": 0.4452,
+      "step": 8501
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.366228070175439e-06,
+      "loss": 0.3641,
+      "step": 8502
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.365679824561404e-06,
+      "loss": 0.4162,
+      "step": 8503
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.365131578947369e-06,
+      "loss": 0.2994,
+      "step": 8504
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.364583333333334e-06,
+      "loss": 0.7246,
+      "step": 8505
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.364035087719298e-06,
+      "loss": 1.0444,
+      "step": 8506
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.363486842105263e-06,
+      "loss": 0.7605,
+      "step": 8507
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.362938596491228e-06,
+      "loss": 0.526,
+      "step": 8508
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.362390350877194e-06,
+      "loss": 0.7447,
+      "step": 8509
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.3618421052631585e-06,
+      "loss": 0.7246,
+      "step": 8510
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.361293859649123e-06,
+      "loss": 1.523,
+      "step": 8511
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.360745614035087e-06,
+      "loss": 1.3916,
+      "step": 8512
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.360197368421053e-06,
+      "loss": 0.4387,
+      "step": 8513
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.359649122807018e-06,
+      "loss": 0.4468,
+      "step": 8514
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.3591008771929835e-06,
+      "loss": 0.7224,
+      "step": 8515
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.358552631578948e-06,
+      "loss": 0.654,
+      "step": 8516
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.358004385964912e-06,
+      "loss": 0.6987,
+      "step": 8517
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.357456140350877e-06,
+      "loss": 0.793,
+      "step": 8518
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.356907894736843e-06,
+      "loss": 0.6441,
+      "step": 8519
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.356359649122808e-06,
+      "loss": 0.9483,
+      "step": 8520
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.3558114035087725e-06,
+      "loss": 0.8104,
+      "step": 8521
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.355263157894738e-06,
+      "loss": 0.3978,
+      "step": 8522
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.354714912280702e-06,
+      "loss": 0.2559,
+      "step": 8523
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.354166666666667e-06,
+      "loss": 0.5645,
+      "step": 8524
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.353618421052632e-06,
+      "loss": 0.6099,
+      "step": 8525
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.3530701754385974e-06,
+      "loss": 0.3007,
+      "step": 8526
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.352521929824562e-06,
+      "loss": 0.6613,
+      "step": 8527
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.351973684210526e-06,
+      "loss": 0.5464,
+      "step": 8528
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.351425438596491e-06,
+      "loss": 0.9231,
+      "step": 8529
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.350877192982457e-06,
+      "loss": 0.56,
+      "step": 8530
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.3503289473684216e-06,
+      "loss": 0.228,
+      "step": 8531
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.349780701754386e-06,
+      "loss": 0.5181,
+      "step": 8532
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.349232456140351e-06,
+      "loss": 0.5997,
+      "step": 8533
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.348684210526316e-06,
+      "loss": 0.5508,
+      "step": 8534
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.348135964912281e-06,
+      "loss": 0.3253,
+      "step": 8535
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.3475877192982465e-06,
+      "loss": 0.6146,
+      "step": 8536
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.347039473684211e-06,
+      "loss": 0.1632,
+      "step": 8537
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.346491228070176e-06,
+      "loss": 1.9076,
+      "step": 8538
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.34594298245614e-06,
+      "loss": 0.4722,
+      "step": 8539
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.345394736842106e-06,
+      "loss": 0.5881,
+      "step": 8540
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.344846491228071e-06,
+      "loss": 0.3807,
+      "step": 8541
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.3442982456140355e-06,
+      "loss": 0.4005,
+      "step": 8542
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.343750000000001e-06,
+      "loss": 0.6089,
+      "step": 8543
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.343201754385965e-06,
+      "loss": 0.5173,
+      "step": 8544
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.34265350877193e-06,
+      "loss": 0.4202,
+      "step": 8545
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.342105263157895e-06,
+      "loss": 0.5795,
+      "step": 8546
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.3415570175438604e-06,
+      "loss": 0.3664,
+      "step": 8547
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.341008771929825e-06,
+      "loss": 0.4463,
+      "step": 8548
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.340460526315789e-06,
+      "loss": 0.7002,
+      "step": 8549
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.339912280701755e-06,
+      "loss": 0.2875,
+      "step": 8550
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.33936403508772e-06,
+      "loss": 0.7648,
+      "step": 8551
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.3388157894736846e-06,
+      "loss": 1.2317,
+      "step": 8552
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.33826754385965e-06,
+      "loss": 0.4352,
+      "step": 8553
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.337719298245615e-06,
+      "loss": 0.4977,
+      "step": 8554
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.337171052631579e-06,
+      "loss": 0.6949,
+      "step": 8555
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.336622807017544e-06,
+      "loss": 0.2953,
+      "step": 8556
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.3360745614035095e-06,
+      "loss": 0.3485,
+      "step": 8557
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.335526315789474e-06,
+      "loss": 0.4514,
+      "step": 8558
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.334978070175439e-06,
+      "loss": 0.391,
+      "step": 8559
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.334429824561403e-06,
+      "loss": 0.2838,
+      "step": 8560
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.333881578947369e-06,
+      "loss": 0.5536,
+      "step": 8561
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.333333333333334e-06,
+      "loss": 0.5905,
+      "step": 8562
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.3327850877192985e-06,
+      "loss": 1.014,
+      "step": 8563
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.332236842105264e-06,
+      "loss": 0.5382,
+      "step": 8564
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.331688596491229e-06,
+      "loss": 0.931,
+      "step": 8565
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.331140350877193e-06,
+      "loss": 0.2872,
+      "step": 8566
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.330592105263158e-06,
+      "loss": 0.8962,
+      "step": 8567
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.3300438596491235e-06,
+      "loss": 1.2744,
+      "step": 8568
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.329495614035088e-06,
+      "loss": 0.2349,
+      "step": 8569
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.328947368421054e-06,
+      "loss": 0.3194,
+      "step": 8570
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.328399122807018e-06,
+      "loss": 0.9242,
+      "step": 8571
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.327850877192983e-06,
+      "loss": 0.4356,
+      "step": 8572
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.3273026315789476e-06,
+      "loss": 0.5746,
+      "step": 8573
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.326754385964913e-06,
+      "loss": 1.3838,
+      "step": 8574
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.326206140350878e-06,
+      "loss": 1.0745,
+      "step": 8575
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.325657894736842e-06,
+      "loss": 0.3661,
+      "step": 8576
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.325109649122807e-06,
+      "loss": 0.2941,
+      "step": 8577
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.3245614035087725e-06,
+      "loss": 0.4375,
+      "step": 8578
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.324013157894737e-06,
+      "loss": 0.6687,
+      "step": 8579
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.323464912280702e-06,
+      "loss": 0.437,
+      "step": 8580
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.322916666666668e-06,
+      "loss": 0.33,
+      "step": 8581
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.322368421052632e-06,
+      "loss": 0.4265,
+      "step": 8582
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.321820175438597e-06,
+      "loss": 0.5855,
+      "step": 8583
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.3212719298245615e-06,
+      "loss": 0.3508,
+      "step": 8584
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.320723684210527e-06,
+      "loss": 0.515,
+      "step": 8585
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.320175438596492e-06,
+      "loss": 0.1468,
+      "step": 8586
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.319627192982456e-06,
+      "loss": 0.2117,
+      "step": 8587
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.319078947368422e-06,
+      "loss": 0.8914,
+      "step": 8588
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.3185307017543865e-06,
+      "loss": 0.5957,
+      "step": 8589
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.317982456140351e-06,
+      "loss": 0.6507,
+      "step": 8590
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.317434210526317e-06,
+      "loss": 1.2243,
+      "step": 8591
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.316885964912281e-06,
+      "loss": 1.0978,
+      "step": 8592
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.316337719298246e-06,
+      "loss": 0.4965,
+      "step": 8593
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.315789473684211e-06,
+      "loss": 0.6585,
+      "step": 8594
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.315241228070176e-06,
+      "loss": 0.4154,
+      "step": 8595
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.314692982456141e-06,
+      "loss": 0.6221,
+      "step": 8596
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.314144736842106e-06,
+      "loss": 0.7477,
+      "step": 8597
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.31359649122807e-06,
+      "loss": 0.7497,
+      "step": 8598
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.3130482456140356e-06,
+      "loss": 0.6488,
+      "step": 8599
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.3125e-06,
+      "loss": 0.4134,
+      "step": 8600
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.311951754385965e-06,
+      "loss": 0.3243,
+      "step": 8601
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.311403508771931e-06,
+      "loss": 0.2637,
+      "step": 8602
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.310855263157895e-06,
+      "loss": 0.4845,
+      "step": 8603
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.31030701754386e-06,
+      "loss": 1.2166,
+      "step": 8604
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.3097587719298245e-06,
+      "loss": 0.3625,
+      "step": 8605
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.30921052631579e-06,
+      "loss": 0.483,
+      "step": 8606
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.308662280701755e-06,
+      "loss": 1.2449,
+      "step": 8607
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.308114035087721e-06,
+      "loss": 0.313,
+      "step": 8608
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.307565789473685e-06,
+      "loss": 0.0938,
+      "step": 8609
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.3070175438596495e-06,
+      "loss": 1.1086,
+      "step": 8610
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.306469298245614e-06,
+      "loss": 0.9886,
+      "step": 8611
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.30592105263158e-06,
+      "loss": 0.8455,
+      "step": 8612
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.305372807017545e-06,
+      "loss": 0.6779,
+      "step": 8613
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.304824561403509e-06,
+      "loss": 0.0845,
+      "step": 8614
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.304276315789474e-06,
+      "loss": 0.1533,
+      "step": 8615
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.303728070175439e-06,
+      "loss": 0.6364,
+      "step": 8616
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.303179824561404e-06,
+      "loss": 1.5236,
+      "step": 8617
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.302631578947369e-06,
+      "loss": 1.0886,
+      "step": 8618
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.302083333333333e-06,
+      "loss": 1.0505,
+      "step": 8619
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.3015350877192986e-06,
+      "loss": 1.1925,
+      "step": 8620
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.300986842105263e-06,
+      "loss": 0.5823,
+      "step": 8621
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.300438596491228e-06,
+      "loss": 0.4942,
+      "step": 8622
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.299890350877194e-06,
+      "loss": 0.4782,
+      "step": 8623
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.299342105263159e-06,
+      "loss": 0.6556,
+      "step": 8624
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.298793859649123e-06,
+      "loss": 1.0025,
+      "step": 8625
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.298245614035088e-06,
+      "loss": 1.0247,
+      "step": 8626
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.297697368421053e-06,
+      "loss": 0.5646,
+      "step": 8627
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.297149122807018e-06,
+      "loss": 0.7076,
+      "step": 8628
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.296600877192984e-06,
+      "loss": 0.8985,
+      "step": 8629
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.296052631578948e-06,
+      "loss": 1.7303,
+      "step": 8630
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.2955043859649125e-06,
+      "loss": 0.6505,
+      "step": 8631
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.294956140350877e-06,
+      "loss": 1.012,
+      "step": 8632
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.294407894736843e-06,
+      "loss": 0.7644,
+      "step": 8633
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.293859649122808e-06,
+      "loss": 0.4876,
+      "step": 8634
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.293311403508772e-06,
+      "loss": 0.651,
+      "step": 8635
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.292763157894737e-06,
+      "loss": 0.7979,
+      "step": 8636
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.292214912280702e-06,
+      "loss": 0.4576,
+      "step": 8637
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.291666666666667e-06,
+      "loss": 0.7076,
+      "step": 8638
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.291118421052632e-06,
+      "loss": 0.8649,
+      "step": 8639
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.290570175438598e-06,
+      "loss": 0.2706,
+      "step": 8640
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.290021929824562e-06,
+      "loss": 0.675,
+      "step": 8641
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.289473684210526e-06,
+      "loss": 0.7285,
+      "step": 8642
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.288925438596492e-06,
+      "loss": 0.8372,
+      "step": 8643
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.288377192982457e-06,
+      "loss": 0.481,
+      "step": 8644
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.287828947368422e-06,
+      "loss": 0.5564,
+      "step": 8645
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.287280701754386e-06,
+      "loss": 0.5973,
+      "step": 8646
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.286732456140351e-06,
+      "loss": 0.6116,
+      "step": 8647
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.286184210526316e-06,
+      "loss": 0.499,
+      "step": 8648
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.285635964912281e-06,
+      "loss": 0.6415,
+      "step": 8649
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.285087719298247e-06,
+      "loss": 0.2781,
+      "step": 8650
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.2845394736842115e-06,
+      "loss": 0.3229,
+      "step": 8651
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.2839912280701755e-06,
+      "loss": 0.7814,
+      "step": 8652
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.28344298245614e-06,
+      "loss": 0.5481,
+      "step": 8653
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.282894736842106e-06,
+      "loss": 0.83,
+      "step": 8654
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.282346491228071e-06,
+      "loss": 0.3456,
+      "step": 8655
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.281798245614036e-06,
+      "loss": 1.1533,
+      "step": 8656
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.28125e-06,
+      "loss": 1.1294,
+      "step": 8657
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.280701754385965e-06,
+      "loss": 1.1586,
+      "step": 8658
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.28015350877193e-06,
+      "loss": 0.6892,
+      "step": 8659
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.279605263157895e-06,
+      "loss": 0.4897,
+      "step": 8660
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.279057017543861e-06,
+      "loss": 0.6545,
+      "step": 8661
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.278508771929825e-06,
+      "loss": 0.3223,
+      "step": 8662
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.277960526315789e-06,
+      "loss": 0.3713,
+      "step": 8663
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.277412280701755e-06,
+      "loss": 0.9921,
+      "step": 8664
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.27686403508772e-06,
+      "loss": 0.6285,
+      "step": 8665
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.276315789473685e-06,
+      "loss": 0.5833,
+      "step": 8666
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.27576754385965e-06,
+      "loss": 0.5483,
+      "step": 8667
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.275219298245614e-06,
+      "loss": 1.3298,
+      "step": 8668
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.274671052631579e-06,
+      "loss": 0.9461,
+      "step": 8669
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.274122807017544e-06,
+      "loss": 0.7727,
+      "step": 8670
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.27357456140351e-06,
+      "loss": 0.4342,
+      "step": 8671
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.2730263157894745e-06,
+      "loss": 0.6974,
+      "step": 8672
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.2724780701754385e-06,
+      "loss": 0.6576,
+      "step": 8673
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.271929824561403e-06,
+      "loss": 0.4039,
+      "step": 8674
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.271381578947369e-06,
+      "loss": 0.943,
+      "step": 8675
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.270833333333334e-06,
+      "loss": 1.452,
+      "step": 8676
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.270285087719299e-06,
+      "loss": 0.9573,
+      "step": 8677
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.269736842105263e-06,
+      "loss": 0.3136,
+      "step": 8678
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.269188596491228e-06,
+      "loss": 1.8635,
+      "step": 8679
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.268640350877193e-06,
+      "loss": 0.3388,
+      "step": 8680
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.268092105263159e-06,
+      "loss": 0.9245,
+      "step": 8681
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.267543859649124e-06,
+      "loss": 0.7499,
+      "step": 8682
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.2669956140350885e-06,
+      "loss": 0.6143,
+      "step": 8683
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.2664473684210524e-06,
+      "loss": 0.6452,
+      "step": 8684
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.265899122807018e-06,
+      "loss": 0.5769,
+      "step": 8685
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.265350877192983e-06,
+      "loss": 0.2971,
+      "step": 8686
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.264802631578948e-06,
+      "loss": 0.693,
+      "step": 8687
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.264254385964913e-06,
+      "loss": 0.5817,
+      "step": 8688
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.263706140350877e-06,
+      "loss": 0.6122,
+      "step": 8689
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.263157894736842e-06,
+      "loss": 0.9595,
+      "step": 8690
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.262609649122807e-06,
+      "loss": 0.3236,
+      "step": 8691
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.262061403508773e-06,
+      "loss": 0.2592,
+      "step": 8692
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.2615131578947375e-06,
+      "loss": 0.3544,
+      "step": 8693
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.260964912280702e-06,
+      "loss": 0.6612,
+      "step": 8694
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.260416666666666e-06,
+      "loss": 0.6975,
+      "step": 8695
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.259868421052632e-06,
+      "loss": 0.9329,
+      "step": 8696
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.259320175438597e-06,
+      "loss": 0.5351,
+      "step": 8697
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.258771929824562e-06,
+      "loss": 0.5641,
+      "step": 8698
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.258223684210527e-06,
+      "loss": 0.6163,
+      "step": 8699
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.257675438596491e-06,
+      "loss": 0.5999,
+      "step": 8700
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.257127192982456e-06,
+      "loss": 1.0915,
+      "step": 8701
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.256578947368422e-06,
+      "loss": 0.591,
+      "step": 8702
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.256030701754387e-06,
+      "loss": 0.2807,
+      "step": 8703
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.2554824561403515e-06,
+      "loss": 0.3946,
+      "step": 8704
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.2549342105263154e-06,
+      "loss": 0.8069,
+      "step": 8705
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.254385964912281e-06,
+      "loss": 0.6605,
+      "step": 8706
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.253837719298246e-06,
+      "loss": 0.6779,
+      "step": 8707
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.253289473684211e-06,
+      "loss": 1.1945,
+      "step": 8708
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.2527412280701764e-06,
+      "loss": 0.5376,
+      "step": 8709
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.252192982456141e-06,
+      "loss": 0.7972,
+      "step": 8710
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.251644736842105e-06,
+      "loss": 0.1391,
+      "step": 8711
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.25109649122807e-06,
+      "loss": 0.6989,
+      "step": 8712
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.250548245614036e-06,
+      "loss": 0.2245,
+      "step": 8713
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.2500000000000006e-06,
+      "loss": 1.0199,
+      "step": 8714
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.249451754385965e-06,
+      "loss": 0.4912,
+      "step": 8715
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.248903508771929e-06,
+      "loss": 0.6905,
+      "step": 8716
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.248355263157895e-06,
+      "loss": 0.4861,
+      "step": 8717
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.24780701754386e-06,
+      "loss": 0.3377,
+      "step": 8718
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.2472587719298255e-06,
+      "loss": 0.8702,
+      "step": 8719
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.24671052631579e-06,
+      "loss": 1.036,
+      "step": 8720
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.246162280701754e-06,
+      "loss": 0.2222,
+      "step": 8721
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.245614035087719e-06,
+      "loss": 0.2588,
+      "step": 8722
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.245065789473685e-06,
+      "loss": 1.6054,
+      "step": 8723
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.24451754385965e-06,
+      "loss": 0.6365,
+      "step": 8724
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.2439692982456145e-06,
+      "loss": 0.7285,
+      "step": 8725
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.24342105263158e-06,
+      "loss": 1.0309,
+      "step": 8726
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.242872807017544e-06,
+      "loss": 0.79,
+      "step": 8727
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.242324561403509e-06,
+      "loss": 0.9689,
+      "step": 8728
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.241776315789474e-06,
+      "loss": 0.5728,
+      "step": 8729
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.2412280701754394e-06,
+      "loss": 0.7074,
+      "step": 8730
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.240679824561404e-06,
+      "loss": 0.5945,
+      "step": 8731
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.240131578947368e-06,
+      "loss": 0.4005,
+      "step": 8732
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.239583333333333e-06,
+      "loss": 0.7003,
+      "step": 8733
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.239035087719299e-06,
+      "loss": 0.5543,
+      "step": 8734
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.2384868421052636e-06,
+      "loss": 1.4926,
+      "step": 8735
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.237938596491229e-06,
+      "loss": 0.4474,
+      "step": 8736
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.237390350877194e-06,
+      "loss": 0.9646,
+      "step": 8737
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.236842105263158e-06,
+      "loss": 0.5406,
+      "step": 8738
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.236293859649123e-06,
+      "loss": 0.7203,
+      "step": 8739
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.2357456140350885e-06,
+      "loss": 0.9658,
+      "step": 8740
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.235197368421053e-06,
+      "loss": 0.2297,
+      "step": 8741
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.234649122807018e-06,
+      "loss": 0.4486,
+      "step": 8742
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.234100877192982e-06,
+      "loss": 0.5017,
+      "step": 8743
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.233552631578948e-06,
+      "loss": 0.9902,
+      "step": 8744
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.233004385964913e-06,
+      "loss": 0.4907,
+      "step": 8745
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.2324561403508775e-06,
+      "loss": 1.058,
+      "step": 8746
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.231907894736843e-06,
+      "loss": 0.3211,
+      "step": 8747
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.231359649122807e-06,
+      "loss": 0.3503,
+      "step": 8748
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.230811403508772e-06,
+      "loss": 0.6115,
+      "step": 8749
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.230263157894737e-06,
+      "loss": 0.5172,
+      "step": 8750
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.2297149122807025e-06,
+      "loss": 0.4899,
+      "step": 8751
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.229166666666667e-06,
+      "loss": 0.5835,
+      "step": 8752
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.228618421052632e-06,
+      "loss": 0.4092,
+      "step": 8753
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.228070175438597e-06,
+      "loss": 0.5113,
+      "step": 8754
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.227521929824562e-06,
+      "loss": 0.4804,
+      "step": 8755
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.226973684210527e-06,
+      "loss": 0.6115,
+      "step": 8756
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.226425438596492e-06,
+      "loss": 1.4323,
+      "step": 8757
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.225877192982457e-06,
+      "loss": 0.6231,
+      "step": 8758
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.225328947368421e-06,
+      "loss": 0.6678,
+      "step": 8759
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.224780701754386e-06,
+      "loss": 0.8321,
+      "step": 8760
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.2242324561403516e-06,
+      "loss": 0.961,
+      "step": 8761
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.223684210526316e-06,
+      "loss": 1.0643,
+      "step": 8762
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.223135964912281e-06,
+      "loss": 0.5128,
+      "step": 8763
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.222587719298247e-06,
+      "loss": 0.4653,
+      "step": 8764
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.222039473684211e-06,
+      "loss": 0.4456,
+      "step": 8765
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.221491228070176e-06,
+      "loss": 1.0371,
+      "step": 8766
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.2209429824561405e-06,
+      "loss": 0.4196,
+      "step": 8767
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.220394736842106e-06,
+      "loss": 1.1285,
+      "step": 8768
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.219846491228071e-06,
+      "loss": 1.2724,
+      "step": 8769
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.219298245614035e-06,
+      "loss": 0.8829,
+      "step": 8770
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.21875e-06,
+      "loss": 0.51,
+      "step": 8771
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.2182017543859655e-06,
+      "loss": 1.9453,
+      "step": 8772
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.21765350877193e-06,
+      "loss": 0.5371,
+      "step": 8773
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.217105263157896e-06,
+      "loss": 0.8709,
+      "step": 8774
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.21655701754386e-06,
+      "loss": 0.8182,
+      "step": 8775
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.216008771929825e-06,
+      "loss": 1.0934,
+      "step": 8776
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.21546052631579e-06,
+      "loss": 0.194,
+      "step": 8777
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.214912280701755e-06,
+      "loss": 0.7047,
+      "step": 8778
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.21436403508772e-06,
+      "loss": 0.5307,
+      "step": 8779
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.213815789473685e-06,
+      "loss": 0.4376,
+      "step": 8780
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.213267543859649e-06,
+      "loss": 0.9845,
+      "step": 8781
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.2127192982456146e-06,
+      "loss": 0.917,
+      "step": 8782
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.212171052631579e-06,
+      "loss": 0.5325,
+      "step": 8783
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.211622807017544e-06,
+      "loss": 1.2355,
+      "step": 8784
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.21107456140351e-06,
+      "loss": 1.0063,
+      "step": 8785
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.210526315789474e-06,
+      "loss": 0.5603,
+      "step": 8786
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.209978070175439e-06,
+      "loss": 0.9611,
+      "step": 8787
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.2094298245614035e-06,
+      "loss": 0.8508,
+      "step": 8788
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.208881578947369e-06,
+      "loss": 0.7329,
+      "step": 8789
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.208333333333334e-06,
+      "loss": 0.6338,
+      "step": 8790
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.207785087719298e-06,
+      "loss": 1.277,
+      "step": 8791
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.207236842105264e-06,
+      "loss": 0.4301,
+      "step": 8792
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.2066885964912285e-06,
+      "loss": 0.6248,
+      "step": 8793
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.206140350877193e-06,
+      "loss": 0.6793,
+      "step": 8794
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.205592105263159e-06,
+      "loss": 0.8967,
+      "step": 8795
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.205043859649124e-06,
+      "loss": 1.0905,
+      "step": 8796
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.204495614035088e-06,
+      "loss": 0.4351,
+      "step": 8797
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.203947368421053e-06,
+      "loss": 1.098,
+      "step": 8798
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.203399122807018e-06,
+      "loss": 0.0999,
+      "step": 8799
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.202850877192983e-06,
+      "loss": 1.1682,
+      "step": 8800
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.202302631578948e-06,
+      "loss": 0.8703,
+      "step": 8801
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.201754385964912e-06,
+      "loss": 0.3502,
+      "step": 8802
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.2012061403508776e-06,
+      "loss": 1.1373,
+      "step": 8803
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.200657894736842e-06,
+      "loss": 0.674,
+      "step": 8804
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.200109649122807e-06,
+      "loss": 0.1279,
+      "step": 8805
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.199561403508773e-06,
+      "loss": 0.9088,
+      "step": 8806
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.199013157894738e-06,
+      "loss": 0.7294,
+      "step": 8807
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.198464912280702e-06,
+      "loss": 0.3982,
+      "step": 8808
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.1979166666666665e-06,
+      "loss": 1.507,
+      "step": 8809
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.197368421052632e-06,
+      "loss": 0.4151,
+      "step": 8810
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.196820175438597e-06,
+      "loss": 0.2434,
+      "step": 8811
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.196271929824563e-06,
+      "loss": 1.166,
+      "step": 8812
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.195723684210527e-06,
+      "loss": 0.7949,
+      "step": 8813
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.1951754385964915e-06,
+      "loss": 0.2897,
+      "step": 8814
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.194627192982456e-06,
+      "loss": 0.3796,
+      "step": 8815
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.194078947368422e-06,
+      "loss": 1.6331,
+      "step": 8816
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.193530701754387e-06,
+      "loss": 0.6462,
+      "step": 8817
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.192982456140351e-06,
+      "loss": 0.8801,
+      "step": 8818
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.192434210526316e-06,
+      "loss": 0.4602,
+      "step": 8819
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.191885964912281e-06,
+      "loss": 1.2801,
+      "step": 8820
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.191337719298246e-06,
+      "loss": 1.2446,
+      "step": 8821
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.190789473684211e-06,
+      "loss": 0.7182,
+      "step": 8822
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.190241228070177e-06,
+      "loss": 1.0264,
+      "step": 8823
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.189692982456141e-06,
+      "loss": 1.8161,
+      "step": 8824
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.189144736842105e-06,
+      "loss": 0.6728,
+      "step": 8825
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.18859649122807e-06,
+      "loss": 1.0617,
+      "step": 8826
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.188048245614036e-06,
+      "loss": 0.3663,
+      "step": 8827
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.187500000000001e-06,
+      "loss": 0.583,
+      "step": 8828
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.186951754385965e-06,
+      "loss": 0.8854,
+      "step": 8829
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.18640350877193e-06,
+      "loss": 0.6641,
+      "step": 8830
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.185855263157895e-06,
+      "loss": 0.6093,
+      "step": 8831
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.18530701754386e-06,
+      "loss": 0.8691,
+      "step": 8832
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.184758771929826e-06,
+      "loss": 0.4192,
+      "step": 8833
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.18421052631579e-06,
+      "loss": 1.158,
+      "step": 8834
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.1836622807017545e-06,
+      "loss": 0.6805,
+      "step": 8835
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.183114035087719e-06,
+      "loss": 0.8052,
+      "step": 8836
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.182565789473685e-06,
+      "loss": 0.4771,
+      "step": 8837
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.18201754385965e-06,
+      "loss": 0.4263,
+      "step": 8838
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.181469298245615e-06,
+      "loss": 0.3822,
+      "step": 8839
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.180921052631579e-06,
+      "loss": 1.2363,
+      "step": 8840
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.180372807017544e-06,
+      "loss": 0.1426,
+      "step": 8841
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.179824561403509e-06,
+      "loss": 0.7064,
+      "step": 8842
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.179276315789474e-06,
+      "loss": 1.9438,
+      "step": 8843
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.17872807017544e-06,
+      "loss": 0.9018,
+      "step": 8844
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.178179824561404e-06,
+      "loss": 0.7646,
+      "step": 8845
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.177631578947368e-06,
+      "loss": 1.1417,
+      "step": 8846
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.177083333333334e-06,
+      "loss": 1.338,
+      "step": 8847
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.176535087719299e-06,
+      "loss": 0.5254,
+      "step": 8848
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.175986842105264e-06,
+      "loss": 0.1556,
+      "step": 8849
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.175438596491229e-06,
+      "loss": 0.4975,
+      "step": 8850
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.174890350877193e-06,
+      "loss": 1.0643,
+      "step": 8851
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.174342105263158e-06,
+      "loss": 0.3276,
+      "step": 8852
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.173793859649123e-06,
+      "loss": 0.8556,
+      "step": 8853
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.173245614035089e-06,
+      "loss": 0.5373,
+      "step": 8854
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.1726973684210535e-06,
+      "loss": 0.3642,
+      "step": 8855
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.1721491228070175e-06,
+      "loss": 0.4118,
+      "step": 8856
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.171600877192982e-06,
+      "loss": 0.7151,
+      "step": 8857
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.171052631578948e-06,
+      "loss": 0.4848,
+      "step": 8858
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.170504385964913e-06,
+      "loss": 0.8088,
+      "step": 8859
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.169956140350878e-06,
+      "loss": 0.5987,
+      "step": 8860
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.169407894736842e-06,
+      "loss": 0.927,
+      "step": 8861
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.168859649122807e-06,
+      "loss": 0.8135,
+      "step": 8862
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.168311403508772e-06,
+      "loss": 1.6519,
+      "step": 8863
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.167763157894737e-06,
+      "loss": 0.4835,
+      "step": 8864
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.167214912280703e-06,
+      "loss": 0.5523,
+      "step": 8865
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.1666666666666675e-06,
+      "loss": 0.2853,
+      "step": 8866
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.1661184210526314e-06,
+      "loss": 0.5942,
+      "step": 8867
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.165570175438597e-06,
+      "loss": 0.9633,
+      "step": 8868
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.165021929824562e-06,
+      "loss": 0.6655,
+      "step": 8869
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.164473684210527e-06,
+      "loss": 0.392,
+      "step": 8870
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.1639254385964924e-06,
+      "loss": 0.6788,
+      "step": 8871
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.163377192982456e-06,
+      "loss": 0.7945,
+      "step": 8872
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.162828947368421e-06,
+      "loss": 0.4811,
+      "step": 8873
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.162280701754386e-06,
+      "loss": 0.6111,
+      "step": 8874
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.161732456140352e-06,
+      "loss": 0.682,
+      "step": 8875
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.1611842105263165e-06,
+      "loss": 1.2791,
+      "step": 8876
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.1606359649122805e-06,
+      "loss": 1.209,
+      "step": 8877
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.160087719298245e-06,
+      "loss": 0.4745,
+      "step": 8878
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.159539473684211e-06,
+      "loss": 0.8913,
+      "step": 8879
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.158991228070176e-06,
+      "loss": 1.4805,
+      "step": 8880
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.158442982456141e-06,
+      "loss": 0.4538,
+      "step": 8881
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.157894736842106e-06,
+      "loss": 1.2283,
+      "step": 8882
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.15734649122807e-06,
+      "loss": 0.6032,
+      "step": 8883
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.156798245614035e-06,
+      "loss": 0.4968,
+      "step": 8884
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.156250000000001e-06,
+      "loss": 0.5595,
+      "step": 8885
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.155701754385966e-06,
+      "loss": 0.509,
+      "step": 8886
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.1551535087719305e-06,
+      "loss": 0.6163,
+      "step": 8887
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.1546052631578944e-06,
+      "loss": 0.5366,
+      "step": 8888
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.15405701754386e-06,
+      "loss": 0.778,
+      "step": 8889
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.153508771929825e-06,
+      "loss": 0.6468,
+      "step": 8890
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.15296052631579e-06,
+      "loss": 1.1775,
+      "step": 8891
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.1524122807017554e-06,
+      "loss": 0.4336,
+      "step": 8892
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.15186403508772e-06,
+      "loss": 0.6076,
+      "step": 8893
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.151315789473684e-06,
+      "loss": 0.9512,
+      "step": 8894
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.150767543859649e-06,
+      "loss": 0.3351,
+      "step": 8895
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.150219298245615e-06,
+      "loss": 0.3186,
+      "step": 8896
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.1496710526315796e-06,
+      "loss": 1.1517,
+      "step": 8897
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.149122807017544e-06,
+      "loss": 0.8689,
+      "step": 8898
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.148574561403508e-06,
+      "loss": 0.5624,
+      "step": 8899
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.148026315789474e-06,
+      "loss": 1.315,
+      "step": 8900
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.147478070175439e-06,
+      "loss": 0.6249,
+      "step": 8901
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.1469298245614045e-06,
+      "loss": 1.5357,
+      "step": 8902
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.146381578947369e-06,
+      "loss": 0.5305,
+      "step": 8903
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.145833333333333e-06,
+      "loss": 0.6432,
+      "step": 8904
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.145285087719298e-06,
+      "loss": 0.3202,
+      "step": 8905
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.144736842105264e-06,
+      "loss": 0.2964,
+      "step": 8906
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.144188596491229e-06,
+      "loss": 1.0543,
+      "step": 8907
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.1436403508771935e-06,
+      "loss": 0.4556,
+      "step": 8908
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.143092105263159e-06,
+      "loss": 0.6081,
+      "step": 8909
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.142543859649123e-06,
+      "loss": 0.239,
+      "step": 8910
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.141995614035088e-06,
+      "loss": 0.9736,
+      "step": 8911
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.141447368421053e-06,
+      "loss": 0.7049,
+      "step": 8912
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.1408991228070185e-06,
+      "loss": 0.2531,
+      "step": 8913
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.140350877192983e-06,
+      "loss": 1.3302,
+      "step": 8914
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.139802631578947e-06,
+      "loss": 0.9019,
+      "step": 8915
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.139254385964912e-06,
+      "loss": 0.6807,
+      "step": 8916
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.138706140350878e-06,
+      "loss": 0.4875,
+      "step": 8917
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.1381578947368426e-06,
+      "loss": 0.4732,
+      "step": 8918
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.137609649122807e-06,
+      "loss": 0.4492,
+      "step": 8919
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.137061403508772e-06,
+      "loss": 1.0554,
+      "step": 8920
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.136513157894737e-06,
+      "loss": 1.1166,
+      "step": 8921
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.135964912280702e-06,
+      "loss": 0.6352,
+      "step": 8922
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.1354166666666675e-06,
+      "loss": 1.0062,
+      "step": 8923
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.134868421052632e-06,
+      "loss": 0.7276,
+      "step": 8924
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.134320175438597e-06,
+      "loss": 1.115,
+      "step": 8925
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.133771929824561e-06,
+      "loss": 0.1801,
+      "step": 8926
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.133223684210527e-06,
+      "loss": 0.7389,
+      "step": 8927
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.132675438596492e-06,
+      "loss": 1.0377,
+      "step": 8928
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.1321271929824565e-06,
+      "loss": 1.2593,
+      "step": 8929
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.131578947368422e-06,
+      "loss": 0.6059,
+      "step": 8930
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.131030701754386e-06,
+      "loss": 0.6039,
+      "step": 8931
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.130482456140351e-06,
+      "loss": 0.4412,
+      "step": 8932
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.129934210526316e-06,
+      "loss": 0.4923,
+      "step": 8933
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.1293859649122815e-06,
+      "loss": 0.3582,
+      "step": 8934
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.128837719298246e-06,
+      "loss": 0.497,
+      "step": 8935
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.128289473684211e-06,
+      "loss": 0.7118,
+      "step": 8936
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.127741228070175e-06,
+      "loss": 0.4716,
+      "step": 8937
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.127192982456141e-06,
+      "loss": 1.0106,
+      "step": 8938
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.126644736842106e-06,
+      "loss": 0.9185,
+      "step": 8939
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.126096491228071e-06,
+      "loss": 1.5185,
+      "step": 8940
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.125548245614036e-06,
+      "loss": 0.1834,
+      "step": 8941
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.125e-06,
+      "loss": 0.6864,
+      "step": 8942
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.124451754385965e-06,
+      "loss": 0.5061,
+      "step": 8943
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.1239035087719306e-06,
+      "loss": 1.1591,
+      "step": 8944
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.123355263157895e-06,
+      "loss": 0.7016,
+      "step": 8945
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.12280701754386e-06,
+      "loss": 0.7382,
+      "step": 8946
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.122258771929824e-06,
+      "loss": 0.7285,
+      "step": 8947
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.12171052631579e-06,
+      "loss": 0.5896,
+      "step": 8948
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.121162280701755e-06,
+      "loss": 0.4888,
+      "step": 8949
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.1206140350877195e-06,
+      "loss": 0.7859,
+      "step": 8950
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.120065789473685e-06,
+      "loss": 0.9795,
+      "step": 8951
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.11951754385965e-06,
+      "loss": 0.4428,
+      "step": 8952
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.118969298245614e-06,
+      "loss": 0.4258,
+      "step": 8953
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.118421052631579e-06,
+      "loss": 0.8626,
+      "step": 8954
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.1178728070175445e-06,
+      "loss": 0.7417,
+      "step": 8955
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.117324561403509e-06,
+      "loss": 0.7302,
+      "step": 8956
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.116776315789474e-06,
+      "loss": 0.7147,
+      "step": 8957
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.116228070175439e-06,
+      "loss": 0.5634,
+      "step": 8958
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.115679824561404e-06,
+      "loss": 0.6724,
+      "step": 8959
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.115131578947369e-06,
+      "loss": 0.2118,
+      "step": 8960
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.114583333333334e-06,
+      "loss": 0.3776,
+      "step": 8961
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.114035087719299e-06,
+      "loss": 0.288,
+      "step": 8962
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.113486842105263e-06,
+      "loss": 0.4862,
+      "step": 8963
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.112938596491228e-06,
+      "loss": 0.3511,
+      "step": 8964
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.1123903508771936e-06,
+      "loss": 0.47,
+      "step": 8965
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.111842105263158e-06,
+      "loss": 0.7819,
+      "step": 8966
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.111293859649123e-06,
+      "loss": 0.5125,
+      "step": 8967
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.110745614035089e-06,
+      "loss": 0.6671,
+      "step": 8968
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.110197368421053e-06,
+      "loss": 0.9009,
+      "step": 8969
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.109649122807018e-06,
+      "loss": 0.9366,
+      "step": 8970
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.1091008771929825e-06,
+      "loss": 0.338,
+      "step": 8971
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.108552631578948e-06,
+      "loss": 1.9346,
+      "step": 8972
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.108004385964913e-06,
+      "loss": 0.6345,
+      "step": 8973
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.107456140350877e-06,
+      "loss": 0.5633,
+      "step": 8974
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.106907894736842e-06,
+      "loss": 0.5968,
+      "step": 8975
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.1063596491228075e-06,
+      "loss": 0.2397,
+      "step": 8976
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.105811403508772e-06,
+      "loss": 0.6361,
+      "step": 8977
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.105263157894738e-06,
+      "loss": 1.5335,
+      "step": 8978
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.104714912280703e-06,
+      "loss": 1.2679,
+      "step": 8979
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.104166666666667e-06,
+      "loss": 0.5646,
+      "step": 8980
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.103618421052632e-06,
+      "loss": 0.6437,
+      "step": 8981
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.103070175438597e-06,
+      "loss": 0.5253,
+      "step": 8982
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.102521929824562e-06,
+      "loss": 0.4417,
+      "step": 8983
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.101973684210527e-06,
+      "loss": 0.3513,
+      "step": 8984
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.101425438596491e-06,
+      "loss": 0.4646,
+      "step": 8985
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.100877192982457e-06,
+      "loss": 0.1595,
+      "step": 8986
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.100328947368421e-06,
+      "loss": 0.3781,
+      "step": 8987
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.099780701754386e-06,
+      "loss": 0.4712,
+      "step": 8988
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.099232456140352e-06,
+      "loss": 0.4413,
+      "step": 8989
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.098684210526316e-06,
+      "loss": 1.1983,
+      "step": 8990
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.098135964912281e-06,
+      "loss": 1.0136,
+      "step": 8991
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.0975877192982455e-06,
+      "loss": 1.731,
+      "step": 8992
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.097039473684211e-06,
+      "loss": 0.9357,
+      "step": 8993
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.096491228070176e-06,
+      "loss": 0.4945,
+      "step": 8994
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.095942982456142e-06,
+      "loss": 0.5865,
+      "step": 8995
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.095394736842106e-06,
+      "loss": 0.2499,
+      "step": 8996
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.0948464912280705e-06,
+      "loss": 1.2133,
+      "step": 8997
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.094298245614035e-06,
+      "loss": 1.0267,
+      "step": 8998
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.093750000000001e-06,
+      "loss": 1.4476,
+      "step": 8999
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.093201754385966e-06,
+      "loss": 0.3506,
+      "step": 9000
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.09265350877193e-06,
+      "loss": 1.4541,
+      "step": 9001
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.092105263157895e-06,
+      "loss": 0.6651,
+      "step": 9002
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.09155701754386e-06,
+      "loss": 1.5148,
+      "step": 9003
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.091008771929825e-06,
+      "loss": 0.6887,
+      "step": 9004
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.09046052631579e-06,
+      "loss": 1.2539,
+      "step": 9005
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.089912280701754e-06,
+      "loss": 1.2946,
+      "step": 9006
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.08936403508772e-06,
+      "loss": 0.6548,
+      "step": 9007
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.088815789473684e-06,
+      "loss": 0.8067,
+      "step": 9008
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.088267543859649e-06,
+      "loss": 0.6169,
+      "step": 9009
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.087719298245615e-06,
+      "loss": 0.5321,
+      "step": 9010
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.08717105263158e-06,
+      "loss": 0.7082,
+      "step": 9011
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.086622807017544e-06,
+      "loss": 0.7895,
+      "step": 9012
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.086074561403509e-06,
+      "loss": 0.9231,
+      "step": 9013
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.085526315789474e-06,
+      "loss": 0.2483,
+      "step": 9014
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.084978070175439e-06,
+      "loss": 0.7391,
+      "step": 9015
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.084429824561405e-06,
+      "loss": 0.8558,
+      "step": 9016
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.083881578947369e-06,
+      "loss": 0.27,
+      "step": 9017
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.0833333333333335e-06,
+      "loss": 0.7498,
+      "step": 9018
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.082785087719298e-06,
+      "loss": 0.4556,
+      "step": 9019
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.082236842105264e-06,
+      "loss": 0.3518,
+      "step": 9020
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.081688596491229e-06,
+      "loss": 0.8334,
+      "step": 9021
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.081140350877194e-06,
+      "loss": 0.165,
+      "step": 9022
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.080592105263158e-06,
+      "loss": 0.4181,
+      "step": 9023
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.080043859649123e-06,
+      "loss": 0.7386,
+      "step": 9024
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.079495614035088e-06,
+      "loss": 1.0263,
+      "step": 9025
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.078947368421053e-06,
+      "loss": 0.315,
+      "step": 9026
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.078399122807019e-06,
+      "loss": 0.6802,
+      "step": 9027
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.077850877192983e-06,
+      "loss": 0.8691,
+      "step": 9028
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.0773026315789474e-06,
+      "loss": 0.3314,
+      "step": 9029
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.076754385964912e-06,
+      "loss": 0.6791,
+      "step": 9030
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.076206140350878e-06,
+      "loss": 0.5901,
+      "step": 9031
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.075657894736843e-06,
+      "loss": 1.1297,
+      "step": 9032
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.075109649122807e-06,
+      "loss": 0.4334,
+      "step": 9033
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.074561403508772e-06,
+      "loss": 0.8664,
+      "step": 9034
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.074013157894737e-06,
+      "loss": 0.2585,
+      "step": 9035
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.073464912280702e-06,
+      "loss": 1.8262,
+      "step": 9036
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.072916666666668e-06,
+      "loss": 0.4709,
+      "step": 9037
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.0723684210526325e-06,
+      "loss": 1.1386,
+      "step": 9038
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.0718201754385965e-06,
+      "loss": 0.9105,
+      "step": 9039
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.071271929824561e-06,
+      "loss": 0.6207,
+      "step": 9040
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.070723684210527e-06,
+      "loss": 0.3897,
+      "step": 9041
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.070175438596492e-06,
+      "loss": 0.3794,
+      "step": 9042
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.069627192982457e-06,
+      "loss": 2.044,
+      "step": 9043
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.069078947368421e-06,
+      "loss": 1.2247,
+      "step": 9044
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.068530701754386e-06,
+      "loss": 1.253,
+      "step": 9045
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.067982456140351e-06,
+      "loss": 0.1201,
+      "step": 9046
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.067434210526316e-06,
+      "loss": 1.619,
+      "step": 9047
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.066885964912282e-06,
+      "loss": 0.5152,
+      "step": 9048
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.0663377192982465e-06,
+      "loss": 0.4933,
+      "step": 9049
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.0657894736842104e-06,
+      "loss": 0.6274,
+      "step": 9050
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.065241228070176e-06,
+      "loss": 0.5813,
+      "step": 9051
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.064692982456141e-06,
+      "loss": 0.9275,
+      "step": 9052
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.064144736842106e-06,
+      "loss": 1.5043,
+      "step": 9053
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.0635964912280714e-06,
+      "loss": 0.4217,
+      "step": 9054
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.063048245614035e-06,
+      "loss": 0.2207,
+      "step": 9055
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.0625e-06,
+      "loss": 0.7226,
+      "step": 9056
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.061951754385965e-06,
+      "loss": 0.5432,
+      "step": 9057
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.061403508771931e-06,
+      "loss": 1.0165,
+      "step": 9058
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.0608552631578956e-06,
+      "loss": 0.7001,
+      "step": 9059
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.0603070175438595e-06,
+      "loss": 0.4381,
+      "step": 9060
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.059758771929824e-06,
+      "loss": 0.4332,
+      "step": 9061
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.05921052631579e-06,
+      "loss": 0.4775,
+      "step": 9062
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.058662280701755e-06,
+      "loss": 0.3296,
+      "step": 9063
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.05811403508772e-06,
+      "loss": 0.9814,
+      "step": 9064
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.057565789473685e-06,
+      "loss": 0.4519,
+      "step": 9065
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.057017543859649e-06,
+      "loss": 0.5417,
+      "step": 9066
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.056469298245614e-06,
+      "loss": 1.1301,
+      "step": 9067
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.055921052631579e-06,
+      "loss": 1.0964,
+      "step": 9068
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.055372807017545e-06,
+      "loss": 0.4278,
+      "step": 9069
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.0548245614035095e-06,
+      "loss": 0.7662,
+      "step": 9070
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.0542763157894735e-06,
+      "loss": 0.6673,
+      "step": 9071
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.053728070175439e-06,
+      "loss": 0.458,
+      "step": 9072
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.053179824561404e-06,
+      "loss": 0.6874,
+      "step": 9073
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.052631578947369e-06,
+      "loss": 0.8,
+      "step": 9074
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.0520833333333344e-06,
+      "loss": 0.2443,
+      "step": 9075
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.051535087719298e-06,
+      "loss": 0.6807,
+      "step": 9076
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.050986842105263e-06,
+      "loss": 0.7907,
+      "step": 9077
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.050438596491228e-06,
+      "loss": 0.8545,
+      "step": 9078
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.049890350877194e-06,
+      "loss": 0.8451,
+      "step": 9079
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.0493421052631586e-06,
+      "loss": 0.7097,
+      "step": 9080
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.048793859649123e-06,
+      "loss": 0.5665,
+      "step": 9081
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.048245614035087e-06,
+      "loss": 0.8087,
+      "step": 9082
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.047697368421053e-06,
+      "loss": 0.3296,
+      "step": 9083
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.047149122807018e-06,
+      "loss": 0.8371,
+      "step": 9084
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.046600877192983e-06,
+      "loss": 1.2106,
+      "step": 9085
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.046052631578948e-06,
+      "loss": 0.2768,
+      "step": 9086
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.045504385964912e-06,
+      "loss": 1.2583,
+      "step": 9087
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.044956140350877e-06,
+      "loss": 0.5637,
+      "step": 9088
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.044407894736843e-06,
+      "loss": 0.5669,
+      "step": 9089
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.043859649122808e-06,
+      "loss": 0.4387,
+      "step": 9090
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.0433114035087725e-06,
+      "loss": 1.774,
+      "step": 9091
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.042763157894738e-06,
+      "loss": 1.0029,
+      "step": 9092
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.042214912280702e-06,
+      "loss": 0.3607,
+      "step": 9093
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.041666666666667e-06,
+      "loss": 0.3157,
+      "step": 9094
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.041118421052632e-06,
+      "loss": 0.696,
+      "step": 9095
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.0405701754385975e-06,
+      "loss": 0.27,
+      "step": 9096
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.040021929824562e-06,
+      "loss": 0.6642,
+      "step": 9097
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.039473684210526e-06,
+      "loss": 1.284,
+      "step": 9098
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.038925438596491e-06,
+      "loss": 0.4943,
+      "step": 9099
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.038377192982457e-06,
+      "loss": 0.7783,
+      "step": 9100
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.037828947368422e-06,
+      "loss": 0.8893,
+      "step": 9101
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.037280701754386e-06,
+      "loss": 0.3637,
+      "step": 9102
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.03673245614035e-06,
+      "loss": 0.9599,
+      "step": 9103
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.036184210526316e-06,
+      "loss": 0.6215,
+      "step": 9104
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.035635964912281e-06,
+      "loss": 0.6672,
+      "step": 9105
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.0350877192982465e-06,
+      "loss": 0.3252,
+      "step": 9106
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.034539473684211e-06,
+      "loss": 1.0463,
+      "step": 9107
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.033991228070176e-06,
+      "loss": 0.5518,
+      "step": 9108
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.03344298245614e-06,
+      "loss": 0.5532,
+      "step": 9109
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.032894736842106e-06,
+      "loss": 0.4628,
+      "step": 9110
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.032346491228071e-06,
+      "loss": 0.8603,
+      "step": 9111
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.0317982456140355e-06,
+      "loss": 1.5204,
+      "step": 9112
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.031250000000001e-06,
+      "loss": 0.5475,
+      "step": 9113
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.030701754385965e-06,
+      "loss": 0.8728,
+      "step": 9114
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.03015350877193e-06,
+      "loss": 0.3309,
+      "step": 9115
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.029605263157895e-06,
+      "loss": 0.5692,
+      "step": 9116
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.0290570175438605e-06,
+      "loss": 0.5535,
+      "step": 9117
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.028508771929825e-06,
+      "loss": 0.2408,
+      "step": 9118
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.027960526315789e-06,
+      "loss": 0.7442,
+      "step": 9119
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.027412280701754e-06,
+      "loss": 0.2974,
+      "step": 9120
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.02686403508772e-06,
+      "loss": 0.6696,
+      "step": 9121
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.026315789473685e-06,
+      "loss": 0.6437,
+      "step": 9122
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.025767543859649e-06,
+      "loss": 0.4203,
+      "step": 9123
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.025219298245615e-06,
+      "loss": 0.8312,
+      "step": 9124
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.024671052631579e-06,
+      "loss": 0.4308,
+      "step": 9125
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.024122807017544e-06,
+      "loss": 1.4266,
+      "step": 9126
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.0235745614035096e-06,
+      "loss": 0.9094,
+      "step": 9127
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.023026315789474e-06,
+      "loss": 0.6906,
+      "step": 9128
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.022478070175439e-06,
+      "loss": 0.8897,
+      "step": 9129
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.021929824561403e-06,
+      "loss": 0.2267,
+      "step": 9130
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.021381578947369e-06,
+      "loss": 0.9009,
+      "step": 9131
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.020833333333334e-06,
+      "loss": 0.4917,
+      "step": 9132
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.0202850877192985e-06,
+      "loss": 0.588,
+      "step": 9133
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.019736842105264e-06,
+      "loss": 0.6497,
+      "step": 9134
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.019188596491229e-06,
+      "loss": 0.4351,
+      "step": 9135
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.018640350877193e-06,
+      "loss": 0.4004,
+      "step": 9136
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.018092105263158e-06,
+      "loss": 0.3511,
+      "step": 9137
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.0175438596491235e-06,
+      "loss": 0.4453,
+      "step": 9138
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.016995614035088e-06,
+      "loss": 0.949,
+      "step": 9139
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.016447368421053e-06,
+      "loss": 0.5693,
+      "step": 9140
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.015899122807017e-06,
+      "loss": 0.3201,
+      "step": 9141
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.015350877192983e-06,
+      "loss": 0.7798,
+      "step": 9142
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.014802631578948e-06,
+      "loss": 0.6328,
+      "step": 9143
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.014254385964913e-06,
+      "loss": 0.1649,
+      "step": 9144
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.013706140350878e-06,
+      "loss": 0.8897,
+      "step": 9145
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.7338527783473446,
+      "eval_loss": 0.7286343574523926,
+      "eval_roc_auc": 0.9290818319435464,
+      "eval_runtime": 191.3042,
+      "eval_samples_per_second": 25.493,
+      "eval_steps_per_second": 3.189,
+      "step": 9145
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.013157894736842e-06,
+      "loss": 0.4182,
+      "step": 9146
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.012609649122807e-06,
+      "loss": 0.6683,
+      "step": 9147
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.0120614035087726e-06,
+      "loss": 0.328,
+      "step": 9148
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.011513157894737e-06,
+      "loss": 0.2632,
+      "step": 9149
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.010964912280702e-06,
+      "loss": 0.2713,
+      "step": 9150
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.010416666666668e-06,
+      "loss": 0.4876,
+      "step": 9151
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.009868421052632e-06,
+      "loss": 1.297,
+      "step": 9152
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.009320175438597e-06,
+      "loss": 1.1923,
+      "step": 9153
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.0087719298245615e-06,
+      "loss": 0.719,
+      "step": 9154
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 5.008223684210527e-06,
+      "loss": 0.3029,
+      "step": 9155
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 5.007675438596492e-06,
+      "loss": 0.2202,
+      "step": 9156
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 5.007127192982456e-06,
+      "loss": 1.4455,
+      "step": 9157
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 5.006578947368421e-06,
+      "loss": 0.5858,
+      "step": 9158
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 5.0060307017543865e-06,
+      "loss": 0.6618,
+      "step": 9159
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 5.005482456140351e-06,
+      "loss": 0.6679,
+      "step": 9160
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 5.004934210526316e-06,
+      "loss": 0.9237,
+      "step": 9161
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 5.004385964912281e-06,
+      "loss": 0.4888,
+      "step": 9162
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 5.003837719298246e-06,
+      "loss": 0.7096,
+      "step": 9163
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 5.003289473684211e-06,
+      "loss": 0.5341,
+      "step": 9164
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 5.002741228070176e-06,
+      "loss": 1.499,
+      "step": 9165
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 5.002192982456141e-06,
+      "loss": 0.5074,
+      "step": 9166
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 5.001644736842106e-06,
+      "loss": 0.5955,
+      "step": 9167
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 5.00109649122807e-06,
+      "loss": 1.2628,
+      "step": 9168
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 5.000548245614036e-06,
+      "loss": 0.2626,
+      "step": 9169
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 5e-06,
+      "loss": 0.3016,
+      "step": 9170
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 4.999451754385965e-06,
+      "loss": 0.5907,
+      "step": 9171
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 4.99890350877193e-06,
+      "loss": 0.6803,
+      "step": 9172
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.998355263157895e-06,
+      "loss": 0.4053,
+      "step": 9173
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.9978070175438606e-06,
+      "loss": 1.0528,
+      "step": 9174
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.9972587719298245e-06,
+      "loss": 0.9581,
+      "step": 9175
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.99671052631579e-06,
+      "loss": 0.8557,
+      "step": 9176
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.996162280701754e-06,
+      "loss": 0.82,
+      "step": 9177
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.99561403508772e-06,
+      "loss": 0.3998,
+      "step": 9178
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.995065789473685e-06,
+      "loss": 0.9901,
+      "step": 9179
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.9945175438596495e-06,
+      "loss": 0.9631,
+      "step": 9180
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.993969298245614e-06,
+      "loss": 0.5402,
+      "step": 9181
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.99342105263158e-06,
+      "loss": 0.9418,
+      "step": 9182
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.992872807017544e-06,
+      "loss": 1.245,
+      "step": 9183
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.99232456140351e-06,
+      "loss": 1.366,
+      "step": 9184
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.991776315789474e-06,
+      "loss": 0.8483,
+      "step": 9185
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.991228070175439e-06,
+      "loss": 0.4549,
+      "step": 9186
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.990679824561404e-06,
+      "loss": 0.3909,
+      "step": 9187
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.990131578947369e-06,
+      "loss": 0.34,
+      "step": 9188
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.989583333333334e-06,
+      "loss": 0.6348,
+      "step": 9189
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.989035087719299e-06,
+      "loss": 1.3131,
+      "step": 9190
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.988486842105263e-06,
+      "loss": 0.3086,
+      "step": 9191
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.987938596491228e-06,
+      "loss": 0.2717,
+      "step": 9192
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.987390350877193e-06,
+      "loss": 0.9225,
+      "step": 9193
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.986842105263158e-06,
+      "loss": 1.0321,
+      "step": 9194
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.9862938596491236e-06,
+      "loss": 0.5304,
+      "step": 9195
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.9857456140350875e-06,
+      "loss": 0.3608,
+      "step": 9196
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.985197368421053e-06,
+      "loss": 0.1156,
+      "step": 9197
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.984649122807018e-06,
+      "loss": 0.8546,
+      "step": 9198
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.984100877192983e-06,
+      "loss": 0.7611,
+      "step": 9199
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.983552631578948e-06,
+      "loss": 0.5358,
+      "step": 9200
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.9830043859649125e-06,
+      "loss": 0.4005,
+      "step": 9201
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.982456140350877e-06,
+      "loss": 0.6127,
+      "step": 9202
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.981907894736843e-06,
+      "loss": 0.567,
+      "step": 9203
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.981359649122807e-06,
+      "loss": 0.8455,
+      "step": 9204
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.980811403508773e-06,
+      "loss": 0.3205,
+      "step": 9205
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.9802631578947375e-06,
+      "loss": 1.0733,
+      "step": 9206
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.979714912280702e-06,
+      "loss": 0.3529,
+      "step": 9207
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.979166666666667e-06,
+      "loss": 0.6942,
+      "step": 9208
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.978618421052632e-06,
+      "loss": 1.044,
+      "step": 9209
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.978070175438597e-06,
+      "loss": 0.5638,
+      "step": 9210
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.977521929824562e-06,
+      "loss": 0.2594,
+      "step": 9211
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.9769736842105264e-06,
+      "loss": 0.6527,
+      "step": 9212
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.976425438596491e-06,
+      "loss": 0.8685,
+      "step": 9213
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.975877192982457e-06,
+      "loss": 1.5173,
+      "step": 9214
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.975328947368421e-06,
+      "loss": 0.8482,
+      "step": 9215
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.974780701754387e-06,
+      "loss": 0.5308,
+      "step": 9216
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.974232456140351e-06,
+      "loss": 0.9222,
+      "step": 9217
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.973684210526316e-06,
+      "loss": 0.5258,
+      "step": 9218
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.973135964912281e-06,
+      "loss": 1.1447,
+      "step": 9219
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.972587719298246e-06,
+      "loss": 0.4264,
+      "step": 9220
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.972039473684211e-06,
+      "loss": 0.2448,
+      "step": 9221
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.971491228070176e-06,
+      "loss": 0.2826,
+      "step": 9222
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.97094298245614e-06,
+      "loss": 0.6488,
+      "step": 9223
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.970394736842106e-06,
+      "loss": 0.5708,
+      "step": 9224
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.969846491228071e-06,
+      "loss": 0.5308,
+      "step": 9225
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.969298245614036e-06,
+      "loss": 0.7679,
+      "step": 9226
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.9687500000000005e-06,
+      "loss": 0.2643,
+      "step": 9227
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.968201754385965e-06,
+      "loss": 0.9095,
+      "step": 9228
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.96765350877193e-06,
+      "loss": 0.2203,
+      "step": 9229
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.967105263157895e-06,
+      "loss": 0.2347,
+      "step": 9230
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.96655701754386e-06,
+      "loss": 0.6756,
+      "step": 9231
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.966008771929825e-06,
+      "loss": 0.7347,
+      "step": 9232
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.96546052631579e-06,
+      "loss": 0.1892,
+      "step": 9233
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.964912280701754e-06,
+      "loss": 0.4273,
+      "step": 9234
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.96436403508772e-06,
+      "loss": 0.4271,
+      "step": 9235
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.963815789473685e-06,
+      "loss": 0.6849,
+      "step": 9236
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.96326754385965e-06,
+      "loss": 0.7879,
+      "step": 9237
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.962719298245614e-06,
+      "loss": 0.4538,
+      "step": 9238
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.962171052631579e-06,
+      "loss": 0.7287,
+      "step": 9239
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.961622807017544e-06,
+      "loss": 0.6957,
+      "step": 9240
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.96107456140351e-06,
+      "loss": 0.6665,
+      "step": 9241
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.960526315789474e-06,
+      "loss": 0.5175,
+      "step": 9242
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.959978070175439e-06,
+      "loss": 0.455,
+      "step": 9243
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.959429824561403e-06,
+      "loss": 1.1002,
+      "step": 9244
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.958881578947369e-06,
+      "loss": 0.756,
+      "step": 9245
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.958333333333334e-06,
+      "loss": 0.5912,
+      "step": 9246
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.957785087719299e-06,
+      "loss": 0.814,
+      "step": 9247
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.9572368421052635e-06,
+      "loss": 0.1903,
+      "step": 9248
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.956688596491228e-06,
+      "loss": 0.399,
+      "step": 9249
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.956140350877193e-06,
+      "loss": 0.2348,
+      "step": 9250
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.955592105263158e-06,
+      "loss": 0.1715,
+      "step": 9251
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.955043859649124e-06,
+      "loss": 1.1409,
+      "step": 9252
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.954495614035088e-06,
+      "loss": 0.4641,
+      "step": 9253
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.953947368421053e-06,
+      "loss": 0.5331,
+      "step": 9254
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.953399122807018e-06,
+      "loss": 0.8444,
+      "step": 9255
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.952850877192983e-06,
+      "loss": 0.6884,
+      "step": 9256
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.952302631578948e-06,
+      "loss": 1.2282,
+      "step": 9257
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.951754385964913e-06,
+      "loss": 0.9805,
+      "step": 9258
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.9512061403508774e-06,
+      "loss": 0.619,
+      "step": 9259
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.950657894736843e-06,
+      "loss": 0.1011,
+      "step": 9260
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.950109649122807e-06,
+      "loss": 0.9808,
+      "step": 9261
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.949561403508773e-06,
+      "loss": 0.5709,
+      "step": 9262
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.949013157894737e-06,
+      "loss": 0.7401,
+      "step": 9263
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.948464912280702e-06,
+      "loss": 0.2111,
+      "step": 9264
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.947916666666667e-06,
+      "loss": 0.1497,
+      "step": 9265
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.947368421052632e-06,
+      "loss": 0.1889,
+      "step": 9266
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.946820175438597e-06,
+      "loss": 0.2892,
+      "step": 9267
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.946271929824562e-06,
+      "loss": 0.2964,
+      "step": 9268
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.9457236842105265e-06,
+      "loss": 0.6994,
+      "step": 9269
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.945175438596491e-06,
+      "loss": 0.211,
+      "step": 9270
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.944627192982456e-06,
+      "loss": 0.4501,
+      "step": 9271
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.944078947368422e-06,
+      "loss": 0.5687,
+      "step": 9272
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.943530701754387e-06,
+      "loss": 2.0553,
+      "step": 9273
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.9429824561403515e-06,
+      "loss": 0.2493,
+      "step": 9274
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.942434210526316e-06,
+      "loss": 0.6028,
+      "step": 9275
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.941885964912281e-06,
+      "loss": 0.5446,
+      "step": 9276
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.941337719298246e-06,
+      "loss": 1.0816,
+      "step": 9277
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.940789473684211e-06,
+      "loss": 0.7199,
+      "step": 9278
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.940241228070176e-06,
+      "loss": 0.4603,
+      "step": 9279
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.9396929824561404e-06,
+      "loss": 0.7209,
+      "step": 9280
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.939144736842106e-06,
+      "loss": 1.1611,
+      "step": 9281
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.93859649122807e-06,
+      "loss": 0.45,
+      "step": 9282
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.938048245614036e-06,
+      "loss": 0.4843,
+      "step": 9283
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.937500000000001e-06,
+      "loss": 0.6284,
+      "step": 9284
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.936951754385965e-06,
+      "loss": 0.6599,
+      "step": 9285
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.93640350877193e-06,
+      "loss": 0.6459,
+      "step": 9286
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.935855263157895e-06,
+      "loss": 0.7036,
+      "step": 9287
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.93530701754386e-06,
+      "loss": 0.2601,
+      "step": 9288
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.934758771929825e-06,
+      "loss": 0.3073,
+      "step": 9289
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.9342105263157895e-06,
+      "loss": 0.4219,
+      "step": 9290
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.933662280701755e-06,
+      "loss": 0.7785,
+      "step": 9291
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.93311403508772e-06,
+      "loss": 0.8408,
+      "step": 9292
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.932565789473685e-06,
+      "loss": 0.241,
+      "step": 9293
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.93201754385965e-06,
+      "loss": 0.797,
+      "step": 9294
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.9314692982456145e-06,
+      "loss": 0.8077,
+      "step": 9295
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.930921052631579e-06,
+      "loss": 0.9579,
+      "step": 9296
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.930372807017544e-06,
+      "loss": 0.8224,
+      "step": 9297
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.929824561403509e-06,
+      "loss": 0.586,
+      "step": 9298
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.929276315789474e-06,
+      "loss": 0.5608,
+      "step": 9299
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.9287280701754395e-06,
+      "loss": 0.8633,
+      "step": 9300
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.9281798245614035e-06,
+      "loss": 0.1506,
+      "step": 9301
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.927631578947369e-06,
+      "loss": 0.2279,
+      "step": 9302
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.927083333333334e-06,
+      "loss": 0.9098,
+      "step": 9303
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.926535087719299e-06,
+      "loss": 0.6746,
+      "step": 9304
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.925986842105264e-06,
+      "loss": 0.5279,
+      "step": 9305
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.925438596491228e-06,
+      "loss": 0.9399,
+      "step": 9306
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.924890350877193e-06,
+      "loss": 0.2719,
+      "step": 9307
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.924342105263158e-06,
+      "loss": 0.8554,
+      "step": 9308
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.923793859649123e-06,
+      "loss": 0.4294,
+      "step": 9309
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.9232456140350886e-06,
+      "loss": 0.3945,
+      "step": 9310
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.922697368421053e-06,
+      "loss": 1.5798,
+      "step": 9311
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.922149122807018e-06,
+      "loss": 0.4852,
+      "step": 9312
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.921600877192983e-06,
+      "loss": 0.6765,
+      "step": 9313
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.921052631578948e-06,
+      "loss": 0.829,
+      "step": 9314
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.920504385964913e-06,
+      "loss": 0.7081,
+      "step": 9315
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.9199561403508775e-06,
+      "loss": 0.8179,
+      "step": 9316
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.919407894736842e-06,
+      "loss": 0.5439,
+      "step": 9317
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.918859649122807e-06,
+      "loss": 0.603,
+      "step": 9318
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.918311403508773e-06,
+      "loss": 0.5768,
+      "step": 9319
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.917763157894737e-06,
+      "loss": 0.5613,
+      "step": 9320
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.9172149122807025e-06,
+      "loss": 0.5505,
+      "step": 9321
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.9166666666666665e-06,
+      "loss": 0.466,
+      "step": 9322
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.916118421052632e-06,
+      "loss": 0.4554,
+      "step": 9323
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.915570175438597e-06,
+      "loss": 0.6669,
+      "step": 9324
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.915021929824562e-06,
+      "loss": 0.3351,
+      "step": 9325
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.914473684210527e-06,
+      "loss": 0.5014,
+      "step": 9326
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.9139254385964914e-06,
+      "loss": 0.6711,
+      "step": 9327
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.913377192982456e-06,
+      "loss": 0.7611,
+      "step": 9328
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.912828947368422e-06,
+      "loss": 0.1961,
+      "step": 9329
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.912280701754386e-06,
+      "loss": 0.5962,
+      "step": 9330
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.911732456140352e-06,
+      "loss": 0.3898,
+      "step": 9331
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.911184210526316e-06,
+      "loss": 1.2261,
+      "step": 9332
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.910635964912281e-06,
+      "loss": 0.5523,
+      "step": 9333
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.910087719298246e-06,
+      "loss": 0.2369,
+      "step": 9334
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.909539473684211e-06,
+      "loss": 0.7343,
+      "step": 9335
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.908991228070176e-06,
+      "loss": 0.6889,
+      "step": 9336
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.9084429824561405e-06,
+      "loss": 0.9096,
+      "step": 9337
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.907894736842106e-06,
+      "loss": 0.4199,
+      "step": 9338
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.90734649122807e-06,
+      "loss": 1.1334,
+      "step": 9339
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.906798245614036e-06,
+      "loss": 0.2186,
+      "step": 9340
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.90625e-06,
+      "loss": 1.1213,
+      "step": 9341
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.9057017543859655e-06,
+      "loss": 1.4472,
+      "step": 9342
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.90515350877193e-06,
+      "loss": 0.4138,
+      "step": 9343
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.904605263157895e-06,
+      "loss": 0.4762,
+      "step": 9344
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.90405701754386e-06,
+      "loss": 0.5956,
+      "step": 9345
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.903508771929825e-06,
+      "loss": 0.8237,
+      "step": 9346
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.90296052631579e-06,
+      "loss": 0.5904,
+      "step": 9347
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.902412280701755e-06,
+      "loss": 0.4494,
+      "step": 9348
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.901864035087719e-06,
+      "loss": 0.2167,
+      "step": 9349
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.901315789473685e-06,
+      "loss": 0.5492,
+      "step": 9350
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.90076754385965e-06,
+      "loss": 0.4193,
+      "step": 9351
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.900219298245615e-06,
+      "loss": 0.4534,
+      "step": 9352
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.899671052631579e-06,
+      "loss": 0.4221,
+      "step": 9353
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.899122807017544e-06,
+      "loss": 0.3984,
+      "step": 9354
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.898574561403509e-06,
+      "loss": 0.6042,
+      "step": 9355
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.898026315789474e-06,
+      "loss": 0.1784,
+      "step": 9356
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.897478070175439e-06,
+      "loss": 0.4439,
+      "step": 9357
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.8969298245614035e-06,
+      "loss": 1.0018,
+      "step": 9358
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.896381578947369e-06,
+      "loss": 1.5329,
+      "step": 9359
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.895833333333333e-06,
+      "loss": 0.2663,
+      "step": 9360
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.895285087719299e-06,
+      "loss": 0.7116,
+      "step": 9361
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.894736842105264e-06,
+      "loss": 0.4582,
+      "step": 9362
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.8941885964912285e-06,
+      "loss": 0.8891,
+      "step": 9363
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.893640350877193e-06,
+      "loss": 0.483,
+      "step": 9364
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.893092105263158e-06,
+      "loss": 0.8819,
+      "step": 9365
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.892543859649123e-06,
+      "loss": 0.6487,
+      "step": 9366
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.891995614035089e-06,
+      "loss": 0.2841,
+      "step": 9367
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.891447368421053e-06,
+      "loss": 1.1063,
+      "step": 9368
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.890899122807018e-06,
+      "loss": 0.2497,
+      "step": 9369
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.890350877192983e-06,
+      "loss": 0.3783,
+      "step": 9370
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.889802631578948e-06,
+      "loss": 0.9577,
+      "step": 9371
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.889254385964913e-06,
+      "loss": 0.8023,
+      "step": 9372
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.888706140350878e-06,
+      "loss": 0.7235,
+      "step": 9373
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.8881578947368424e-06,
+      "loss": 0.2705,
+      "step": 9374
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.887609649122807e-06,
+      "loss": 1.2467,
+      "step": 9375
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.887061403508772e-06,
+      "loss": 0.512,
+      "step": 9376
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.886513157894737e-06,
+      "loss": 0.6455,
+      "step": 9377
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.8859649122807026e-06,
+      "loss": 0.5886,
+      "step": 9378
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.8854166666666665e-06,
+      "loss": 0.6647,
+      "step": 9379
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.884868421052632e-06,
+      "loss": 0.8456,
+      "step": 9380
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.884320175438597e-06,
+      "loss": 0.3902,
+      "step": 9381
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.883771929824562e-06,
+      "loss": 0.5992,
+      "step": 9382
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.883223684210527e-06,
+      "loss": 0.4302,
+      "step": 9383
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.8826754385964915e-06,
+      "loss": 0.5755,
+      "step": 9384
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.882127192982456e-06,
+      "loss": 1.0671,
+      "step": 9385
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.881578947368422e-06,
+      "loss": 0.6685,
+      "step": 9386
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.881030701754386e-06,
+      "loss": 0.4358,
+      "step": 9387
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.880482456140352e-06,
+      "loss": 1.3523,
+      "step": 9388
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.8799342105263165e-06,
+      "loss": 0.9048,
+      "step": 9389
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.879385964912281e-06,
+      "loss": 0.2216,
+      "step": 9390
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.878837719298246e-06,
+      "loss": 1.2204,
+      "step": 9391
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.878289473684211e-06,
+      "loss": 0.3099,
+      "step": 9392
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.877741228070176e-06,
+      "loss": 0.2337,
+      "step": 9393
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.877192982456141e-06,
+      "loss": 0.1077,
+      "step": 9394
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.8766447368421054e-06,
+      "loss": 0.13,
+      "step": 9395
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.87609649122807e-06,
+      "loss": 1.0365,
+      "step": 9396
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.875548245614036e-06,
+      "loss": 0.7947,
+      "step": 9397
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.875e-06,
+      "loss": 1.4513,
+      "step": 9398
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.874451754385966e-06,
+      "loss": 0.8491,
+      "step": 9399
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.8739035087719296e-06,
+      "loss": 1.1183,
+      "step": 9400
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.873355263157895e-06,
+      "loss": 0.4802,
+      "step": 9401
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.87280701754386e-06,
+      "loss": 0.9924,
+      "step": 9402
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.872258771929825e-06,
+      "loss": 0.9112,
+      "step": 9403
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.87171052631579e-06,
+      "loss": 0.2998,
+      "step": 9404
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.871162280701755e-06,
+      "loss": 1.1693,
+      "step": 9405
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.870614035087719e-06,
+      "loss": 0.5869,
+      "step": 9406
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.870065789473685e-06,
+      "loss": 0.2731,
+      "step": 9407
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.869517543859649e-06,
+      "loss": 0.5306,
+      "step": 9408
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.868969298245615e-06,
+      "loss": 0.3013,
+      "step": 9409
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.8684210526315795e-06,
+      "loss": 1.0703,
+      "step": 9410
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.867872807017544e-06,
+      "loss": 0.9728,
+      "step": 9411
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.867324561403509e-06,
+      "loss": 0.6433,
+      "step": 9412
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.866776315789474e-06,
+      "loss": 0.5971,
+      "step": 9413
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.866228070175439e-06,
+      "loss": 0.702,
+      "step": 9414
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.865679824561404e-06,
+      "loss": 0.8498,
+      "step": 9415
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.865131578947369e-06,
+      "loss": 1.2176,
+      "step": 9416
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.864583333333333e-06,
+      "loss": 0.8501,
+      "step": 9417
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.864035087719299e-06,
+      "loss": 0.6151,
+      "step": 9418
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.863486842105263e-06,
+      "loss": 0.7703,
+      "step": 9419
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.862938596491229e-06,
+      "loss": 0.6828,
+      "step": 9420
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.862390350877193e-06,
+      "loss": 0.4835,
+      "step": 9421
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.861842105263158e-06,
+      "loss": 0.6219,
+      "step": 9422
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.861293859649123e-06,
+      "loss": 0.2401,
+      "step": 9423
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.860745614035089e-06,
+      "loss": 0.9794,
+      "step": 9424
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.860197368421053e-06,
+      "loss": 0.8457,
+      "step": 9425
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.859649122807018e-06,
+      "loss": 0.2674,
+      "step": 9426
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.859100877192982e-06,
+      "loss": 0.5385,
+      "step": 9427
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.858552631578948e-06,
+      "loss": 0.1966,
+      "step": 9428
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.858004385964913e-06,
+      "loss": 1.5161,
+      "step": 9429
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.857456140350878e-06,
+      "loss": 0.2928,
+      "step": 9430
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.8569078947368425e-06,
+      "loss": 1.1169,
+      "step": 9431
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.856359649122807e-06,
+      "loss": 1.575,
+      "step": 9432
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.855811403508772e-06,
+      "loss": 0.524,
+      "step": 9433
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.855263157894737e-06,
+      "loss": 0.3355,
+      "step": 9434
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.854714912280702e-06,
+      "loss": 0.3111,
+      "step": 9435
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.854166666666667e-06,
+      "loss": 0.4913,
+      "step": 9436
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.853618421052632e-06,
+      "loss": 0.6318,
+      "step": 9437
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.853070175438596e-06,
+      "loss": 0.4488,
+      "step": 9438
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.852521929824562e-06,
+      "loss": 0.2957,
+      "step": 9439
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.851973684210527e-06,
+      "loss": 0.7256,
+      "step": 9440
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.851425438596492e-06,
+      "loss": 0.981,
+      "step": 9441
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.8508771929824564e-06,
+      "loss": 0.7194,
+      "step": 9442
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.850328947368421e-06,
+      "loss": 0.7088,
+      "step": 9443
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.849780701754386e-06,
+      "loss": 0.5726,
+      "step": 9444
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.849232456140352e-06,
+      "loss": 0.5743,
+      "step": 9445
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.848684210526316e-06,
+      "loss": 0.7683,
+      "step": 9446
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.848135964912281e-06,
+      "loss": 0.5604,
+      "step": 9447
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.847587719298246e-06,
+      "loss": 0.5158,
+      "step": 9448
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.847039473684211e-06,
+      "loss": 0.231,
+      "step": 9449
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.846491228070176e-06,
+      "loss": 1.2862,
+      "step": 9450
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.845942982456141e-06,
+      "loss": 0.1449,
+      "step": 9451
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.8453947368421055e-06,
+      "loss": 1.0511,
+      "step": 9452
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.84484649122807e-06,
+      "loss": 0.4256,
+      "step": 9453
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.844298245614035e-06,
+      "loss": 0.3082,
+      "step": 9454
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.84375e-06,
+      "loss": 0.8099,
+      "step": 9455
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.843201754385966e-06,
+      "loss": 1.409,
+      "step": 9456
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.8426535087719305e-06,
+      "loss": 0.9988,
+      "step": 9457
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.842105263157895e-06,
+      "loss": 0.4909,
+      "step": 9458
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.84155701754386e-06,
+      "loss": 0.3109,
+      "step": 9459
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.841008771929825e-06,
+      "loss": 0.501,
+      "step": 9460
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.84046052631579e-06,
+      "loss": 0.1516,
+      "step": 9461
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.839912280701755e-06,
+      "loss": 1.2107,
+      "step": 9462
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.8393640350877194e-06,
+      "loss": 1.393,
+      "step": 9463
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.838815789473685e-06,
+      "loss": 0.4929,
+      "step": 9464
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.838267543859649e-06,
+      "loss": 0.7237,
+      "step": 9465
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.837719298245615e-06,
+      "loss": 0.2893,
+      "step": 9466
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.83717105263158e-06,
+      "loss": 0.4643,
+      "step": 9467
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.836622807017544e-06,
+      "loss": 0.3017,
+      "step": 9468
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.836074561403509e-06,
+      "loss": 0.8354,
+      "step": 9469
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.835526315789474e-06,
+      "loss": 0.6668,
+      "step": 9470
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.834978070175439e-06,
+      "loss": 0.5673,
+      "step": 9471
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.834429824561404e-06,
+      "loss": 0.7419,
+      "step": 9472
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.8338815789473685e-06,
+      "loss": 0.6564,
+      "step": 9473
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.833333333333333e-06,
+      "loss": 0.7355,
+      "step": 9474
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.832785087719299e-06,
+      "loss": 0.3984,
+      "step": 9475
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.832236842105264e-06,
+      "loss": 0.2464,
+      "step": 9476
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.831688596491229e-06,
+      "loss": 0.5095,
+      "step": 9477
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.8311403508771935e-06,
+      "loss": 1.047,
+      "step": 9478
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.830592105263158e-06,
+      "loss": 0.5655,
+      "step": 9479
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.830043859649123e-06,
+      "loss": 0.9571,
+      "step": 9480
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.829495614035088e-06,
+      "loss": 1.3722,
+      "step": 9481
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.828947368421053e-06,
+      "loss": 0.6198,
+      "step": 9482
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.8283991228070185e-06,
+      "loss": 0.6761,
+      "step": 9483
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.8278508771929825e-06,
+      "loss": 0.7598,
+      "step": 9484
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.827302631578948e-06,
+      "loss": 0.5347,
+      "step": 9485
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.826754385964912e-06,
+      "loss": 0.9263,
+      "step": 9486
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.826206140350878e-06,
+      "loss": 0.3467,
+      "step": 9487
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.825657894736843e-06,
+      "loss": 0.42,
+      "step": 9488
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.8251096491228074e-06,
+      "loss": 0.753,
+      "step": 9489
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.824561403508772e-06,
+      "loss": 0.8599,
+      "step": 9490
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.824013157894737e-06,
+      "loss": 0.2837,
+      "step": 9491
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.823464912280702e-06,
+      "loss": 0.5952,
+      "step": 9492
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.822916666666667e-06,
+      "loss": 0.5862,
+      "step": 9493
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.8223684210526315e-06,
+      "loss": 1.043,
+      "step": 9494
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.821820175438597e-06,
+      "loss": 0.7096,
+      "step": 9495
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.821271929824562e-06,
+      "loss": 0.398,
+      "step": 9496
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.820723684210527e-06,
+      "loss": 0.2035,
+      "step": 9497
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.820175438596492e-06,
+      "loss": 0.7969,
+      "step": 9498
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.8196271929824565e-06,
+      "loss": 1.4038,
+      "step": 9499
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.819078947368421e-06,
+      "loss": 1.4029,
+      "step": 9500
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.818530701754386e-06,
+      "loss": 0.2281,
+      "step": 9501
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.817982456140352e-06,
+      "loss": 0.9562,
+      "step": 9502
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.817434210526316e-06,
+      "loss": 0.5891,
+      "step": 9503
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.8168859649122815e-06,
+      "loss": 0.4735,
+      "step": 9504
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.8163377192982455e-06,
+      "loss": 1.1518,
+      "step": 9505
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.815789473684211e-06,
+      "loss": 1.0141,
+      "step": 9506
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.815241228070176e-06,
+      "loss": 0.6546,
+      "step": 9507
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.814692982456141e-06,
+      "loss": 0.5482,
+      "step": 9508
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.814144736842106e-06,
+      "loss": 0.8942,
+      "step": 9509
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.8135964912280704e-06,
+      "loss": 1.2846,
+      "step": 9510
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.813048245614035e-06,
+      "loss": 0.3537,
+      "step": 9511
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.8125e-06,
+      "loss": 0.3714,
+      "step": 9512
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.811951754385965e-06,
+      "loss": 0.3159,
+      "step": 9513
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.811403508771931e-06,
+      "loss": 1.0618,
+      "step": 9514
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.810855263157895e-06,
+      "loss": 0.8142,
+      "step": 9515
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.81030701754386e-06,
+      "loss": 0.4331,
+      "step": 9516
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.809758771929825e-06,
+      "loss": 1.304,
+      "step": 9517
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.80921052631579e-06,
+      "loss": 0.9397,
+      "step": 9518
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.808662280701755e-06,
+      "loss": 0.5521,
+      "step": 9519
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.8081140350877195e-06,
+      "loss": 0.6013,
+      "step": 9520
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.807565789473684e-06,
+      "loss": 0.3484,
+      "step": 9521
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.807017543859649e-06,
+      "loss": 0.9328,
+      "step": 9522
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.806469298245615e-06,
+      "loss": 0.237,
+      "step": 9523
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.805921052631579e-06,
+      "loss": 0.3777,
+      "step": 9524
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.8053728070175445e-06,
+      "loss": 0.6879,
+      "step": 9525
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.804824561403509e-06,
+      "loss": 0.4701,
+      "step": 9526
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.804276315789474e-06,
+      "loss": 1.0091,
+      "step": 9527
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.803728070175439e-06,
+      "loss": 0.4141,
+      "step": 9528
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.803179824561404e-06,
+      "loss": 0.6318,
+      "step": 9529
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.802631578947369e-06,
+      "loss": 1.2413,
+      "step": 9530
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.802083333333334e-06,
+      "loss": 1.0431,
+      "step": 9531
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.801535087719298e-06,
+      "loss": 0.7228,
+      "step": 9532
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.800986842105264e-06,
+      "loss": 1.1746,
+      "step": 9533
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.800438596491229e-06,
+      "loss": 0.9558,
+      "step": 9534
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.799890350877194e-06,
+      "loss": 0.7806,
+      "step": 9535
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.799342105263158e-06,
+      "loss": 0.156,
+      "step": 9536
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.798793859649123e-06,
+      "loss": 0.8623,
+      "step": 9537
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.798245614035088e-06,
+      "loss": 1.0434,
+      "step": 9538
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.797697368421053e-06,
+      "loss": 0.9297,
+      "step": 9539
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.797149122807018e-06,
+      "loss": 0.4798,
+      "step": 9540
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.7966008771929825e-06,
+      "loss": 1.1519,
+      "step": 9541
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.796052631578948e-06,
+      "loss": 0.7902,
+      "step": 9542
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.795504385964912e-06,
+      "loss": 0.4436,
+      "step": 9543
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.794956140350878e-06,
+      "loss": 0.4983,
+      "step": 9544
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.794407894736843e-06,
+      "loss": 0.9824,
+      "step": 9545
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.7938596491228075e-06,
+      "loss": 0.4798,
+      "step": 9546
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.793311403508772e-06,
+      "loss": 0.8458,
+      "step": 9547
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.792763157894737e-06,
+      "loss": 0.5952,
+      "step": 9548
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.792214912280702e-06,
+      "loss": 0.9273,
+      "step": 9549
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.791666666666668e-06,
+      "loss": 1.022,
+      "step": 9550
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.791118421052632e-06,
+      "loss": 0.0981,
+      "step": 9551
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.790570175438597e-06,
+      "loss": 0.4254,
+      "step": 9552
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.790021929824562e-06,
+      "loss": 0.5096,
+      "step": 9553
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.789473684210527e-06,
+      "loss": 0.523,
+      "step": 9554
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.788925438596492e-06,
+      "loss": 1.6494,
+      "step": 9555
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.788377192982457e-06,
+      "loss": 1.2237,
+      "step": 9556
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.7878289473684214e-06,
+      "loss": 0.2992,
+      "step": 9557
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.787280701754386e-06,
+      "loss": 0.8712,
+      "step": 9558
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.786732456140351e-06,
+      "loss": 0.4405,
+      "step": 9559
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.786184210526316e-06,
+      "loss": 0.9404,
+      "step": 9560
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.785635964912282e-06,
+      "loss": 0.9547,
+      "step": 9561
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.7850877192982456e-06,
+      "loss": 0.5441,
+      "step": 9562
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.784539473684211e-06,
+      "loss": 0.3398,
+      "step": 9563
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.783991228070175e-06,
+      "loss": 0.3511,
+      "step": 9564
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.783442982456141e-06,
+      "loss": 0.5043,
+      "step": 9565
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.782894736842106e-06,
+      "loss": 0.7118,
+      "step": 9566
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.7823464912280705e-06,
+      "loss": 0.5972,
+      "step": 9567
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.781798245614035e-06,
+      "loss": 0.5374,
+      "step": 9568
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.781250000000001e-06,
+      "loss": 1.235,
+      "step": 9569
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.780701754385965e-06,
+      "loss": 0.2756,
+      "step": 9570
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.780153508771931e-06,
+      "loss": 0.7776,
+      "step": 9571
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.779605263157895e-06,
+      "loss": 0.4884,
+      "step": 9572
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.77905701754386e-06,
+      "loss": 0.1717,
+      "step": 9573
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.778508771929825e-06,
+      "loss": 0.4446,
+      "step": 9574
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.77796052631579e-06,
+      "loss": 0.6022,
+      "step": 9575
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.777412280701755e-06,
+      "loss": 0.6007,
+      "step": 9576
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.77686403508772e-06,
+      "loss": 0.8297,
+      "step": 9577
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.7763157894736844e-06,
+      "loss": 0.6271,
+      "step": 9578
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.775767543859649e-06,
+      "loss": 0.1265,
+      "step": 9579
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.775219298245615e-06,
+      "loss": 0.3279,
+      "step": 9580
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.774671052631579e-06,
+      "loss": 0.918,
+      "step": 9581
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.774122807017545e-06,
+      "loss": 0.9086,
+      "step": 9582
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.7735745614035086e-06,
+      "loss": 0.3208,
+      "step": 9583
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.773026315789474e-06,
+      "loss": 0.4686,
+      "step": 9584
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.772478070175439e-06,
+      "loss": 0.4785,
+      "step": 9585
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.771929824561404e-06,
+      "loss": 0.5231,
+      "step": 9586
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.771381578947369e-06,
+      "loss": 0.4208,
+      "step": 9587
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.770833333333334e-06,
+      "loss": 0.6361,
+      "step": 9588
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.770285087719298e-06,
+      "loss": 0.7678,
+      "step": 9589
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.769736842105264e-06,
+      "loss": 1.4183,
+      "step": 9590
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.769188596491228e-06,
+      "loss": 0.7461,
+      "step": 9591
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.768640350877194e-06,
+      "loss": 0.8195,
+      "step": 9592
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.7680921052631585e-06,
+      "loss": 1.2509,
+      "step": 9593
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.767543859649123e-06,
+      "loss": 0.6999,
+      "step": 9594
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.766995614035088e-06,
+      "loss": 1.3044,
+      "step": 9595
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.766447368421053e-06,
+      "loss": 0.4722,
+      "step": 9596
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.765899122807018e-06,
+      "loss": 1.0653,
+      "step": 9597
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.765350877192983e-06,
+      "loss": 0.7998,
+      "step": 9598
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.7648026315789475e-06,
+      "loss": 0.2844,
+      "step": 9599
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.764254385964912e-06,
+      "loss": 0.2088,
+      "step": 9600
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.763706140350878e-06,
+      "loss": 0.33,
+      "step": 9601
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.763157894736842e-06,
+      "loss": 1.218,
+      "step": 9602
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.762609649122808e-06,
+      "loss": 0.3692,
+      "step": 9603
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.7620614035087724e-06,
+      "loss": 0.5695,
+      "step": 9604
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.761513157894737e-06,
+      "loss": 0.7988,
+      "step": 9605
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.760964912280702e-06,
+      "loss": 0.5387,
+      "step": 9606
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.760416666666667e-06,
+      "loss": 0.512,
+      "step": 9607
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.759868421052632e-06,
+      "loss": 0.7591,
+      "step": 9608
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.759320175438597e-06,
+      "loss": 1.0077,
+      "step": 9609
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.758771929824561e-06,
+      "loss": 0.1283,
+      "step": 9610
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.758223684210527e-06,
+      "loss": 1.1129,
+      "step": 9611
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.757675438596492e-06,
+      "loss": 1.1701,
+      "step": 9612
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.757127192982457e-06,
+      "loss": 0.2684,
+      "step": 9613
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.7565789473684215e-06,
+      "loss": 0.4281,
+      "step": 9614
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.756030701754386e-06,
+      "loss": 0.493,
+      "step": 9615
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.755482456140351e-06,
+      "loss": 1.0181,
+      "step": 9616
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.754934210526316e-06,
+      "loss": 0.3399,
+      "step": 9617
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.754385964912281e-06,
+      "loss": 0.6754,
+      "step": 9618
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.753837719298246e-06,
+      "loss": 0.5031,
+      "step": 9619
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.753289473684211e-06,
+      "loss": 0.6091,
+      "step": 9620
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.752741228070175e-06,
+      "loss": 0.2999,
+      "step": 9621
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.752192982456141e-06,
+      "loss": 0.4636,
+      "step": 9622
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.751644736842106e-06,
+      "loss": 0.645,
+      "step": 9623
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.751096491228071e-06,
+      "loss": 0.9608,
+      "step": 9624
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.7505482456140354e-06,
+      "loss": 0.651,
+      "step": 9625
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.75e-06,
+      "loss": 1.1811,
+      "step": 9626
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.749451754385965e-06,
+      "loss": 1.2306,
+      "step": 9627
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.748903508771931e-06,
+      "loss": 1.0213,
+      "step": 9628
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.748355263157895e-06,
+      "loss": 0.3999,
+      "step": 9629
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.74780701754386e-06,
+      "loss": 0.5344,
+      "step": 9630
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.747258771929825e-06,
+      "loss": 0.5092,
+      "step": 9631
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.74671052631579e-06,
+      "loss": 0.4493,
+      "step": 9632
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.746162280701755e-06,
+      "loss": 0.7287,
+      "step": 9633
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.74561403508772e-06,
+      "loss": 0.3667,
+      "step": 9634
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.7450657894736845e-06,
+      "loss": 0.3043,
+      "step": 9635
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.744517543859649e-06,
+      "loss": 0.5579,
+      "step": 9636
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.743969298245614e-06,
+      "loss": 0.6477,
+      "step": 9637
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.743421052631579e-06,
+      "loss": 0.6539,
+      "step": 9638
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.742872807017545e-06,
+      "loss": 0.2704,
+      "step": 9639
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.742324561403509e-06,
+      "loss": 1.7092,
+      "step": 9640
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.741776315789474e-06,
+      "loss": 0.2163,
+      "step": 9641
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.741228070175439e-06,
+      "loss": 1.1243,
+      "step": 9642
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.740679824561404e-06,
+      "loss": 0.5403,
+      "step": 9643
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.740131578947369e-06,
+      "loss": 0.8249,
+      "step": 9644
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.739583333333334e-06,
+      "loss": 0.5753,
+      "step": 9645
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.7390350877192984e-06,
+      "loss": 0.7763,
+      "step": 9646
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.738486842105264e-06,
+      "loss": 0.7065,
+      "step": 9647
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.737938596491228e-06,
+      "loss": 1.008,
+      "step": 9648
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.737390350877194e-06,
+      "loss": 0.1798,
+      "step": 9649
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.736842105263158e-06,
+      "loss": 0.197,
+      "step": 9650
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.736293859649123e-06,
+      "loss": 0.6869,
+      "step": 9651
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.735745614035088e-06,
+      "loss": 1.2138,
+      "step": 9652
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.735197368421053e-06,
+      "loss": 0.6246,
+      "step": 9653
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.734649122807018e-06,
+      "loss": 0.6875,
+      "step": 9654
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.734100877192983e-06,
+      "loss": 0.853,
+      "step": 9655
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.7335526315789475e-06,
+      "loss": 0.6484,
+      "step": 9656
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.733004385964912e-06,
+      "loss": 0.7161,
+      "step": 9657
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.732456140350877e-06,
+      "loss": 0.8348,
+      "step": 9658
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.731907894736842e-06,
+      "loss": 0.2559,
+      "step": 9659
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.731359649122808e-06,
+      "loss": 0.7921,
+      "step": 9660
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.7308114035087725e-06,
+      "loss": 0.2209,
+      "step": 9661
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.730263157894737e-06,
+      "loss": 0.7526,
+      "step": 9662
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.729714912280702e-06,
+      "loss": 0.8097,
+      "step": 9663
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.729166666666667e-06,
+      "loss": 0.5644,
+      "step": 9664
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.728618421052632e-06,
+      "loss": 0.238,
+      "step": 9665
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.7280701754385975e-06,
+      "loss": 0.815,
+      "step": 9666
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.7275219298245615e-06,
+      "loss": 0.5944,
+      "step": 9667
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.726973684210527e-06,
+      "loss": 0.723,
+      "step": 9668
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.726425438596491e-06,
+      "loss": 0.4329,
+      "step": 9669
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.725877192982457e-06,
+      "loss": 0.8011,
+      "step": 9670
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.725328947368422e-06,
+      "loss": 0.4406,
+      "step": 9671
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.7247807017543864e-06,
+      "loss": 0.5033,
+      "step": 9672
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.724232456140351e-06,
+      "loss": 0.8978,
+      "step": 9673
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.723684210526316e-06,
+      "loss": 0.2783,
+      "step": 9674
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.723135964912281e-06,
+      "loss": 0.9,
+      "step": 9675
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.722587719298246e-06,
+      "loss": 0.4386,
+      "step": 9676
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.7220394736842106e-06,
+      "loss": 0.7224,
+      "step": 9677
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.721491228070175e-06,
+      "loss": 0.2564,
+      "step": 9678
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.720942982456141e-06,
+      "loss": 0.7693,
+      "step": 9679
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.720394736842106e-06,
+      "loss": 0.9639,
+      "step": 9680
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.719846491228071e-06,
+      "loss": 0.4863,
+      "step": 9681
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.7192982456140355e-06,
+      "loss": 0.2937,
+      "step": 9682
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.71875e-06,
+      "loss": 0.7455,
+      "step": 9683
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.718201754385965e-06,
+      "loss": 0.5677,
+      "step": 9684
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.71765350877193e-06,
+      "loss": 0.3497,
+      "step": 9685
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.717105263157895e-06,
+      "loss": 0.3791,
+      "step": 9686
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.7165570175438605e-06,
+      "loss": 0.4018,
+      "step": 9687
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.7160087719298245e-06,
+      "loss": 0.5249,
+      "step": 9688
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.71546052631579e-06,
+      "loss": 0.4767,
+      "step": 9689
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.714912280701755e-06,
+      "loss": 0.8657,
+      "step": 9690
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.71436403508772e-06,
+      "loss": 0.622,
+      "step": 9691
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.713815789473685e-06,
+      "loss": 0.8841,
+      "step": 9692
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.7132675438596494e-06,
+      "loss": 0.7292,
+      "step": 9693
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.712719298245614e-06,
+      "loss": 0.4926,
+      "step": 9694
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.712171052631579e-06,
+      "loss": 0.2245,
+      "step": 9695
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.711622807017544e-06,
+      "loss": 0.3112,
+      "step": 9696
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.711074561403509e-06,
+      "loss": 0.6482,
+      "step": 9697
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.710526315789474e-06,
+      "loss": 0.5267,
+      "step": 9698
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.709978070175439e-06,
+      "loss": 0.268,
+      "step": 9699
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.709429824561404e-06,
+      "loss": 1.999,
+      "step": 9700
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.708881578947369e-06,
+      "loss": 0.2782,
+      "step": 9701
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.708333333333334e-06,
+      "loss": 0.6433,
+      "step": 9702
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.7077850877192985e-06,
+      "loss": 0.2475,
+      "step": 9703
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.707236842105263e-06,
+      "loss": 0.5216,
+      "step": 9704
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.706688596491228e-06,
+      "loss": 1.3971,
+      "step": 9705
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.706140350877194e-06,
+      "loss": 0.5767,
+      "step": 9706
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.705592105263158e-06,
+      "loss": 1.7814,
+      "step": 9707
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.7050438596491235e-06,
+      "loss": 0.3988,
+      "step": 9708
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.704495614035088e-06,
+      "loss": 0.7045,
+      "step": 9709
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.703947368421053e-06,
+      "loss": 0.2615,
+      "step": 9710
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.703399122807018e-06,
+      "loss": 0.2932,
+      "step": 9711
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.702850877192983e-06,
+      "loss": 0.9892,
+      "step": 9712
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.702302631578948e-06,
+      "loss": 0.1514,
+      "step": 9713
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.7017543859649125e-06,
+      "loss": 1.0906,
+      "step": 9714
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.701206140350877e-06,
+      "loss": 0.4067,
+      "step": 9715
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.700657894736843e-06,
+      "loss": 0.2768,
+      "step": 9716
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.700109649122808e-06,
+      "loss": 0.4281,
+      "step": 9717
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.699561403508773e-06,
+      "loss": 0.2398,
+      "step": 9718
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.6990131578947374e-06,
+      "loss": 0.5434,
+      "step": 9719
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.698464912280702e-06,
+      "loss": 0.8911,
+      "step": 9720
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.697916666666667e-06,
+      "loss": 0.4049,
+      "step": 9721
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.697368421052632e-06,
+      "loss": 0.74,
+      "step": 9722
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.696820175438597e-06,
+      "loss": 0.4881,
+      "step": 9723
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.6962719298245615e-06,
+      "loss": 0.5511,
+      "step": 9724
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.695723684210527e-06,
+      "loss": 0.4245,
+      "step": 9725
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.695175438596491e-06,
+      "loss": 0.3494,
+      "step": 9726
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.694627192982457e-06,
+      "loss": 0.7844,
+      "step": 9727
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.694078947368421e-06,
+      "loss": 1.0386,
+      "step": 9728
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.6935307017543865e-06,
+      "loss": 0.5707,
+      "step": 9729
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.692982456140351e-06,
+      "loss": 0.5449,
+      "step": 9730
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.692434210526316e-06,
+      "loss": 0.601,
+      "step": 9731
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.691885964912281e-06,
+      "loss": 0.1566,
+      "step": 9732
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.691337719298246e-06,
+      "loss": 1.3203,
+      "step": 9733
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.690789473684211e-06,
+      "loss": 0.1314,
+      "step": 9734
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.690241228070176e-06,
+      "loss": 1.8924,
+      "step": 9735
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.68969298245614e-06,
+      "loss": 0.5801,
+      "step": 9736
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.689144736842106e-06,
+      "loss": 0.7325,
+      "step": 9737
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.688596491228071e-06,
+      "loss": 1.7371,
+      "step": 9738
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.688048245614036e-06,
+      "loss": 0.3962,
+      "step": 9739
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.6875000000000004e-06,
+      "loss": 0.4461,
+      "step": 9740
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.686951754385965e-06,
+      "loss": 0.8907,
+      "step": 9741
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.68640350877193e-06,
+      "loss": 0.626,
+      "step": 9742
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.685855263157895e-06,
+      "loss": 0.3982,
+      "step": 9743
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.685307017543861e-06,
+      "loss": 0.5731,
+      "step": 9744
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.6847587719298246e-06,
+      "loss": 0.501,
+      "step": 9745
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.68421052631579e-06,
+      "loss": 0.5524,
+      "step": 9746
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.683662280701754e-06,
+      "loss": 1.0032,
+      "step": 9747
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.68311403508772e-06,
+      "loss": 0.6979,
+      "step": 9748
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.682565789473685e-06,
+      "loss": 0.5672,
+      "step": 9749
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.6820175438596495e-06,
+      "loss": 1.245,
+      "step": 9750
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.681469298245614e-06,
+      "loss": 0.2705,
+      "step": 9751
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.680921052631579e-06,
+      "loss": 0.442,
+      "step": 9752
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.680372807017544e-06,
+      "loss": 0.4486,
+      "step": 9753
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.67982456140351e-06,
+      "loss": 0.9987,
+      "step": 9754
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.679276315789474e-06,
+      "loss": 0.5683,
+      "step": 9755
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.678728070175439e-06,
+      "loss": 0.2523,
+      "step": 9756
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.678179824561404e-06,
+      "loss": 1.0114,
+      "step": 9757
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.677631578947369e-06,
+      "loss": 0.937,
+      "step": 9758
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.677083333333334e-06,
+      "loss": 0.2304,
+      "step": 9759
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.676535087719299e-06,
+      "loss": 0.4904,
+      "step": 9760
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.6759868421052634e-06,
+      "loss": 1.2365,
+      "step": 9761
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.675438596491228e-06,
+      "loss": 0.4253,
+      "step": 9762
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.674890350877193e-06,
+      "loss": 0.8611,
+      "step": 9763
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.674342105263158e-06,
+      "loss": 0.8974,
+      "step": 9764
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.673793859649124e-06,
+      "loss": 0.7597,
+      "step": 9765
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.6732456140350876e-06,
+      "loss": 1.0399,
+      "step": 9766
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.672697368421053e-06,
+      "loss": 0.4279,
+      "step": 9767
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.672149122807018e-06,
+      "loss": 0.5317,
+      "step": 9768
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.671600877192983e-06,
+      "loss": 0.5972,
+      "step": 9769
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.671052631578948e-06,
+      "loss": 0.7961,
+      "step": 9770
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.6705043859649125e-06,
+      "loss": 0.4081,
+      "step": 9771
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.669956140350877e-06,
+      "loss": 0.7617,
+      "step": 9772
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.669407894736843e-06,
+      "loss": 0.3648,
+      "step": 9773
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.668859649122807e-06,
+      "loss": 0.5087,
+      "step": 9774
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.668311403508773e-06,
+      "loss": 0.2828,
+      "step": 9775
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.6677631578947375e-06,
+      "loss": 0.4152,
+      "step": 9776
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.667214912280702e-06,
+      "loss": 0.6187,
+      "step": 9777
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.666666666666667e-06,
+      "loss": 1.5008,
+      "step": 9778
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.666118421052632e-06,
+      "loss": 0.4504,
+      "step": 9779
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.665570175438597e-06,
+      "loss": 0.574,
+      "step": 9780
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.665021929824562e-06,
+      "loss": 0.4236,
+      "step": 9781
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.6644736842105265e-06,
+      "loss": 1.0386,
+      "step": 9782
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.663925438596491e-06,
+      "loss": 0.6007,
+      "step": 9783
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.663377192982457e-06,
+      "loss": 0.4199,
+      "step": 9784
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.662828947368421e-06,
+      "loss": 1.6732,
+      "step": 9785
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.662280701754387e-06,
+      "loss": 0.3855,
+      "step": 9786
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.6617324561403514e-06,
+      "loss": 0.9164,
+      "step": 9787
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.661184210526316e-06,
+      "loss": 0.7095,
+      "step": 9788
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.660635964912281e-06,
+      "loss": 0.8971,
+      "step": 9789
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.660087719298246e-06,
+      "loss": 0.4758,
+      "step": 9790
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.659539473684211e-06,
+      "loss": 1.5665,
+      "step": 9791
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.658991228070176e-06,
+      "loss": 0.8342,
+      "step": 9792
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.65844298245614e-06,
+      "loss": 1.0883,
+      "step": 9793
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.657894736842106e-06,
+      "loss": 1.9146,
+      "step": 9794
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.657346491228071e-06,
+      "loss": 0.307,
+      "step": 9795
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.656798245614036e-06,
+      "loss": 0.4859,
+      "step": 9796
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.6562500000000005e-06,
+      "loss": 0.466,
+      "step": 9797
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.655701754385965e-06,
+      "loss": 0.2195,
+      "step": 9798
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.65515350877193e-06,
+      "loss": 0.5405,
+      "step": 9799
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.654605263157895e-06,
+      "loss": 0.251,
+      "step": 9800
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.65405701754386e-06,
+      "loss": 0.8848,
+      "step": 9801
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.653508771929825e-06,
+      "loss": 0.6816,
+      "step": 9802
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.65296052631579e-06,
+      "loss": 0.6044,
+      "step": 9803
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.652412280701754e-06,
+      "loss": 0.3548,
+      "step": 9804
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.65186403508772e-06,
+      "loss": 0.7647,
+      "step": 9805
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.651315789473684e-06,
+      "loss": 0.4306,
+      "step": 9806
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.65076754385965e-06,
+      "loss": 0.6988,
+      "step": 9807
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.6502192982456144e-06,
+      "loss": 0.4075,
+      "step": 9808
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.649671052631579e-06,
+      "loss": 1.0186,
+      "step": 9809
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.649122807017544e-06,
+      "loss": 1.0362,
+      "step": 9810
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.64857456140351e-06,
+      "loss": 0.7942,
+      "step": 9811
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.648026315789474e-06,
+      "loss": 0.2884,
+      "step": 9812
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.647478070175439e-06,
+      "loss": 0.787,
+      "step": 9813
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.646929824561403e-06,
+      "loss": 0.7525,
+      "step": 9814
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.646381578947369e-06,
+      "loss": 0.4511,
+      "step": 9815
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.645833333333334e-06,
+      "loss": 1.025,
+      "step": 9816
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.645285087719299e-06,
+      "loss": 0.5574,
+      "step": 9817
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.6447368421052635e-06,
+      "loss": 0.3143,
+      "step": 9818
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.644188596491228e-06,
+      "loss": 0.3386,
+      "step": 9819
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.643640350877193e-06,
+      "loss": 0.3198,
+      "step": 9820
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.643092105263158e-06,
+      "loss": 1.1931,
+      "step": 9821
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.642543859649124e-06,
+      "loss": 0.8446,
+      "step": 9822
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.641995614035088e-06,
+      "loss": 1.0972,
+      "step": 9823
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.641447368421053e-06,
+      "loss": 0.3264,
+      "step": 9824
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.640899122807017e-06,
+      "loss": 0.3324,
+      "step": 9825
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.640350877192983e-06,
+      "loss": 1.0037,
+      "step": 9826
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.639802631578948e-06,
+      "loss": 0.3687,
+      "step": 9827
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.639254385964913e-06,
+      "loss": 0.586,
+      "step": 9828
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.6387061403508775e-06,
+      "loss": 0.6843,
+      "step": 9829
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.638157894736843e-06,
+      "loss": 1.0746,
+      "step": 9830
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.637609649122807e-06,
+      "loss": 1.1537,
+      "step": 9831
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.637061403508773e-06,
+      "loss": 0.1067,
+      "step": 9832
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.636513157894737e-06,
+      "loss": 0.271,
+      "step": 9833
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.6359649122807024e-06,
+      "loss": 0.4164,
+      "step": 9834
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.635416666666667e-06,
+      "loss": 0.5233,
+      "step": 9835
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.634868421052632e-06,
+      "loss": 0.5432,
+      "step": 9836
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.634320175438597e-06,
+      "loss": 1.2601,
+      "step": 9837
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.633771929824562e-06,
+      "loss": 0.5837,
+      "step": 9838
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.6332236842105265e-06,
+      "loss": 0.4804,
+      "step": 9839
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.632675438596491e-06,
+      "loss": 0.9129,
+      "step": 9840
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.632127192982456e-06,
+      "loss": 1.1462,
+      "step": 9841
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.631578947368421e-06,
+      "loss": 0.7256,
+      "step": 9842
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.631030701754387e-06,
+      "loss": 0.5707,
+      "step": 9843
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.630482456140351e-06,
+      "loss": 0.8662,
+      "step": 9844
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.629934210526316e-06,
+      "loss": 0.5689,
+      "step": 9845
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.629385964912281e-06,
+      "loss": 0.951,
+      "step": 9846
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.628837719298246e-06,
+      "loss": 0.2442,
+      "step": 9847
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.628289473684211e-06,
+      "loss": 0.3723,
+      "step": 9848
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.627741228070176e-06,
+      "loss": 1.5913,
+      "step": 9849
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.6271929824561405e-06,
+      "loss": 0.2411,
+      "step": 9850
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.626644736842106e-06,
+      "loss": 0.1675,
+      "step": 9851
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.62609649122807e-06,
+      "loss": 0.4443,
+      "step": 9852
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.625548245614036e-06,
+      "loss": 0.3907,
+      "step": 9853
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.625000000000001e-06,
+      "loss": 1.4504,
+      "step": 9854
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.6244517543859654e-06,
+      "loss": 0.2566,
+      "step": 9855
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.62390350877193e-06,
+      "loss": 0.1744,
+      "step": 9856
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.623355263157895e-06,
+      "loss": 0.1024,
+      "step": 9857
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.62280701754386e-06,
+      "loss": 0.4336,
+      "step": 9858
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.622258771929825e-06,
+      "loss": 1.5125,
+      "step": 9859
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.6217105263157896e-06,
+      "loss": 0.2776,
+      "step": 9860
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.621162280701754e-06,
+      "loss": 0.3025,
+      "step": 9861
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.62061403508772e-06,
+      "loss": 0.5655,
+      "step": 9862
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.620065789473684e-06,
+      "loss": 0.5936,
+      "step": 9863
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.61951754385965e-06,
+      "loss": 0.5434,
+      "step": 9864
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.6189692982456145e-06,
+      "loss": 1.3215,
+      "step": 9865
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.618421052631579e-06,
+      "loss": 1.3113,
+      "step": 9866
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.617872807017544e-06,
+      "loss": 0.25,
+      "step": 9867
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.617324561403509e-06,
+      "loss": 1.4091,
+      "step": 9868
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.616776315789474e-06,
+      "loss": 1.2269,
+      "step": 9869
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.6162280701754395e-06,
+      "loss": 0.6091,
+      "step": 9870
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.6156798245614035e-06,
+      "loss": 0.328,
+      "step": 9871
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.615131578947369e-06,
+      "loss": 0.2938,
+      "step": 9872
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.614583333333334e-06,
+      "loss": 0.6708,
+      "step": 9873
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.614035087719299e-06,
+      "loss": 0.2722,
+      "step": 9874
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.613486842105264e-06,
+      "loss": 0.526,
+      "step": 9875
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.6129385964912284e-06,
+      "loss": 0.5898,
+      "step": 9876
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.612390350877193e-06,
+      "loss": 0.8048,
+      "step": 9877
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.611842105263158e-06,
+      "loss": 0.5212,
+      "step": 9878
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.611293859649123e-06,
+      "loss": 0.5512,
+      "step": 9879
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.610745614035088e-06,
+      "loss": 1.0194,
+      "step": 9880
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.610197368421053e-06,
+      "loss": 0.5701,
+      "step": 9881
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.609649122807017e-06,
+      "loss": 0.7137,
+      "step": 9882
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.609100877192983e-06,
+      "loss": 1.0833,
+      "step": 9883
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.608552631578948e-06,
+      "loss": 0.6796,
+      "step": 9884
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.608004385964913e-06,
+      "loss": 0.5797,
+      "step": 9885
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.6074561403508775e-06,
+      "loss": 0.7316,
+      "step": 9886
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.606907894736842e-06,
+      "loss": 0.3904,
+      "step": 9887
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.606359649122807e-06,
+      "loss": 0.6243,
+      "step": 9888
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.605811403508773e-06,
+      "loss": 0.4774,
+      "step": 9889
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.605263157894737e-06,
+      "loss": 0.7618,
+      "step": 9890
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.6047149122807025e-06,
+      "loss": 0.5851,
+      "step": 9891
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.6041666666666665e-06,
+      "loss": 0.5141,
+      "step": 9892
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.603618421052632e-06,
+      "loss": 0.4161,
+      "step": 9893
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.603070175438597e-06,
+      "loss": 0.1136,
+      "step": 9894
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.602521929824562e-06,
+      "loss": 1.048,
+      "step": 9895
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.601973684210527e-06,
+      "loss": 0.7655,
+      "step": 9896
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.6014254385964915e-06,
+      "loss": 0.1995,
+      "step": 9897
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.600877192982456e-06,
+      "loss": 0.6308,
+      "step": 9898
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.600328947368421e-06,
+      "loss": 0.6656,
+      "step": 9899
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.599780701754386e-06,
+      "loss": 1.7222,
+      "step": 9900
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.599232456140352e-06,
+      "loss": 0.317,
+      "step": 9901
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.5986842105263164e-06,
+      "loss": 0.8463,
+      "step": 9902
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.598135964912281e-06,
+      "loss": 0.1716,
+      "step": 9903
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.597587719298246e-06,
+      "loss": 1.0116,
+      "step": 9904
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.597039473684211e-06,
+      "loss": 0.4592,
+      "step": 9905
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.596491228070176e-06,
+      "loss": 1.1577,
+      "step": 9906
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.5959429824561406e-06,
+      "loss": 0.5884,
+      "step": 9907
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.595394736842106e-06,
+      "loss": 0.8445,
+      "step": 9908
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.59484649122807e-06,
+      "loss": 0.5254,
+      "step": 9909
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.594298245614036e-06,
+      "loss": 0.8342,
+      "step": 9910
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.59375e-06,
+      "loss": 0.8772,
+      "step": 9911
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.5932017543859655e-06,
+      "loss": 0.1975,
+      "step": 9912
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.59265350877193e-06,
+      "loss": 1.4447,
+      "step": 9913
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.592105263157895e-06,
+      "loss": 0.3941,
+      "step": 9914
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.59155701754386e-06,
+      "loss": 0.7374,
+      "step": 9915
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.591008771929825e-06,
+      "loss": 0.9829,
+      "step": 9916
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.59046052631579e-06,
+      "loss": 1.0638,
+      "step": 9917
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.5899122807017545e-06,
+      "loss": 0.5556,
+      "step": 9918
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.589364035087719e-06,
+      "loss": 0.4485,
+      "step": 9919
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.588815789473685e-06,
+      "loss": 0.4101,
+      "step": 9920
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.58826754385965e-06,
+      "loss": 0.4762,
+      "step": 9921
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.587719298245615e-06,
+      "loss": 0.4549,
+      "step": 9922
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.5871710526315794e-06,
+      "loss": 1.4606,
+      "step": 9923
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.586622807017544e-06,
+      "loss": 0.4215,
+      "step": 9924
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.586074561403509e-06,
+      "loss": 0.2579,
+      "step": 9925
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.585526315789474e-06,
+      "loss": 0.2117,
+      "step": 9926
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.584978070175439e-06,
+      "loss": 0.3304,
+      "step": 9927
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.5844298245614036e-06,
+      "loss": 0.9719,
+      "step": 9928
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.583881578947369e-06,
+      "loss": 0.6859,
+      "step": 9929
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.583333333333333e-06,
+      "loss": 0.6046,
+      "step": 9930
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.582785087719299e-06,
+      "loss": 0.8359,
+      "step": 9931
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.582236842105264e-06,
+      "loss": 0.2726,
+      "step": 9932
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.5816885964912285e-06,
+      "loss": 0.2577,
+      "step": 9933
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.581140350877193e-06,
+      "loss": 0.2273,
+      "step": 9934
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.580592105263158e-06,
+      "loss": 0.14,
+      "step": 9935
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.580043859649123e-06,
+      "loss": 0.7484,
+      "step": 9936
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.579495614035088e-06,
+      "loss": 0.398,
+      "step": 9937
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.578947368421053e-06,
+      "loss": 0.5089,
+      "step": 9938
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.578399122807018e-06,
+      "loss": 0.1993,
+      "step": 9939
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.577850877192983e-06,
+      "loss": 0.4897,
+      "step": 9940
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.577302631578948e-06,
+      "loss": 0.1758,
+      "step": 9941
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.576754385964913e-06,
+      "loss": 1.1178,
+      "step": 9942
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.576206140350878e-06,
+      "loss": 0.682,
+      "step": 9943
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.5756578947368425e-06,
+      "loss": 0.4436,
+      "step": 9944
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.575109649122807e-06,
+      "loss": 0.5713,
+      "step": 9945
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.574561403508772e-06,
+      "loss": 0.4375,
+      "step": 9946
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.574013157894737e-06,
+      "loss": 0.3947,
+      "step": 9947
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.573464912280703e-06,
+      "loss": 0.84,
+      "step": 9948
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.572916666666667e-06,
+      "loss": 0.3681,
+      "step": 9949
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.572368421052632e-06,
+      "loss": 0.3257,
+      "step": 9950
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.571820175438597e-06,
+      "loss": 0.1163,
+      "step": 9951
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.571271929824562e-06,
+      "loss": 0.1677,
+      "step": 9952
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.570723684210527e-06,
+      "loss": 0.9758,
+      "step": 9953
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.5701754385964915e-06,
+      "loss": 0.5449,
+      "step": 9954
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.569627192982456e-06,
+      "loss": 1.2314,
+      "step": 9955
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.569078947368421e-06,
+      "loss": 0.8158,
+      "step": 9956
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.568530701754386e-06,
+      "loss": 0.3797,
+      "step": 9957
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.567982456140352e-06,
+      "loss": 0.6984,
+      "step": 9958
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.5674342105263165e-06,
+      "loss": 1.0482,
+      "step": 9959
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.566885964912281e-06,
+      "loss": 0.7179,
+      "step": 9960
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.566337719298246e-06,
+      "loss": 0.3527,
+      "step": 9961
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.565789473684211e-06,
+      "loss": 0.8435,
+      "step": 9962
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.565241228070176e-06,
+      "loss": 0.8052,
+      "step": 9963
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.564692982456141e-06,
+      "loss": 1.0185,
+      "step": 9964
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.5641447368421055e-06,
+      "loss": 0.9948,
+      "step": 9965
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.56359649122807e-06,
+      "loss": 0.3652,
+      "step": 9966
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.563048245614036e-06,
+      "loss": 0.7415,
+      "step": 9967
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.5625e-06,
+      "loss": 0.7232,
+      "step": 9968
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.561951754385966e-06,
+      "loss": 0.9951,
+      "step": 9969
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.56140350877193e-06,
+      "loss": 0.3974,
+      "step": 9970
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.560855263157895e-06,
+      "loss": 0.8753,
+      "step": 9971
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.56030701754386e-06,
+      "loss": 0.6985,
+      "step": 9972
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.559758771929825e-06,
+      "loss": 0.6553,
+      "step": 9973
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.55921052631579e-06,
+      "loss": 0.9692,
+      "step": 9974
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.558662280701755e-06,
+      "loss": 0.2266,
+      "step": 9975
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.558114035087719e-06,
+      "loss": 0.4529,
+      "step": 9976
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.557565789473685e-06,
+      "loss": 0.6217,
+      "step": 9977
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.557017543859649e-06,
+      "loss": 0.142,
+      "step": 9978
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.556469298245615e-06,
+      "loss": 1.2603,
+      "step": 9979
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.5559210526315795e-06,
+      "loss": 1.1894,
+      "step": 9980
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.555372807017544e-06,
+      "loss": 0.4915,
+      "step": 9981
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.554824561403509e-06,
+      "loss": 1.0994,
+      "step": 9982
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.554276315789474e-06,
+      "loss": 0.2206,
+      "step": 9983
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.553728070175439e-06,
+      "loss": 0.1754,
+      "step": 9984
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.553179824561404e-06,
+      "loss": 0.8543,
+      "step": 9985
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.552631578947369e-06,
+      "loss": 0.5658,
+      "step": 9986
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.552083333333333e-06,
+      "loss": 1.0249,
+      "step": 9987
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.551535087719299e-06,
+      "loss": 0.6415,
+      "step": 9988
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.550986842105263e-06,
+      "loss": 0.392,
+      "step": 9989
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.550438596491229e-06,
+      "loss": 0.1854,
+      "step": 9990
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.5498903508771934e-06,
+      "loss": 0.8814,
+      "step": 9991
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.549342105263158e-06,
+      "loss": 1.6482,
+      "step": 9992
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.548793859649123e-06,
+      "loss": 1.0409,
+      "step": 9993
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.548245614035089e-06,
+      "loss": 0.9183,
+      "step": 9994
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.547697368421053e-06,
+      "loss": 0.2597,
+      "step": 9995
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.547149122807018e-06,
+      "loss": 0.461,
+      "step": 9996
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.546600877192982e-06,
+      "loss": 0.2121,
+      "step": 9997
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.546052631578948e-06,
+      "loss": 0.537,
+      "step": 9998
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.545504385964913e-06,
+      "loss": 0.5976,
+      "step": 9999
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.544956140350878e-06,
+      "loss": 0.5636,
+      "step": 10000
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.5444078947368425e-06,
+      "loss": 0.8642,
+      "step": 10001
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.543859649122807e-06,
+      "loss": 0.1922,
+      "step": 10002
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.543311403508772e-06,
+      "loss": 0.9846,
+      "step": 10003
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.542763157894737e-06,
+      "loss": 0.9762,
+      "step": 10004
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.542214912280702e-06,
+      "loss": 0.464,
+      "step": 10005
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.541666666666667e-06,
+      "loss": 0.5205,
+      "step": 10006
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.541118421052632e-06,
+      "loss": 0.6628,
+      "step": 10007
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.540570175438596e-06,
+      "loss": 0.2171,
+      "step": 10008
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.540021929824562e-06,
+      "loss": 0.5547,
+      "step": 10009
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.539473684210527e-06,
+      "loss": 0.0427,
+      "step": 10010
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.538925438596492e-06,
+      "loss": 1.6344,
+      "step": 10011
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.5383771929824565e-06,
+      "loss": 0.2977,
+      "step": 10012
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.537828947368421e-06,
+      "loss": 1.2381,
+      "step": 10013
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.537280701754386e-06,
+      "loss": 0.339,
+      "step": 10014
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.536732456140352e-06,
+      "loss": 0.1898,
+      "step": 10015
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.536184210526316e-06,
+      "loss": 0.8223,
+      "step": 10016
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.5356359649122814e-06,
+      "loss": 0.4899,
+      "step": 10017
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.535087719298246e-06,
+      "loss": 0.338,
+      "step": 10018
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.534539473684211e-06,
+      "loss": 0.5844,
+      "step": 10019
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.533991228070176e-06,
+      "loss": 0.3566,
+      "step": 10020
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.533442982456141e-06,
+      "loss": 0.5371,
+      "step": 10021
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.5328947368421056e-06,
+      "loss": 1.0665,
+      "step": 10022
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.53234649122807e-06,
+      "loss": 0.1384,
+      "step": 10023
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.531798245614035e-06,
+      "loss": 0.1247,
+      "step": 10024
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.53125e-06,
+      "loss": 0.5636,
+      "step": 10025
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.530701754385966e-06,
+      "loss": 0.688,
+      "step": 10026
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.53015350877193e-06,
+      "loss": 0.8487,
+      "step": 10027
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.529605263157895e-06,
+      "loss": 1.0131,
+      "step": 10028
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.52905701754386e-06,
+      "loss": 0.391,
+      "step": 10029
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.528508771929825e-06,
+      "loss": 0.7143,
+      "step": 10030
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.52796052631579e-06,
+      "loss": 0.4216,
+      "step": 10031
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.527412280701755e-06,
+      "loss": 0.313,
+      "step": 10032
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.5268640350877195e-06,
+      "loss": 0.995,
+      "step": 10033
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.526315789473685e-06,
+      "loss": 0.1207,
+      "step": 10034
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.525767543859649e-06,
+      "loss": 0.2687,
+      "step": 10035
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.525219298245615e-06,
+      "loss": 0.3766,
+      "step": 10036
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.52467105263158e-06,
+      "loss": 0.4537,
+      "step": 10037
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.5241228070175444e-06,
+      "loss": 1.0986,
+      "step": 10038
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.523574561403509e-06,
+      "loss": 0.3159,
+      "step": 10039
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.523026315789474e-06,
+      "loss": 0.6027,
+      "step": 10040
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.522478070175439e-06,
+      "loss": 0.7537,
+      "step": 10041
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.521929824561404e-06,
+      "loss": 0.7616,
+      "step": 10042
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.5213815789473686e-06,
+      "loss": 0.9027,
+      "step": 10043
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.520833333333333e-06,
+      "loss": 0.5962,
+      "step": 10044
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.520285087719299e-06,
+      "loss": 0.6257,
+      "step": 10045
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.519736842105263e-06,
+      "loss": 0.467,
+      "step": 10046
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.519188596491229e-06,
+      "loss": 0.5613,
+      "step": 10047
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.518640350877193e-06,
+      "loss": 0.7951,
+      "step": 10048
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.518092105263158e-06,
+      "loss": 1.2704,
+      "step": 10049
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.517543859649123e-06,
+      "loss": 0.4022,
+      "step": 10050
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.516995614035088e-06,
+      "loss": 0.6111,
+      "step": 10051
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.516447368421053e-06,
+      "loss": 0.8033,
+      "step": 10052
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.5158991228070185e-06,
+      "loss": 0.0883,
+      "step": 10053
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.5153508771929825e-06,
+      "loss": 0.4498,
+      "step": 10054
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.514802631578948e-06,
+      "loss": 0.6395,
+      "step": 10055
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.514254385964912e-06,
+      "loss": 0.9197,
+      "step": 10056
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.513706140350878e-06,
+      "loss": 0.1665,
+      "step": 10057
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.513157894736843e-06,
+      "loss": 1.6139,
+      "step": 10058
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.5126096491228075e-06,
+      "loss": 0.1765,
+      "step": 10059
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.512061403508772e-06,
+      "loss": 0.538,
+      "step": 10060
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.511513157894737e-06,
+      "loss": 0.4553,
+      "step": 10061
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.510964912280702e-06,
+      "loss": 0.9733,
+      "step": 10062
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.510416666666667e-06,
+      "loss": 0.8996,
+      "step": 10063
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.5098684210526316e-06,
+      "loss": 0.6336,
+      "step": 10064
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.509320175438596e-06,
+      "loss": 0.8751,
+      "step": 10065
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.508771929824562e-06,
+      "loss": 1.0009,
+      "step": 10066
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.508223684210527e-06,
+      "loss": 0.7193,
+      "step": 10067
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.507675438596492e-06,
+      "loss": 1.2861,
+      "step": 10068
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.5071271929824565e-06,
+      "loss": 0.4049,
+      "step": 10069
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.506578947368421e-06,
+      "loss": 1.7538,
+      "step": 10070
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.506030701754386e-06,
+      "loss": 0.2995,
+      "step": 10071
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.505482456140352e-06,
+      "loss": 0.6961,
+      "step": 10072
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.504934210526316e-06,
+      "loss": 0.1319,
+      "step": 10073
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.5043859649122815e-06,
+      "loss": 0.6835,
+      "step": 10074
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.5038377192982455e-06,
+      "loss": 0.5254,
+      "step": 10075
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.503289473684211e-06,
+      "loss": 0.446,
+      "step": 10076
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.502741228070176e-06,
+      "loss": 1.2667,
+      "step": 10077
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.502192982456141e-06,
+      "loss": 0.9121,
+      "step": 10078
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.501644736842106e-06,
+      "loss": 1.1151,
+      "step": 10079
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.5010964912280705e-06,
+      "loss": 0.4066,
+      "step": 10080
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.500548245614035e-06,
+      "loss": 1.1189,
+      "step": 10081
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.5e-06,
+      "loss": 0.3341,
+      "step": 10082
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.499451754385965e-06,
+      "loss": 0.2683,
+      "step": 10083
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.49890350877193e-06,
+      "loss": 0.4112,
+      "step": 10084
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.4983552631578954e-06,
+      "loss": 1.0607,
+      "step": 10085
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.49780701754386e-06,
+      "loss": 0.3109,
+      "step": 10086
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.497258771929825e-06,
+      "loss": 0.4637,
+      "step": 10087
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.49671052631579e-06,
+      "loss": 0.578,
+      "step": 10088
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.496162280701755e-06,
+      "loss": 0.5955,
+      "step": 10089
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.4956140350877196e-06,
+      "loss": 0.1044,
+      "step": 10090
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.495065789473684e-06,
+      "loss": 0.2502,
+      "step": 10091
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.494517543859649e-06,
+      "loss": 0.8985,
+      "step": 10092
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.493969298245615e-06,
+      "loss": 0.7871,
+      "step": 10093
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.493421052631579e-06,
+      "loss": 0.4529,
+      "step": 10094
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.4928728070175445e-06,
+      "loss": 0.7826,
+      "step": 10095
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.492324561403509e-06,
+      "loss": 0.5815,
+      "step": 10096
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.491776315789474e-06,
+      "loss": 1.0706,
+      "step": 10097
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.491228070175439e-06,
+      "loss": 0.6738,
+      "step": 10098
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.490679824561404e-06,
+      "loss": 0.683,
+      "step": 10099
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.490131578947369e-06,
+      "loss": 0.4774,
+      "step": 10100
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.4895833333333335e-06,
+      "loss": 0.358,
+      "step": 10101
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.489035087719298e-06,
+      "loss": 0.2395,
+      "step": 10102
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.488486842105263e-06,
+      "loss": 1.0586,
+      "step": 10103
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.487938596491229e-06,
+      "loss": 1.0641,
+      "step": 10104
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.487390350877194e-06,
+      "loss": 0.5698,
+      "step": 10105
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.4868421052631584e-06,
+      "loss": 0.3796,
+      "step": 10106
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.486293859649123e-06,
+      "loss": 1.5373,
+      "step": 10107
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.485745614035088e-06,
+      "loss": 0.7206,
+      "step": 10108
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.485197368421053e-06,
+      "loss": 0.1112,
+      "step": 10109
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.484649122807018e-06,
+      "loss": 1.3465,
+      "step": 10110
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.4841008771929826e-06,
+      "loss": 0.4408,
+      "step": 10111
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.483552631578948e-06,
+      "loss": 1.1117,
+      "step": 10112
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.483004385964912e-06,
+      "loss": 1.0188,
+      "step": 10113
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.482456140350878e-06,
+      "loss": 0.4914,
+      "step": 10114
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.481907894736843e-06,
+      "loss": 1.0301,
+      "step": 10115
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.4813596491228075e-06,
+      "loss": 0.8444,
+      "step": 10116
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.480811403508772e-06,
+      "loss": 0.9779,
+      "step": 10117
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.480263157894737e-06,
+      "loss": 1.3395,
+      "step": 10118
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.479714912280702e-06,
+      "loss": 0.2227,
+      "step": 10119
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.479166666666667e-06,
+      "loss": 0.9542,
+      "step": 10120
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.478618421052632e-06,
+      "loss": 0.9657,
+      "step": 10121
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.4780701754385965e-06,
+      "loss": 0.283,
+      "step": 10122
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.477521929824562e-06,
+      "loss": 1.1386,
+      "step": 10123
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.476973684210527e-06,
+      "loss": 0.4885,
+      "step": 10124
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.476425438596492e-06,
+      "loss": 0.3952,
+      "step": 10125
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.475877192982457e-06,
+      "loss": 1.0697,
+      "step": 10126
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.4753289473684215e-06,
+      "loss": 0.6119,
+      "step": 10127
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.474780701754386e-06,
+      "loss": 0.5134,
+      "step": 10128
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.474232456140351e-06,
+      "loss": 0.6406,
+      "step": 10129
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.473684210526316e-06,
+      "loss": 0.3779,
+      "step": 10130
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.473135964912282e-06,
+      "loss": 1.2447,
+      "step": 10131
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.472587719298246e-06,
+      "loss": 1.2306,
+      "step": 10132
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.472039473684211e-06,
+      "loss": 1.4921,
+      "step": 10133
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.471491228070175e-06,
+      "loss": 0.6468,
+      "step": 10134
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.470942982456141e-06,
+      "loss": 0.3775,
+      "step": 10135
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.470394736842106e-06,
+      "loss": 1.0028,
+      "step": 10136
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.4698464912280706e-06,
+      "loss": 0.4581,
+      "step": 10137
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.469298245614035e-06,
+      "loss": 0.4942,
+      "step": 10138
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.46875e-06,
+      "loss": 0.8134,
+      "step": 10139
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.468201754385965e-06,
+      "loss": 0.4259,
+      "step": 10140
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.46765350877193e-06,
+      "loss": 0.4165,
+      "step": 10141
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.467105263157895e-06,
+      "loss": 0.7515,
+      "step": 10142
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.46655701754386e-06,
+      "loss": 0.2478,
+      "step": 10143
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.466008771929825e-06,
+      "loss": 0.864,
+      "step": 10144
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.46546052631579e-06,
+      "loss": 1.1844,
+      "step": 10145
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.464912280701755e-06,
+      "loss": 0.2775,
+      "step": 10146
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.46436403508772e-06,
+      "loss": 0.3968,
+      "step": 10147
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.4638157894736845e-06,
+      "loss": 0.3646,
+      "step": 10148
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.463267543859649e-06,
+      "loss": 0.2002,
+      "step": 10149
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.462719298245615e-06,
+      "loss": 0.6822,
+      "step": 10150
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.462171052631579e-06,
+      "loss": 0.8164,
+      "step": 10151
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.461622807017545e-06,
+      "loss": 0.8212,
+      "step": 10152
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.461074561403509e-06,
+      "loss": 0.4705,
+      "step": 10153
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.460526315789474e-06,
+      "loss": 0.5596,
+      "step": 10154
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.459978070175439e-06,
+      "loss": 1.1883,
+      "step": 10155
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.459429824561404e-06,
+      "loss": 0.8795,
+      "step": 10156
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.458881578947369e-06,
+      "loss": 0.2453,
+      "step": 10157
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.4583333333333336e-06,
+      "loss": 0.7618,
+      "step": 10158
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.457785087719298e-06,
+      "loss": 1.1785,
+      "step": 10159
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.457236842105264e-06,
+      "loss": 0.6706,
+      "step": 10160
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.456688596491228e-06,
+      "loss": 0.5387,
+      "step": 10161
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.456140350877194e-06,
+      "loss": 0.965,
+      "step": 10162
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.4555921052631585e-06,
+      "loss": 0.8256,
+      "step": 10163
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.455043859649123e-06,
+      "loss": 0.1756,
+      "step": 10164
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.454495614035088e-06,
+      "loss": 1.023,
+      "step": 10165
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.453947368421053e-06,
+      "loss": 0.6576,
+      "step": 10166
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.453399122807018e-06,
+      "loss": 0.4487,
+      "step": 10167
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.452850877192983e-06,
+      "loss": 0.3245,
+      "step": 10168
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.4523026315789475e-06,
+      "loss": 0.1558,
+      "step": 10169
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.451754385964912e-06,
+      "loss": 0.5285,
+      "step": 10170
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.451206140350878e-06,
+      "loss": 0.6049,
+      "step": 10171
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.450657894736842e-06,
+      "loss": 0.3896,
+      "step": 10172
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.450109649122808e-06,
+      "loss": 0.6411,
+      "step": 10173
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.4495614035087725e-06,
+      "loss": 0.4191,
+      "step": 10174
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.449013157894737e-06,
+      "loss": 0.9649,
+      "step": 10175
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.448464912280702e-06,
+      "loss": 1.2963,
+      "step": 10176
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.447916666666667e-06,
+      "loss": 0.592,
+      "step": 10177
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.447368421052632e-06,
+      "loss": 0.3465,
+      "step": 10178
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.446820175438597e-06,
+      "loss": 0.2924,
+      "step": 10179
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.446271929824561e-06,
+      "loss": 0.8631,
+      "step": 10180
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.445723684210527e-06,
+      "loss": 0.7514,
+      "step": 10181
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.445175438596492e-06,
+      "loss": 0.3414,
+      "step": 10182
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.444627192982457e-06,
+      "loss": 0.1847,
+      "step": 10183
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.4440789473684215e-06,
+      "loss": 0.8347,
+      "step": 10184
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.443530701754386e-06,
+      "loss": 0.7461,
+      "step": 10185
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.442982456140351e-06,
+      "loss": 0.3836,
+      "step": 10186
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.442434210526316e-06,
+      "loss": 0.4174,
+      "step": 10187
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.441885964912281e-06,
+      "loss": 0.7443,
+      "step": 10188
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.441337719298246e-06,
+      "loss": 0.2385,
+      "step": 10189
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.440789473684211e-06,
+      "loss": 0.4566,
+      "step": 10190
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.440241228070175e-06,
+      "loss": 0.4007,
+      "step": 10191
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.439692982456141e-06,
+      "loss": 1.1095,
+      "step": 10192
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.439144736842106e-06,
+      "loss": 0.4288,
+      "step": 10193
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.438596491228071e-06,
+      "loss": 0.7923,
+      "step": 10194
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.4380482456140355e-06,
+      "loss": 0.6315,
+      "step": 10195
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.4375e-06,
+      "loss": 0.968,
+      "step": 10196
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.436951754385965e-06,
+      "loss": 0.9049,
+      "step": 10197
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.436403508771931e-06,
+      "loss": 0.4755,
+      "step": 10198
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.435855263157895e-06,
+      "loss": 0.5241,
+      "step": 10199
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.4353070175438604e-06,
+      "loss": 0.3532,
+      "step": 10200
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.434758771929825e-06,
+      "loss": 0.1945,
+      "step": 10201
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.43421052631579e-06,
+      "loss": 0.7061,
+      "step": 10202
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.433662280701755e-06,
+      "loss": 0.5316,
+      "step": 10203
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.43311403508772e-06,
+      "loss": 0.2212,
+      "step": 10204
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.4325657894736846e-06,
+      "loss": 0.7848,
+      "step": 10205
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.432017543859649e-06,
+      "loss": 0.302,
+      "step": 10206
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.431469298245614e-06,
+      "loss": 0.2284,
+      "step": 10207
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.430921052631579e-06,
+      "loss": 1.0573,
+      "step": 10208
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.430372807017545e-06,
+      "loss": 0.5777,
+      "step": 10209
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.429824561403509e-06,
+      "loss": 1.2535,
+      "step": 10210
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.429276315789474e-06,
+      "loss": 0.7903,
+      "step": 10211
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.428728070175438e-06,
+      "loss": 0.1859,
+      "step": 10212
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.428179824561404e-06,
+      "loss": 0.268,
+      "step": 10213
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.427631578947369e-06,
+      "loss": 1.0018,
+      "step": 10214
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.427083333333334e-06,
+      "loss": 0.5247,
+      "step": 10215
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.4265350877192985e-06,
+      "loss": 0.3149,
+      "step": 10216
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.425986842105264e-06,
+      "loss": 0.7488,
+      "step": 10217
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.425438596491228e-06,
+      "loss": 0.1966,
+      "step": 10218
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.424890350877194e-06,
+      "loss": 0.4884,
+      "step": 10219
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.424342105263158e-06,
+      "loss": 0.4689,
+      "step": 10220
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.4237938596491234e-06,
+      "loss": 0.4235,
+      "step": 10221
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.423245614035088e-06,
+      "loss": 0.7481,
+      "step": 10222
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.422697368421053e-06,
+      "loss": 0.8358,
+      "step": 10223
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.422149122807018e-06,
+      "loss": 0.273,
+      "step": 10224
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.421600877192983e-06,
+      "loss": 0.7952,
+      "step": 10225
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.4210526315789476e-06,
+      "loss": 0.3908,
+      "step": 10226
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.420504385964912e-06,
+      "loss": 0.3753,
+      "step": 10227
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.419956140350877e-06,
+      "loss": 0.2937,
+      "step": 10228
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.419407894736842e-06,
+      "loss": 0.0841,
+      "step": 10229
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.418859649122808e-06,
+      "loss": 0.7445,
+      "step": 10230
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.418311403508772e-06,
+      "loss": 0.2018,
+      "step": 10231
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.417763157894737e-06,
+      "loss": 0.8803,
+      "step": 10232
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.417214912280702e-06,
+      "loss": 0.2983,
+      "step": 10233
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.416666666666667e-06,
+      "loss": 0.743,
+      "step": 10234
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.416118421052632e-06,
+      "loss": 0.7663,
+      "step": 10235
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.4155701754385975e-06,
+      "loss": 0.5684,
+      "step": 10236
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.4150219298245615e-06,
+      "loss": 0.9399,
+      "step": 10237
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.414473684210527e-06,
+      "loss": 0.549,
+      "step": 10238
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.413925438596491e-06,
+      "loss": 0.4669,
+      "step": 10239
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.413377192982457e-06,
+      "loss": 0.4514,
+      "step": 10240
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.412828947368422e-06,
+      "loss": 0.9515,
+      "step": 10241
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.4122807017543865e-06,
+      "loss": 0.395,
+      "step": 10242
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.411732456140351e-06,
+      "loss": 0.6046,
+      "step": 10243
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.411184210526316e-06,
+      "loss": 0.6569,
+      "step": 10244
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.410635964912281e-06,
+      "loss": 0.6796,
+      "step": 10245
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.410087719298246e-06,
+      "loss": 0.2387,
+      "step": 10246
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.409539473684211e-06,
+      "loss": 0.9095,
+      "step": 10247
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.408991228070175e-06,
+      "loss": 0.3617,
+      "step": 10248
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.408442982456141e-06,
+      "loss": 0.3808,
+      "step": 10249
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.407894736842105e-06,
+      "loss": 0.6367,
+      "step": 10250
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.407346491228071e-06,
+      "loss": 0.6826,
+      "step": 10251
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.4067982456140356e-06,
+      "loss": 0.2602,
+      "step": 10252
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.40625e-06,
+      "loss": 1.1161,
+      "step": 10253
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.405701754385965e-06,
+      "loss": 0.4725,
+      "step": 10254
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.40515350877193e-06,
+      "loss": 1.0255,
+      "step": 10255
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.404605263157895e-06,
+      "loss": 0.1418,
+      "step": 10256
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.4040570175438605e-06,
+      "loss": 0.3491,
+      "step": 10257
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.4035087719298245e-06,
+      "loss": 0.7609,
+      "step": 10258
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.40296052631579e-06,
+      "loss": 0.4359,
+      "step": 10259
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.402412280701755e-06,
+      "loss": 0.2391,
+      "step": 10260
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.40186403508772e-06,
+      "loss": 0.291,
+      "step": 10261
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.401315789473685e-06,
+      "loss": 0.748,
+      "step": 10262
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.4007675438596495e-06,
+      "loss": 0.2637,
+      "step": 10263
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.400219298245614e-06,
+      "loss": 0.6809,
+      "step": 10264
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.399671052631579e-06,
+      "loss": 0.6981,
+      "step": 10265
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.399122807017544e-06,
+      "loss": 0.9592,
+      "step": 10266
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.398574561403509e-06,
+      "loss": 0.5858,
+      "step": 10267
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.3980263157894744e-06,
+      "loss": 0.8642,
+      "step": 10268
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.397478070175438e-06,
+      "loss": 0.9488,
+      "step": 10269
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.396929824561404e-06,
+      "loss": 0.6003,
+      "step": 10270
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.396381578947369e-06,
+      "loss": 0.1813,
+      "step": 10271
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.395833333333334e-06,
+      "loss": 0.074,
+      "step": 10272
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.3952850877192986e-06,
+      "loss": 0.6609,
+      "step": 10273
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.394736842105263e-06,
+      "loss": 0.8784,
+      "step": 10274
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.394188596491228e-06,
+      "loss": 0.2214,
+      "step": 10275
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.393640350877194e-06,
+      "loss": 0.6125,
+      "step": 10276
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.393092105263158e-06,
+      "loss": 0.4375,
+      "step": 10277
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.3925438596491235e-06,
+      "loss": 1.4458,
+      "step": 10278
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.391995614035088e-06,
+      "loss": 0.3883,
+      "step": 10279
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.391447368421053e-06,
+      "loss": 0.8261,
+      "step": 10280
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.390899122807018e-06,
+      "loss": 0.8873,
+      "step": 10281
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.390350877192983e-06,
+      "loss": 0.2405,
+      "step": 10282
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.389802631578948e-06,
+      "loss": 0.8979,
+      "step": 10283
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.3892543859649125e-06,
+      "loss": 0.1944,
+      "step": 10284
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.388706140350877e-06,
+      "loss": 0.4951,
+      "step": 10285
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.388157894736842e-06,
+      "loss": 0.3433,
+      "step": 10286
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.387609649122808e-06,
+      "loss": 0.1653,
+      "step": 10287
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.387061403508772e-06,
+      "loss": 0.3589,
+      "step": 10288
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.3865131578947375e-06,
+      "loss": 0.7172,
+      "step": 10289
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.385964912280702e-06,
+      "loss": 0.288,
+      "step": 10290
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.385416666666667e-06,
+      "loss": 0.8006,
+      "step": 10291
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.384868421052632e-06,
+      "loss": 0.3338,
+      "step": 10292
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.384320175438597e-06,
+      "loss": 0.5908,
+      "step": 10293
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.3837719298245616e-06,
+      "loss": 0.4166,
+      "step": 10294
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.383223684210527e-06,
+      "loss": 0.176,
+      "step": 10295
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.382675438596491e-06,
+      "loss": 0.3924,
+      "step": 10296
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.382127192982457e-06,
+      "loss": 0.3077,
+      "step": 10297
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.381578947368421e-06,
+      "loss": 0.4904,
+      "step": 10298
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.3810307017543865e-06,
+      "loss": 0.4831,
+      "step": 10299
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.380482456140351e-06,
+      "loss": 0.8753,
+      "step": 10300
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.379934210526316e-06,
+      "loss": 0.8451,
+      "step": 10301
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.379385964912281e-06,
+      "loss": 0.5304,
+      "step": 10302
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.378837719298246e-06,
+      "loss": 0.827,
+      "step": 10303
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.378289473684211e-06,
+      "loss": 0.5301,
+      "step": 10304
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.3777412280701755e-06,
+      "loss": 0.2255,
+      "step": 10305
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.37719298245614e-06,
+      "loss": 0.36,
+      "step": 10306
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.376644736842105e-06,
+      "loss": 0.2217,
+      "step": 10307
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.376096491228071e-06,
+      "loss": 0.5069,
+      "step": 10308
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.375548245614036e-06,
+      "loss": 1.1389,
+      "step": 10309
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.3750000000000005e-06,
+      "loss": 0.6776,
+      "step": 10310
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.374451754385965e-06,
+      "loss": 0.4159,
+      "step": 10311
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.37390350877193e-06,
+      "loss": 0.2379,
+      "step": 10312
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.373355263157895e-06,
+      "loss": 0.7093,
+      "step": 10313
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.372807017543861e-06,
+      "loss": 0.3416,
+      "step": 10314
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.372258771929825e-06,
+      "loss": 0.3352,
+      "step": 10315
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.37171052631579e-06,
+      "loss": 0.3318,
+      "step": 10316
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.371162280701754e-06,
+      "loss": 0.8237,
+      "step": 10317
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.37061403508772e-06,
+      "loss": 0.779,
+      "step": 10318
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.370065789473685e-06,
+      "loss": 0.8253,
+      "step": 10319
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.3695175438596496e-06,
+      "loss": 0.446,
+      "step": 10320
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.368969298245614e-06,
+      "loss": 1.4156,
+      "step": 10321
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.368421052631579e-06,
+      "loss": 0.5226,
+      "step": 10322
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.367872807017544e-06,
+      "loss": 1.1118,
+      "step": 10323
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.367324561403509e-06,
+      "loss": 1.1689,
+      "step": 10324
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.366776315789474e-06,
+      "loss": 0.3628,
+      "step": 10325
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.3662280701754385e-06,
+      "loss": 0.8345,
+      "step": 10326
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.365679824561404e-06,
+      "loss": 0.0878,
+      "step": 10327
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.365131578947369e-06,
+      "loss": 1.2873,
+      "step": 10328
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.364583333333334e-06,
+      "loss": 0.8346,
+      "step": 10329
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.364035087719299e-06,
+      "loss": 0.6782,
+      "step": 10330
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.3634868421052635e-06,
+      "loss": 0.534,
+      "step": 10331
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.362938596491228e-06,
+      "loss": 0.2941,
+      "step": 10332
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.362390350877193e-06,
+      "loss": 0.2385,
+      "step": 10333
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.361842105263158e-06,
+      "loss": 1.1001,
+      "step": 10334
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.361293859649124e-06,
+      "loss": 0.1964,
+      "step": 10335
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.360745614035088e-06,
+      "loss": 0.1702,
+      "step": 10336
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.360197368421053e-06,
+      "loss": 0.2424,
+      "step": 10337
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.359649122807018e-06,
+      "loss": 1.1866,
+      "step": 10338
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.359100877192983e-06,
+      "loss": 1.3463,
+      "step": 10339
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.358552631578948e-06,
+      "loss": 0.6668,
+      "step": 10340
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.3580043859649126e-06,
+      "loss": 0.5853,
+      "step": 10341
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.357456140350877e-06,
+      "loss": 0.4182,
+      "step": 10342
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.356907894736842e-06,
+      "loss": 0.1463,
+      "step": 10343
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.356359649122807e-06,
+      "loss": 0.3444,
+      "step": 10344
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.355811403508773e-06,
+      "loss": 0.4008,
+      "step": 10345
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.3552631578947375e-06,
+      "loss": 0.165,
+      "step": 10346
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.354714912280702e-06,
+      "loss": 0.6638,
+      "step": 10347
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.354166666666667e-06,
+      "loss": 0.5304,
+      "step": 10348
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.353618421052632e-06,
+      "loss": 0.7221,
+      "step": 10349
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.353070175438597e-06,
+      "loss": 0.696,
+      "step": 10350
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.352521929824562e-06,
+      "loss": 1.3852,
+      "step": 10351
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.3519736842105265e-06,
+      "loss": 0.7538,
+      "step": 10352
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.351425438596491e-06,
+      "loss": 0.4621,
+      "step": 10353
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.350877192982457e-06,
+      "loss": 0.6406,
+      "step": 10354
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.350328947368421e-06,
+      "loss": 0.1984,
+      "step": 10355
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.349780701754387e-06,
+      "loss": 0.2079,
+      "step": 10356
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.3492324561403515e-06,
+      "loss": 0.1881,
+      "step": 10357
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.348684210526316e-06,
+      "loss": 0.5895,
+      "step": 10358
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.348135964912281e-06,
+      "loss": 0.4601,
+      "step": 10359
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.347587719298246e-06,
+      "loss": 0.9012,
+      "step": 10360
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.347039473684211e-06,
+      "loss": 0.3008,
+      "step": 10361
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.346491228070176e-06,
+      "loss": 0.3717,
+      "step": 10362
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.34594298245614e-06,
+      "loss": 0.3182,
+      "step": 10363
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.345394736842106e-06,
+      "loss": 0.5913,
+      "step": 10364
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.344846491228071e-06,
+      "loss": 1.1482,
+      "step": 10365
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.344298245614036e-06,
+      "loss": 0.7032,
+      "step": 10366
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.3437500000000006e-06,
+      "loss": 0.5295,
+      "step": 10367
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.343201754385965e-06,
+      "loss": 0.7609,
+      "step": 10368
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.34265350877193e-06,
+      "loss": 0.656,
+      "step": 10369
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.342105263157895e-06,
+      "loss": 0.4732,
+      "step": 10370
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.34155701754386e-06,
+      "loss": 0.5105,
+      "step": 10371
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.341008771929825e-06,
+      "loss": 0.5197,
+      "step": 10372
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.34046052631579e-06,
+      "loss": 0.4824,
+      "step": 10373
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.339912280701754e-06,
+      "loss": 1.4004,
+      "step": 10374
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.33936403508772e-06,
+      "loss": 0.6585,
+      "step": 10375
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.338815789473684e-06,
+      "loss": 1.3693,
+      "step": 10376
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.33826754385965e-06,
+      "loss": 0.5401,
+      "step": 10377
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.3377192982456145e-06,
+      "loss": 1.2997,
+      "step": 10378
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.337171052631579e-06,
+      "loss": 0.1978,
+      "step": 10379
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.336622807017544e-06,
+      "loss": 0.3867,
+      "step": 10380
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.336074561403509e-06,
+      "loss": 0.7097,
+      "step": 10381
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.335526315789474e-06,
+      "loss": 0.7347,
+      "step": 10382
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.3349780701754394e-06,
+      "loss": 0.4583,
+      "step": 10383
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.334429824561403e-06,
+      "loss": 0.3985,
+      "step": 10384
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.333881578947369e-06,
+      "loss": 0.0882,
+      "step": 10385
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.333333333333334e-06,
+      "loss": 0.4662,
+      "step": 10386
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.332785087719299e-06,
+      "loss": 0.6526,
+      "step": 10387
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.3322368421052636e-06,
+      "loss": 0.3483,
+      "step": 10388
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.331688596491228e-06,
+      "loss": 0.5303,
+      "step": 10389
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.331140350877193e-06,
+      "loss": 0.9759,
+      "step": 10390
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.330592105263158e-06,
+      "loss": 0.9951,
+      "step": 10391
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.330043859649124e-06,
+      "loss": 1.3856,
+      "step": 10392
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.329495614035088e-06,
+      "loss": 0.2685,
+      "step": 10393
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.328947368421053e-06,
+      "loss": 0.4014,
+      "step": 10394
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.328399122807017e-06,
+      "loss": 1.545,
+      "step": 10395
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.327850877192983e-06,
+      "loss": 0.0619,
+      "step": 10396
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.327302631578948e-06,
+      "loss": 0.5008,
+      "step": 10397
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.326754385964913e-06,
+      "loss": 1.2727,
+      "step": 10398
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.3262061403508775e-06,
+      "loss": 0.1729,
+      "step": 10399
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.325657894736842e-06,
+      "loss": 0.4029,
+      "step": 10400
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.325109649122807e-06,
+      "loss": 0.4111,
+      "step": 10401
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.324561403508773e-06,
+      "loss": 0.4417,
+      "step": 10402
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.324013157894737e-06,
+      "loss": 0.6772,
+      "step": 10403
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.3234649122807025e-06,
+      "loss": 0.2814,
+      "step": 10404
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.322916666666667e-06,
+      "loss": 0.3725,
+      "step": 10405
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.322368421052632e-06,
+      "loss": 0.2444,
+      "step": 10406
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.321820175438597e-06,
+      "loss": 0.3191,
+      "step": 10407
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.321271929824562e-06,
+      "loss": 0.4704,
+      "step": 10408
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.3207236842105266e-06,
+      "loss": 1.0595,
+      "step": 10409
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.320175438596491e-06,
+      "loss": 0.7241,
+      "step": 10410
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.319627192982456e-06,
+      "loss": 1.3995,
+      "step": 10411
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.319078947368421e-06,
+      "loss": 0.5893,
+      "step": 10412
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.318530701754387e-06,
+      "loss": 0.2907,
+      "step": 10413
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.317982456140351e-06,
+      "loss": 0.1674,
+      "step": 10414
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.317434210526316e-06,
+      "loss": 1.0025,
+      "step": 10415
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.316885964912281e-06,
+      "loss": 1.0307,
+      "step": 10416
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.316337719298246e-06,
+      "loss": 0.76,
+      "step": 10417
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.315789473684211e-06,
+      "loss": 0.3646,
+      "step": 10418
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.315241228070176e-06,
+      "loss": 0.1084,
+      "step": 10419
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.3146929824561405e-06,
+      "loss": 0.3847,
+      "step": 10420
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.314144736842106e-06,
+      "loss": 0.9039,
+      "step": 10421
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.31359649122807e-06,
+      "loss": 0.4045,
+      "step": 10422
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.313048245614036e-06,
+      "loss": 1.0263,
+      "step": 10423
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.312500000000001e-06,
+      "loss": 0.3672,
+      "step": 10424
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.3119517543859655e-06,
+      "loss": 0.2955,
+      "step": 10425
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.31140350877193e-06,
+      "loss": 0.3079,
+      "step": 10426
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.310855263157895e-06,
+      "loss": 0.113,
+      "step": 10427
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.31030701754386e-06,
+      "loss": 0.7615,
+      "step": 10428
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.309758771929825e-06,
+      "loss": 0.6233,
+      "step": 10429
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.30921052631579e-06,
+      "loss": 0.817,
+      "step": 10430
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.308662280701754e-06,
+      "loss": 0.4096,
+      "step": 10431
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.30811403508772e-06,
+      "loss": 0.4515,
+      "step": 10432
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.307565789473684e-06,
+      "loss": 0.2364,
+      "step": 10433
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.30701754385965e-06,
+      "loss": 0.5395,
+      "step": 10434
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.3064692982456146e-06,
+      "loss": 0.4759,
+      "step": 10435
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.305921052631579e-06,
+      "loss": 0.9336,
+      "step": 10436
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.305372807017544e-06,
+      "loss": 0.2085,
+      "step": 10437
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.304824561403509e-06,
+      "loss": 0.6775,
+      "step": 10438
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.304276315789474e-06,
+      "loss": 0.3942,
+      "step": 10439
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.3037280701754395e-06,
+      "loss": 0.9699,
+      "step": 10440
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.3031798245614035e-06,
+      "loss": 0.4385,
+      "step": 10441
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.302631578947369e-06,
+      "loss": 0.4803,
+      "step": 10442
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.302083333333334e-06,
+      "loss": 0.806,
+      "step": 10443
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.301535087719299e-06,
+      "loss": 0.4293,
+      "step": 10444
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.300986842105264e-06,
+      "loss": 0.1872,
+      "step": 10445
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.3004385964912285e-06,
+      "loss": 0.5785,
+      "step": 10446
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.299890350877193e-06,
+      "loss": 0.3841,
+      "step": 10447
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.299342105263158e-06,
+      "loss": 0.3326,
+      "step": 10448
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.298793859649123e-06,
+      "loss": 0.1258,
+      "step": 10449
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.298245614035088e-06,
+      "loss": 1.5769,
+      "step": 10450
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.2976973684210534e-06,
+      "loss": 0.8353,
+      "step": 10451
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.297149122807017e-06,
+      "loss": 0.1817,
+      "step": 10452
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.296600877192983e-06,
+      "loss": 1.5354,
+      "step": 10453
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.296052631578947e-06,
+      "loss": 0.7218,
+      "step": 10454
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.295504385964913e-06,
+      "loss": 0.5161,
+      "step": 10455
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.2949561403508776e-06,
+      "loss": 0.5866,
+      "step": 10456
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.294407894736842e-06,
+      "loss": 0.8662,
+      "step": 10457
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.293859649122807e-06,
+      "loss": 0.685,
+      "step": 10458
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.293311403508773e-06,
+      "loss": 1.0077,
+      "step": 10459
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.292763157894737e-06,
+      "loss": 0.8483,
+      "step": 10460
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.2922149122807025e-06,
+      "loss": 0.2382,
+      "step": 10461
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.2916666666666665e-06,
+      "loss": 1.4172,
+      "step": 10462
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.291118421052632e-06,
+      "loss": 0.4237,
+      "step": 10463
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.290570175438597e-06,
+      "loss": 0.342,
+      "step": 10464
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.290021929824562e-06,
+      "loss": 0.3362,
+      "step": 10465
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.289473684210527e-06,
+      "loss": 1.2624,
+      "step": 10466
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.2889254385964915e-06,
+      "loss": 0.1949,
+      "step": 10467
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.288377192982456e-06,
+      "loss": 0.6232,
+      "step": 10468
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.287828947368421e-06,
+      "loss": 0.5998,
+      "step": 10469
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.287280701754386e-06,
+      "loss": 1.1227,
+      "step": 10470
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.286732456140351e-06,
+      "loss": 1.1901,
+      "step": 10471
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.2861842105263165e-06,
+      "loss": 1.1988,
+      "step": 10472
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.2856359649122804e-06,
+      "loss": 0.7952,
+      "step": 10473
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.285087719298246e-06,
+      "loss": 0.8289,
+      "step": 10474
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.284539473684211e-06,
+      "loss": 0.3495,
+      "step": 10475
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.283991228070176e-06,
+      "loss": 0.9996,
+      "step": 10476
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.283442982456141e-06,
+      "loss": 0.5598,
+      "step": 10477
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.282894736842106e-06,
+      "loss": 0.5845,
+      "step": 10478
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.28234649122807e-06,
+      "loss": 0.6557,
+      "step": 10479
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.281798245614036e-06,
+      "loss": 0.2825,
+      "step": 10480
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.28125e-06,
+      "loss": 0.5733,
+      "step": 10481
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.2807017543859656e-06,
+      "loss": 0.7686,
+      "step": 10482
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.28015350877193e-06,
+      "loss": 0.4242,
+      "step": 10483
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.279605263157895e-06,
+      "loss": 0.2691,
+      "step": 10484
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.27905701754386e-06,
+      "loss": 0.2613,
+      "step": 10485
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.278508771929825e-06,
+      "loss": 0.5262,
+      "step": 10486
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.27796052631579e-06,
+      "loss": 1.2619,
+      "step": 10487
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.2774122807017545e-06,
+      "loss": 0.667,
+      "step": 10488
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.276864035087719e-06,
+      "loss": 1.0831,
+      "step": 10489
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.276315789473684e-06,
+      "loss": 0.9889,
+      "step": 10490
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.27576754385965e-06,
+      "loss": 0.3784,
+      "step": 10491
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.275219298245614e-06,
+      "loss": 0.3856,
+      "step": 10492
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.2746710526315795e-06,
+      "loss": 0.5263,
+      "step": 10493
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.274122807017544e-06,
+      "loss": 0.178,
+      "step": 10494
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.273574561403509e-06,
+      "loss": 0.5784,
+      "step": 10495
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.273026315789474e-06,
+      "loss": 0.4448,
+      "step": 10496
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.272478070175439e-06,
+      "loss": 0.6033,
+      "step": 10497
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.271929824561404e-06,
+      "loss": 0.3713,
+      "step": 10498
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.271381578947369e-06,
+      "loss": 0.8355,
+      "step": 10499
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.270833333333333e-06,
+      "loss": 0.4817,
+      "step": 10500
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.270285087719299e-06,
+      "loss": 0.9974,
+      "step": 10501
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.269736842105264e-06,
+      "loss": 0.8329,
+      "step": 10502
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.2691885964912286e-06,
+      "loss": 0.5037,
+      "step": 10503
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.268640350877193e-06,
+      "loss": 0.4588,
+      "step": 10504
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.268092105263158e-06,
+      "loss": 1.0027,
+      "step": 10505
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.267543859649123e-06,
+      "loss": 0.5488,
+      "step": 10506
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.266995614035088e-06,
+      "loss": 1.453,
+      "step": 10507
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.266447368421053e-06,
+      "loss": 1.0706,
+      "step": 10508
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.2658991228070175e-06,
+      "loss": 1.305,
+      "step": 10509
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.265350877192983e-06,
+      "loss": 1.1742,
+      "step": 10510
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.264802631578947e-06,
+      "loss": 0.8954,
+      "step": 10511
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.264254385964913e-06,
+      "loss": 0.2938,
+      "step": 10512
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.263706140350878e-06,
+      "loss": 0.8131,
+      "step": 10513
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.2631578947368425e-06,
+      "loss": 0.7965,
+      "step": 10514
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.262609649122807e-06,
+      "loss": 0.5836,
+      "step": 10515
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.262061403508772e-06,
+      "loss": 0.9255,
+      "step": 10516
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.261513157894737e-06,
+      "loss": 0.1408,
+      "step": 10517
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.260964912280703e-06,
+      "loss": 1.3008,
+      "step": 10518
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.260416666666667e-06,
+      "loss": 1.5302,
+      "step": 10519
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.259868421052632e-06,
+      "loss": 0.8975,
+      "step": 10520
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.259320175438597e-06,
+      "loss": 0.227,
+      "step": 10521
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.258771929824562e-06,
+      "loss": 0.7583,
+      "step": 10522
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.258223684210527e-06,
+      "loss": 0.7748,
+      "step": 10523
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.2576754385964916e-06,
+      "loss": 0.7111,
+      "step": 10524
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.257127192982456e-06,
+      "loss": 1.1244,
+      "step": 10525
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.256578947368421e-06,
+      "loss": 0.4907,
+      "step": 10526
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.256030701754386e-06,
+      "loss": 1.1216,
+      "step": 10527
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.255482456140351e-06,
+      "loss": 0.3936,
+      "step": 10528
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.2549342105263165e-06,
+      "loss": 0.4507,
+      "step": 10529
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.254385964912281e-06,
+      "loss": 0.5258,
+      "step": 10530
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.253837719298246e-06,
+      "loss": 0.9098,
+      "step": 10531
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.253289473684211e-06,
+      "loss": 1.1267,
+      "step": 10532
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.252741228070176e-06,
+      "loss": 0.5871,
+      "step": 10533
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.252192982456141e-06,
+      "loss": 0.64,
+      "step": 10534
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.2516447368421055e-06,
+      "loss": 0.1333,
+      "step": 10535
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.25109649122807e-06,
+      "loss": 1.7785,
+      "step": 10536
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.250548245614036e-06,
+      "loss": 0.6131,
+      "step": 10537
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.25e-06,
+      "loss": 0.6443,
+      "step": 10538
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.249451754385966e-06,
+      "loss": 0.5079,
+      "step": 10539
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.24890350877193e-06,
+      "loss": 1.0633,
+      "step": 10540
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.248355263157895e-06,
+      "loss": 0.3194,
+      "step": 10541
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.24780701754386e-06,
+      "loss": 0.8232,
+      "step": 10542
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.247258771929825e-06,
+      "loss": 0.7514,
+      "step": 10543
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.24671052631579e-06,
+      "loss": 0.694,
+      "step": 10544
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.246162280701755e-06,
+      "loss": 0.1378,
+      "step": 10545
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.245614035087719e-06,
+      "loss": 0.4544,
+      "step": 10546
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.245065789473684e-06,
+      "loss": 0.6476,
+      "step": 10547
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.244517543859649e-06,
+      "loss": 0.418,
+      "step": 10548
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.243969298245615e-06,
+      "loss": 0.3605,
+      "step": 10549
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.2434210526315796e-06,
+      "loss": 1.1797,
+      "step": 10550
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.242872807017544e-06,
+      "loss": 0.3556,
+      "step": 10551
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.242324561403509e-06,
+      "loss": 0.4522,
+      "step": 10552
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.241776315789474e-06,
+      "loss": 1.0912,
+      "step": 10553
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.241228070175439e-06,
+      "loss": 0.2941,
+      "step": 10554
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.240679824561404e-06,
+      "loss": 0.5328,
+      "step": 10555
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.240131578947369e-06,
+      "loss": 0.2338,
+      "step": 10556
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.239583333333333e-06,
+      "loss": 0.7784,
+      "step": 10557
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.239035087719299e-06,
+      "loss": 0.7078,
+      "step": 10558
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.238486842105263e-06,
+      "loss": 0.6536,
+      "step": 10559
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.237938596491229e-06,
+      "loss": 1.0799,
+      "step": 10560
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.2373903508771935e-06,
+      "loss": 0.6312,
+      "step": 10561
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.236842105263158e-06,
+      "loss": 0.3454,
+      "step": 10562
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.236293859649123e-06,
+      "loss": 0.3707,
+      "step": 10563
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.235745614035088e-06,
+      "loss": 0.5266,
+      "step": 10564
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.235197368421053e-06,
+      "loss": 0.4381,
+      "step": 10565
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.234649122807018e-06,
+      "loss": 1.5778,
+      "step": 10566
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.234100877192982e-06,
+      "loss": 0.5465,
+      "step": 10567
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.233552631578948e-06,
+      "loss": 1.2125,
+      "step": 10568
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.233004385964913e-06,
+      "loss": 0.6237,
+      "step": 10569
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.232456140350878e-06,
+      "loss": 0.3833,
+      "step": 10570
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.2319078947368426e-06,
+      "loss": 0.5676,
+      "step": 10571
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.231359649122807e-06,
+      "loss": 0.8887,
+      "step": 10572
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.230811403508772e-06,
+      "loss": 0.3561,
+      "step": 10573
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.230263157894737e-06,
+      "loss": 0.6211,
+      "step": 10574
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.229714912280702e-06,
+      "loss": 0.6638,
+      "step": 10575
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.229166666666667e-06,
+      "loss": 0.5751,
+      "step": 10576
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.228618421052632e-06,
+      "loss": 0.5176,
+      "step": 10577
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.228070175438596e-06,
+      "loss": 0.9528,
+      "step": 10578
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.227521929824562e-06,
+      "loss": 0.5661,
+      "step": 10579
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.226973684210527e-06,
+      "loss": 0.5637,
+      "step": 10580
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.226425438596492e-06,
+      "loss": 1.2539,
+      "step": 10581
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.2258771929824565e-06,
+      "loss": 0.1185,
+      "step": 10582
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.225328947368421e-06,
+      "loss": 0.324,
+      "step": 10583
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.224780701754386e-06,
+      "loss": 0.3028,
+      "step": 10584
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.224232456140351e-06,
+      "loss": 0.6571,
+      "step": 10585
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.223684210526316e-06,
+      "loss": 0.3487,
+      "step": 10586
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.2231359649122815e-06,
+      "loss": 0.626,
+      "step": 10587
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.222587719298246e-06,
+      "loss": 1.0828,
+      "step": 10588
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.222039473684211e-06,
+      "loss": 0.4895,
+      "step": 10589
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.221491228070176e-06,
+      "loss": 0.5106,
+      "step": 10590
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.220942982456141e-06,
+      "loss": 0.696,
+      "step": 10591
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.220394736842106e-06,
+      "loss": 0.9611,
+      "step": 10592
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.21984649122807e-06,
+      "loss": 0.6224,
+      "step": 10593
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.219298245614035e-06,
+      "loss": 1.448,
+      "step": 10594
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.21875e-06,
+      "loss": 0.3717,
+      "step": 10595
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.218201754385966e-06,
+      "loss": 0.4798,
+      "step": 10596
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.21765350877193e-06,
+      "loss": 1.2799,
+      "step": 10597
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.217105263157895e-06,
+      "loss": 0.453,
+      "step": 10598
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.21655701754386e-06,
+      "loss": 0.4288,
+      "step": 10599
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.216008771929825e-06,
+      "loss": 0.5752,
+      "step": 10600
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.21546052631579e-06,
+      "loss": 0.3379,
+      "step": 10601
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.214912280701755e-06,
+      "loss": 0.3381,
+      "step": 10602
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.2143640350877195e-06,
+      "loss": 0.6603,
+      "step": 10603
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.213815789473685e-06,
+      "loss": 0.3582,
+      "step": 10604
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.213267543859649e-06,
+      "loss": 0.7287,
+      "step": 10605
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.212719298245615e-06,
+      "loss": 0.3232,
+      "step": 10606
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.21217105263158e-06,
+      "loss": 0.6159,
+      "step": 10607
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.2116228070175445e-06,
+      "loss": 0.8181,
+      "step": 10608
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.211074561403509e-06,
+      "loss": 0.4671,
+      "step": 10609
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.210526315789474e-06,
+      "loss": 0.2361,
+      "step": 10610
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.209978070175439e-06,
+      "loss": 1.271,
+      "step": 10611
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.209429824561404e-06,
+      "loss": 0.9808,
+      "step": 10612
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.208881578947369e-06,
+      "loss": 0.3256,
+      "step": 10613
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.208333333333333e-06,
+      "loss": 0.8254,
+      "step": 10614
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.207785087719299e-06,
+      "loss": 0.9748,
+      "step": 10615
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.207236842105263e-06,
+      "loss": 0.5386,
+      "step": 10616
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.206688596491229e-06,
+      "loss": 0.9474,
+      "step": 10617
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.206140350877193e-06,
+      "loss": 0.4711,
+      "step": 10618
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.205592105263158e-06,
+      "loss": 0.9491,
+      "step": 10619
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.205043859649123e-06,
+      "loss": 0.3054,
+      "step": 10620
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.204495614035088e-06,
+      "loss": 0.6846,
+      "step": 10621
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.203947368421053e-06,
+      "loss": 0.4469,
+      "step": 10622
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.2033991228070185e-06,
+      "loss": 0.2491,
+      "step": 10623
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.2028508771929825e-06,
+      "loss": 0.4099,
+      "step": 10624
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.202302631578948e-06,
+      "loss": 0.3895,
+      "step": 10625
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.201754385964912e-06,
+      "loss": 0.7569,
+      "step": 10626
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.201206140350878e-06,
+      "loss": 0.5113,
+      "step": 10627
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.200657894736843e-06,
+      "loss": 0.4815,
+      "step": 10628
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.2001096491228075e-06,
+      "loss": 0.8421,
+      "step": 10629
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.199561403508772e-06,
+      "loss": 0.4706,
+      "step": 10630
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.199013157894737e-06,
+      "loss": 0.4917,
+      "step": 10631
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.198464912280702e-06,
+      "loss": 0.6673,
+      "step": 10632
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.197916666666667e-06,
+      "loss": 0.8544,
+      "step": 10633
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.197368421052632e-06,
+      "loss": 0.6455,
+      "step": 10634
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.1968201754385964e-06,
+      "loss": 0.1332,
+      "step": 10635
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.196271929824562e-06,
+      "loss": 0.3872,
+      "step": 10636
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.195723684210526e-06,
+      "loss": 0.453,
+      "step": 10637
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.195175438596492e-06,
+      "loss": 0.116,
+      "step": 10638
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.1946271929824566e-06,
+      "loss": 0.9992,
+      "step": 10639
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.194078947368421e-06,
+      "loss": 0.8773,
+      "step": 10640
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.193530701754386e-06,
+      "loss": 0.3445,
+      "step": 10641
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.192982456140352e-06,
+      "loss": 1.0917,
+      "step": 10642
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.192434210526316e-06,
+      "loss": 0.2647,
+      "step": 10643
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.1918859649122815e-06,
+      "loss": 0.4221,
+      "step": 10644
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.1913377192982455e-06,
+      "loss": 0.7482,
+      "step": 10645
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.190789473684211e-06,
+      "loss": 0.2226,
+      "step": 10646
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.190241228070176e-06,
+      "loss": 0.0728,
+      "step": 10647
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.189692982456141e-06,
+      "loss": 0.1947,
+      "step": 10648
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.189144736842106e-06,
+      "loss": 0.8613,
+      "step": 10649
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.1885964912280705e-06,
+      "loss": 0.85,
+      "step": 10650
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.188048245614035e-06,
+      "loss": 0.6368,
+      "step": 10651
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.1875e-06,
+      "loss": 0.7317,
+      "step": 10652
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.186951754385965e-06,
+      "loss": 1.0659,
+      "step": 10653
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.18640350877193e-06,
+      "loss": 0.3051,
+      "step": 10654
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.1858552631578955e-06,
+      "loss": 0.1337,
+      "step": 10655
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.1853070175438594e-06,
+      "loss": 1.025,
+      "step": 10656
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.184758771929825e-06,
+      "loss": 0.5909,
+      "step": 10657
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.18421052631579e-06,
+      "loss": 0.9508,
+      "step": 10658
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.183662280701755e-06,
+      "loss": 0.6857,
+      "step": 10659
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.18311403508772e-06,
+      "loss": 0.3821,
+      "step": 10660
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.182565789473684e-06,
+      "loss": 0.4266,
+      "step": 10661
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.182017543859649e-06,
+      "loss": 0.6053,
+      "step": 10662
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.181469298245615e-06,
+      "loss": 0.2431,
+      "step": 10663
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.180921052631579e-06,
+      "loss": 0.4432,
+      "step": 10664
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.1803728070175446e-06,
+      "loss": 1.1553,
+      "step": 10665
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.179824561403509e-06,
+      "loss": 0.2483,
+      "step": 10666
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.179276315789474e-06,
+      "loss": 1.2139,
+      "step": 10667
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.178728070175439e-06,
+      "loss": 0.6881,
+      "step": 10668
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.178179824561404e-06,
+      "loss": 0.3156,
+      "step": 10669
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.177631578947369e-06,
+      "loss": 0.4165,
+      "step": 10670
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.1770833333333335e-06,
+      "loss": 0.3592,
+      "step": 10671
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.176535087719298e-06,
+      "loss": 0.3496,
+      "step": 10672
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.175986842105263e-06,
+      "loss": 0.3346,
+      "step": 10673
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.175438596491229e-06,
+      "loss": 0.9145,
+      "step": 10674
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.174890350877193e-06,
+      "loss": 0.8454,
+      "step": 10675
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.1743421052631585e-06,
+      "loss": 0.3222,
+      "step": 10676
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.173793859649123e-06,
+      "loss": 0.9567,
+      "step": 10677
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.173245614035088e-06,
+      "loss": 0.0858,
+      "step": 10678
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.172697368421053e-06,
+      "loss": 0.94,
+      "step": 10679
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.172149122807018e-06,
+      "loss": 1.1807,
+      "step": 10680
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.171600877192983e-06,
+      "loss": 0.9047,
+      "step": 10681
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.171052631578948e-06,
+      "loss": 0.2265,
+      "step": 10682
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.170504385964912e-06,
+      "loss": 0.6088,
+      "step": 10683
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.169956140350878e-06,
+      "loss": 0.5139,
+      "step": 10684
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.169407894736843e-06,
+      "loss": 0.7013,
+      "step": 10685
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.1688596491228076e-06,
+      "loss": 0.2709,
+      "step": 10686
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.168311403508772e-06,
+      "loss": 2.0823,
+      "step": 10687
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.167763157894737e-06,
+      "loss": 0.6893,
+      "step": 10688
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.167214912280702e-06,
+      "loss": 0.7223,
+      "step": 10689
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.166666666666667e-06,
+      "loss": 0.2753,
+      "step": 10690
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.166118421052632e-06,
+      "loss": 0.7041,
+      "step": 10691
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.1655701754385965e-06,
+      "loss": 0.4963,
+      "step": 10692
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.165021929824562e-06,
+      "loss": 0.5168,
+      "step": 10693
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.164473684210526e-06,
+      "loss": 0.5578,
+      "step": 10694
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.163925438596492e-06,
+      "loss": 1.0267,
+      "step": 10695
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.163377192982457e-06,
+      "loss": 0.1805,
+      "step": 10696
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.1628289473684215e-06,
+      "loss": 0.1162,
+      "step": 10697
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.162280701754386e-06,
+      "loss": 0.3956,
+      "step": 10698
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.161732456140351e-06,
+      "loss": 0.4614,
+      "step": 10699
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.161184210526316e-06,
+      "loss": 0.862,
+      "step": 10700
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.160635964912282e-06,
+      "loss": 1.0043,
+      "step": 10701
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.160087719298246e-06,
+      "loss": 0.4026,
+      "step": 10702
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.159539473684211e-06,
+      "loss": 0.8522,
+      "step": 10703
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.158991228070175e-06,
+      "loss": 0.6136,
+      "step": 10704
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.158442982456141e-06,
+      "loss": 0.4552,
+      "step": 10705
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.157894736842106e-06,
+      "loss": 1.1039,
+      "step": 10706
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.157346491228071e-06,
+      "loss": 0.5429,
+      "step": 10707
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.156798245614035e-06,
+      "loss": 0.5325,
+      "step": 10708
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.15625e-06,
+      "loss": 0.26,
+      "step": 10709
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.155701754385965e-06,
+      "loss": 0.3241,
+      "step": 10710
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.15515350877193e-06,
+      "loss": 0.8957,
+      "step": 10711
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.154605263157895e-06,
+      "loss": 0.4114,
+      "step": 10712
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.1540570175438595e-06,
+      "loss": 0.6517,
+      "step": 10713
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.153508771929825e-06,
+      "loss": 0.4323,
+      "step": 10714
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.15296052631579e-06,
+      "loss": 0.6785,
+      "step": 10715
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.152412280701755e-06,
+      "loss": 0.3629,
+      "step": 10716
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.15186403508772e-06,
+      "loss": 0.4424,
+      "step": 10717
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.1513157894736845e-06,
+      "loss": 0.9804,
+      "step": 10718
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.150767543859649e-06,
+      "loss": 1.1987,
+      "step": 10719
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.150219298245615e-06,
+      "loss": 0.6373,
+      "step": 10720
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.149671052631579e-06,
+      "loss": 1.1969,
+      "step": 10721
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.149122807017545e-06,
+      "loss": 0.468,
+      "step": 10722
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.148574561403509e-06,
+      "loss": 1.046,
+      "step": 10723
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.148026315789474e-06,
+      "loss": 0.3684,
+      "step": 10724
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.147478070175439e-06,
+      "loss": 0.3105,
+      "step": 10725
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.146929824561404e-06,
+      "loss": 0.1159,
+      "step": 10726
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.146381578947369e-06,
+      "loss": 1.1195,
+      "step": 10727
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.145833333333334e-06,
+      "loss": 0.2082,
+      "step": 10728
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.145285087719298e-06,
+      "loss": 0.6773,
+      "step": 10729
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.144736842105263e-06,
+      "loss": 0.3627,
+      "step": 10730
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.144188596491228e-06,
+      "loss": 0.7246,
+      "step": 10731
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.143640350877193e-06,
+      "loss": 0.8476,
+      "step": 10732
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.1430921052631586e-06,
+      "loss": 1.1723,
+      "step": 10733
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.142543859649123e-06,
+      "loss": 0.4399,
+      "step": 10734
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.141995614035088e-06,
+      "loss": 0.1429,
+      "step": 10735
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.141447368421053e-06,
+      "loss": 0.4112,
+      "step": 10736
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.140899122807018e-06,
+      "loss": 0.3988,
+      "step": 10737
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.140350877192983e-06,
+      "loss": 0.4856,
+      "step": 10738
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.1398026315789475e-06,
+      "loss": 0.4728,
+      "step": 10739
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.139254385964912e-06,
+      "loss": 0.3446,
+      "step": 10740
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.138706140350878e-06,
+      "loss": 0.7196,
+      "step": 10741
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.138157894736842e-06,
+      "loss": 0.6691,
+      "step": 10742
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.137609649122808e-06,
+      "loss": 0.1929,
+      "step": 10743
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.1370614035087725e-06,
+      "loss": 0.3209,
+      "step": 10744
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.136513157894737e-06,
+      "loss": 0.1444,
+      "step": 10745
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.135964912280702e-06,
+      "loss": 0.6006,
+      "step": 10746
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.135416666666667e-06,
+      "loss": 1.6217,
+      "step": 10747
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.134868421052632e-06,
+      "loss": 0.6865,
+      "step": 10748
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.134320175438597e-06,
+      "loss": 0.9763,
+      "step": 10749
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.1337719298245614e-06,
+      "loss": 0.9155,
+      "step": 10750
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.133223684210526e-06,
+      "loss": 0.2052,
+      "step": 10751
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.132675438596492e-06,
+      "loss": 0.919,
+      "step": 10752
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.132127192982457e-06,
+      "loss": 0.5839,
+      "step": 10753
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.1315789473684216e-06,
+      "loss": 0.9652,
+      "step": 10754
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.131030701754386e-06,
+      "loss": 0.6508,
+      "step": 10755
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.130482456140351e-06,
+      "loss": 0.479,
+      "step": 10756
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.129934210526316e-06,
+      "loss": 0.9231,
+      "step": 10757
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.129385964912281e-06,
+      "loss": 0.479,
+      "step": 10758
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.128837719298246e-06,
+      "loss": 0.7753,
+      "step": 10759
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.128289473684211e-06,
+      "loss": 0.5495,
+      "step": 10760
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.127741228070175e-06,
+      "loss": 0.3764,
+      "step": 10761
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.127192982456141e-06,
+      "loss": 0.7704,
+      "step": 10762
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.126644736842106e-06,
+      "loss": 0.8357,
+      "step": 10763
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.126096491228071e-06,
+      "loss": 0.7272,
+      "step": 10764
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.1255482456140355e-06,
+      "loss": 0.4743,
+      "step": 10765
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.125e-06,
+      "loss": 0.4895,
+      "step": 10766
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.124451754385965e-06,
+      "loss": 0.94,
+      "step": 10767
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.12390350877193e-06,
+      "loss": 0.3858,
+      "step": 10768
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.123355263157895e-06,
+      "loss": 0.8829,
+      "step": 10769
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.12280701754386e-06,
+      "loss": 0.1664,
+      "step": 10770
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.122258771929825e-06,
+      "loss": 0.2111,
+      "step": 10771
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.12171052631579e-06,
+      "loss": 0.641,
+      "step": 10772
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.121162280701755e-06,
+      "loss": 0.1248,
+      "step": 10773
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.12061403508772e-06,
+      "loss": 0.8044,
+      "step": 10774
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.120065789473685e-06,
+      "loss": 0.4628,
+      "step": 10775
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.119517543859649e-06,
+      "loss": 0.4301,
+      "step": 10776
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.118969298245614e-06,
+      "loss": 0.772,
+      "step": 10777
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.118421052631579e-06,
+      "loss": 0.7171,
+      "step": 10778
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.117872807017545e-06,
+      "loss": 0.2529,
+      "step": 10779
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.117324561403509e-06,
+      "loss": 0.3712,
+      "step": 10780
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.116776315789474e-06,
+      "loss": 0.3166,
+      "step": 10781
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.116228070175438e-06,
+      "loss": 1.0712,
+      "step": 10782
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.115679824561404e-06,
+      "loss": 0.7724,
+      "step": 10783
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.115131578947369e-06,
+      "loss": 0.2536,
+      "step": 10784
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.114583333333334e-06,
+      "loss": 1.0264,
+      "step": 10785
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.1140350877192985e-06,
+      "loss": 0.9393,
+      "step": 10786
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.113486842105263e-06,
+      "loss": 1.6325,
+      "step": 10787
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.112938596491228e-06,
+      "loss": 0.2578,
+      "step": 10788
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.112390350877194e-06,
+      "loss": 0.2468,
+      "step": 10789
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.111842105263158e-06,
+      "loss": 0.5086,
+      "step": 10790
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.1112938596491235e-06,
+      "loss": 1.0931,
+      "step": 10791
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.110745614035088e-06,
+      "loss": 0.5848,
+      "step": 10792
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.110197368421053e-06,
+      "loss": 0.5553,
+      "step": 10793
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.109649122807018e-06,
+      "loss": 0.4984,
+      "step": 10794
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.109100877192983e-06,
+      "loss": 0.7409,
+      "step": 10795
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.108552631578948e-06,
+      "loss": 0.3221,
+      "step": 10796
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.108004385964912e-06,
+      "loss": 0.572,
+      "step": 10797
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.107456140350877e-06,
+      "loss": 0.3805,
+      "step": 10798
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.106907894736842e-06,
+      "loss": 0.9701,
+      "step": 10799
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.106359649122808e-06,
+      "loss": 0.8091,
+      "step": 10800
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.105811403508772e-06,
+      "loss": 0.3949,
+      "step": 10801
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.105263157894737e-06,
+      "loss": 0.3844,
+      "step": 10802
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.104714912280702e-06,
+      "loss": 0.4878,
+      "step": 10803
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.104166666666667e-06,
+      "loss": 0.3401,
+      "step": 10804
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.103618421052632e-06,
+      "loss": 0.8046,
+      "step": 10805
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.103070175438597e-06,
+      "loss": 0.9418,
+      "step": 10806
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.1025219298245615e-06,
+      "loss": 0.5656,
+      "step": 10807
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.101973684210527e-06,
+      "loss": 0.6124,
+      "step": 10808
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.101425438596491e-06,
+      "loss": 1.2385,
+      "step": 10809
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.100877192982457e-06,
+      "loss": 0.6906,
+      "step": 10810
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.100328947368422e-06,
+      "loss": 1.138,
+      "step": 10811
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.0997807017543865e-06,
+      "loss": 1.2223,
+      "step": 10812
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.099232456140351e-06,
+      "loss": 0.398,
+      "step": 10813
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.098684210526316e-06,
+      "loss": 0.4566,
+      "step": 10814
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.098135964912281e-06,
+      "loss": 0.6678,
+      "step": 10815
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.097587719298246e-06,
+      "loss": 0.5934,
+      "step": 10816
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.097039473684211e-06,
+      "loss": 0.6256,
+      "step": 10817
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.0964912280701754e-06,
+      "loss": 0.8238,
+      "step": 10818
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.095942982456141e-06,
+      "loss": 0.7674,
+      "step": 10819
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.095394736842105e-06,
+      "loss": 0.7096,
+      "step": 10820
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.094846491228071e-06,
+      "loss": 0.4977,
+      "step": 10821
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.094298245614036e-06,
+      "loss": 0.5854,
+      "step": 10822
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.09375e-06,
+      "loss": 0.6791,
+      "step": 10823
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.093201754385965e-06,
+      "loss": 1.0426,
+      "step": 10824
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.09265350877193e-06,
+      "loss": 0.3633,
+      "step": 10825
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.092105263157895e-06,
+      "loss": 0.607,
+      "step": 10826
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.0915570175438605e-06,
+      "loss": 0.3974,
+      "step": 10827
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.0910087719298245e-06,
+      "loss": 0.1826,
+      "step": 10828
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.09046052631579e-06,
+      "loss": 0.6297,
+      "step": 10829
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.089912280701755e-06,
+      "loss": 0.5924,
+      "step": 10830
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.08936403508772e-06,
+      "loss": 0.3099,
+      "step": 10831
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.088815789473685e-06,
+      "loss": 0.2915,
+      "step": 10832
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.0882675438596495e-06,
+      "loss": 0.502,
+      "step": 10833
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.087719298245614e-06,
+      "loss": 0.2156,
+      "step": 10834
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.087171052631579e-06,
+      "loss": 1.2588,
+      "step": 10835
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.086622807017544e-06,
+      "loss": 0.4821,
+      "step": 10836
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.086074561403509e-06,
+      "loss": 0.2669,
+      "step": 10837
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.0855263157894745e-06,
+      "loss": 0.212,
+      "step": 10838
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.0849780701754384e-06,
+      "loss": 1.1869,
+      "step": 10839
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.084429824561404e-06,
+      "loss": 0.1712,
+      "step": 10840
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.083881578947369e-06,
+      "loss": 0.6284,
+      "step": 10841
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.083333333333334e-06,
+      "loss": 0.8632,
+      "step": 10842
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.082785087719299e-06,
+      "loss": 0.7368,
+      "step": 10843
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.082236842105263e-06,
+      "loss": 1.3415,
+      "step": 10844
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.081688596491228e-06,
+      "loss": 1.1712,
+      "step": 10845
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.081140350877194e-06,
+      "loss": 0.2857,
+      "step": 10846
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.080592105263158e-06,
+      "loss": 0.9337,
+      "step": 10847
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.0800438596491236e-06,
+      "loss": 0.6776,
+      "step": 10848
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.079495614035088e-06,
+      "loss": 0.2598,
+      "step": 10849
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.078947368421053e-06,
+      "loss": 0.9131,
+      "step": 10850
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.078399122807018e-06,
+      "loss": 0.4378,
+      "step": 10851
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.077850877192983e-06,
+      "loss": 0.6303,
+      "step": 10852
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.077302631578948e-06,
+      "loss": 0.348,
+      "step": 10853
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.0767543859649125e-06,
+      "loss": 0.3399,
+      "step": 10854
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.076206140350877e-06,
+      "loss": 0.7504,
+      "step": 10855
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.075657894736842e-06,
+      "loss": 0.2922,
+      "step": 10856
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.075109649122808e-06,
+      "loss": 0.6074,
+      "step": 10857
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.074561403508772e-06,
+      "loss": 0.5635,
+      "step": 10858
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.0740131578947375e-06,
+      "loss": 0.153,
+      "step": 10859
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.0734649122807015e-06,
+      "loss": 1.1031,
+      "step": 10860
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.072916666666667e-06,
+      "loss": 0.5965,
+      "step": 10861
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.072368421052632e-06,
+      "loss": 0.4105,
+      "step": 10862
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.071820175438597e-06,
+      "loss": 0.5312,
+      "step": 10863
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.071271929824562e-06,
+      "loss": 0.575,
+      "step": 10864
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.070723684210527e-06,
+      "loss": 0.8666,
+      "step": 10865
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.070175438596491e-06,
+      "loss": 0.5894,
+      "step": 10866
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.069627192982457e-06,
+      "loss": 0.7484,
+      "step": 10867
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.069078947368421e-06,
+      "loss": 1.4474,
+      "step": 10868
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.0685307017543866e-06,
+      "loss": 0.2834,
+      "step": 10869
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.067982456140351e-06,
+      "loss": 0.3797,
+      "step": 10870
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.067434210526316e-06,
+      "loss": 0.1689,
+      "step": 10871
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.066885964912281e-06,
+      "loss": 0.7952,
+      "step": 10872
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.066337719298246e-06,
+      "loss": 0.7251,
+      "step": 10873
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.065789473684211e-06,
+      "loss": 0.2056,
+      "step": 10874
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.0652412280701755e-06,
+      "loss": 0.9378,
+      "step": 10875
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.06469298245614e-06,
+      "loss": 0.5753,
+      "step": 10876
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.064144736842105e-06,
+      "loss": 0.5043,
+      "step": 10877
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.063596491228071e-06,
+      "loss": 1.2037,
+      "step": 10878
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.063048245614035e-06,
+      "loss": 0.7334,
+      "step": 10879
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.0625000000000005e-06,
+      "loss": 0.5938,
+      "step": 10880
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.061951754385965e-06,
+      "loss": 1.1752,
+      "step": 10881
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.06140350877193e-06,
+      "loss": 0.2854,
+      "step": 10882
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.060855263157895e-06,
+      "loss": 0.1773,
+      "step": 10883
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.060307017543861e-06,
+      "loss": 0.7049,
+      "step": 10884
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.059758771929825e-06,
+      "loss": 1.0935,
+      "step": 10885
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.05921052631579e-06,
+      "loss": 0.8094,
+      "step": 10886
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.058662280701754e-06,
+      "loss": 0.3411,
+      "step": 10887
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.05811403508772e-06,
+      "loss": 0.5224,
+      "step": 10888
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.057565789473685e-06,
+      "loss": 0.7597,
+      "step": 10889
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.05701754385965e-06,
+      "loss": 0.6131,
+      "step": 10890
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.056469298245614e-06,
+      "loss": 0.1756,
+      "step": 10891
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.055921052631579e-06,
+      "loss": 0.2789,
+      "step": 10892
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.055372807017544e-06,
+      "loss": 0.7939,
+      "step": 10893
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.054824561403509e-06,
+      "loss": 0.49,
+      "step": 10894
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.054276315789474e-06,
+      "loss": 2.1441,
+      "step": 10895
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.0537280701754385e-06,
+      "loss": 0.5908,
+      "step": 10896
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.053179824561404e-06,
+      "loss": 0.9249,
+      "step": 10897
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.052631578947368e-06,
+      "loss": 0.2669,
+      "step": 10898
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.052083333333334e-06,
+      "loss": 0.5699,
+      "step": 10899
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.051535087719299e-06,
+      "loss": 0.253,
+      "step": 10900
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.0509868421052635e-06,
+      "loss": 0.6633,
+      "step": 10901
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.050438596491228e-06,
+      "loss": 0.217,
+      "step": 10902
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.049890350877193e-06,
+      "loss": 1.0729,
+      "step": 10903
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.049342105263158e-06,
+      "loss": 0.4213,
+      "step": 10904
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.048793859649124e-06,
+      "loss": 1.2493,
+      "step": 10905
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.048245614035088e-06,
+      "loss": 0.6988,
+      "step": 10906
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.047697368421053e-06,
+      "loss": 0.4367,
+      "step": 10907
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.047149122807018e-06,
+      "loss": 1.2755,
+      "step": 10908
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.046600877192983e-06,
+      "loss": 1.5181,
+      "step": 10909
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.046052631578948e-06,
+      "loss": 0.5617,
+      "step": 10910
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.045504385964913e-06,
+      "loss": 0.439,
+      "step": 10911
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.044956140350877e-06,
+      "loss": 1.0103,
+      "step": 10912
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.044407894736842e-06,
+      "loss": 1.0329,
+      "step": 10913
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.043859649122807e-06,
+      "loss": 0.3221,
+      "step": 10914
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.043311403508772e-06,
+      "loss": 1.3915,
+      "step": 10915
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.0427631578947376e-06,
+      "loss": 0.569,
+      "step": 10916
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.0422149122807015e-06,
+      "loss": 0.5822,
+      "step": 10917
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.041666666666667e-06,
+      "loss": 0.2559,
+      "step": 10918
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.041118421052632e-06,
+      "loss": 0.6688,
+      "step": 10919
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.040570175438597e-06,
+      "loss": 0.7176,
+      "step": 10920
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.040021929824562e-06,
+      "loss": 0.478,
+      "step": 10921
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.0394736842105265e-06,
+      "loss": 0.7138,
+      "step": 10922
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.038925438596491e-06,
+      "loss": 0.3273,
+      "step": 10923
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.038377192982457e-06,
+      "loss": 0.6535,
+      "step": 10924
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.037828947368421e-06,
+      "loss": 0.8917,
+      "step": 10925
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.037280701754387e-06,
+      "loss": 0.722,
+      "step": 10926
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.0367324561403515e-06,
+      "loss": 0.8356,
+      "step": 10927
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.036184210526316e-06,
+      "loss": 0.9904,
+      "step": 10928
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.035635964912281e-06,
+      "loss": 0.5785,
+      "step": 10929
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.035087719298246e-06,
+      "loss": 0.6773,
+      "step": 10930
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.034539473684211e-06,
+      "loss": 0.4049,
+      "step": 10931
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.033991228070176e-06,
+      "loss": 1.1152,
+      "step": 10932
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.0334429824561404e-06,
+      "loss": 0.358,
+      "step": 10933
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.032894736842105e-06,
+      "loss": 1.334,
+      "step": 10934
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.032346491228071e-06,
+      "loss": 0.7587,
+      "step": 10935
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.031798245614035e-06,
+      "loss": 0.7903,
+      "step": 10936
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.031250000000001e-06,
+      "loss": 0.8582,
+      "step": 10937
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.030701754385965e-06,
+      "loss": 0.2362,
+      "step": 10938
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.03015350877193e-06,
+      "loss": 1.0387,
+      "step": 10939
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.029605263157895e-06,
+      "loss": 0.9282,
+      "step": 10940
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.02905701754386e-06,
+      "loss": 0.3446,
+      "step": 10941
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.028508771929825e-06,
+      "loss": 0.1285,
+      "step": 10942
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.02796052631579e-06,
+      "loss": 0.2833,
+      "step": 10943
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.027412280701754e-06,
+      "loss": 0.4537,
+      "step": 10944
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.02686403508772e-06,
+      "loss": 0.7202,
+      "step": 10945
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.026315789473684e-06,
+      "loss": 0.3402,
+      "step": 10946
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.02576754385965e-06,
+      "loss": 0.5267,
+      "step": 10947
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.0252192982456145e-06,
+      "loss": 0.1086,
+      "step": 10948
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.024671052631579e-06,
+      "loss": 0.6519,
+      "step": 10949
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.024122807017544e-06,
+      "loss": 0.1412,
+      "step": 10950
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.023574561403509e-06,
+      "loss": 1.114,
+      "step": 10951
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.023026315789474e-06,
+      "loss": 0.8485,
+      "step": 10952
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.022478070175439e-06,
+      "loss": 0.6994,
+      "step": 10953
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.0219298245614034e-06,
+      "loss": 0.7169,
+      "step": 10954
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.021381578947368e-06,
+      "loss": 0.2596,
+      "step": 10955
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.020833333333334e-06,
+      "loss": 0.5347,
+      "step": 10956
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.020285087719299e-06,
+      "loss": 0.7886,
+      "step": 10957
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.019736842105264e-06,
+      "loss": 0.6909,
+      "step": 10958
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.019188596491228e-06,
+      "loss": 0.6063,
+      "step": 10959
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.018640350877193e-06,
+      "loss": 1.398,
+      "step": 10960
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.018092105263158e-06,
+      "loss": 0.775,
+      "step": 10961
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.017543859649124e-06,
+      "loss": 1.1879,
+      "step": 10962
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.016995614035088e-06,
+      "loss": 0.5853,
+      "step": 10963
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.016447368421053e-06,
+      "loss": 0.6866,
+      "step": 10964
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.015899122807017e-06,
+      "loss": 0.3152,
+      "step": 10965
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.015350877192983e-06,
+      "loss": 0.4044,
+      "step": 10966
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.014802631578948e-06,
+      "loss": 1.3469,
+      "step": 10967
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.014254385964913e-06,
+      "loss": 0.3049,
+      "step": 10968
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.0137061403508775e-06,
+      "loss": 0.3561,
+      "step": 10969
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.013157894736842e-06,
+      "loss": 0.9122,
+      "step": 10970
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.012609649122807e-06,
+      "loss": 0.1059,
+      "step": 10971
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.012061403508772e-06,
+      "loss": 1.1563,
+      "step": 10972
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.011513157894737e-06,
+      "loss": 0.4094,
+      "step": 10973
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.0109649122807025e-06,
+      "loss": 0.7683,
+      "step": 10974
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.7498462169366413,
+      "eval_loss": 0.712877631187439,
+      "eval_roc_auc": 0.9357763788811468,
+      "eval_runtime": 190.4379,
+      "eval_samples_per_second": 25.609,
+      "eval_steps_per_second": 3.203,
+      "step": 10974
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.010416666666667e-06,
+      "loss": 1.0908,
+      "step": 10975
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.009868421052632e-06,
+      "loss": 0.7775,
+      "step": 10976
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.009320175438597e-06,
+      "loss": 0.5325,
+      "step": 10977
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.008771929824562e-06,
+      "loss": 0.4472,
+      "step": 10978
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.008223684210527e-06,
+      "loss": 1.8439,
+      "step": 10979
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.0076754385964914e-06,
+      "loss": 0.7025,
+      "step": 10980
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.007127192982456e-06,
+      "loss": 1.3364,
+      "step": 10981
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.006578947368421e-06,
+      "loss": 0.7507,
+      "step": 10982
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.006030701754387e-06,
+      "loss": 0.4256,
+      "step": 10983
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 4.005482456140351e-06,
+      "loss": 0.2631,
+      "step": 10984
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 4.004934210526316e-06,
+      "loss": 1.783,
+      "step": 10985
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 4.004385964912281e-06,
+      "loss": 1.7672,
+      "step": 10986
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 4.003837719298246e-06,
+      "loss": 0.6194,
+      "step": 10987
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 4.003289473684211e-06,
+      "loss": 0.6288,
+      "step": 10988
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 4.002741228070176e-06,
+      "loss": 0.2008,
+      "step": 10989
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 4.0021929824561405e-06,
+      "loss": 0.429,
+      "step": 10990
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 4.001644736842105e-06,
+      "loss": 0.2191,
+      "step": 10991
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 4.00109649122807e-06,
+      "loss": 0.38,
+      "step": 10992
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 4.000548245614036e-06,
+      "loss": 0.4109,
+      "step": 10993
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.3167,
+      "step": 10994
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.9994517543859655e-06,
+      "loss": 0.6781,
+      "step": 10995
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.99890350877193e-06,
+      "loss": 1.3187,
+      "step": 10996
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.998355263157895e-06,
+      "loss": 0.6156,
+      "step": 10997
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.99780701754386e-06,
+      "loss": 0.087,
+      "step": 10998
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.997258771929825e-06,
+      "loss": 1.3793,
+      "step": 10999
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.99671052631579e-06,
+      "loss": 0.1306,
+      "step": 11000
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.9961622807017544e-06,
+      "loss": 0.866,
+      "step": 11001
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.99561403508772e-06,
+      "loss": 0.2587,
+      "step": 11002
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.995065789473684e-06,
+      "loss": 0.3713,
+      "step": 11003
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.99451754385965e-06,
+      "loss": 0.4918,
+      "step": 11004
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.993969298245615e-06,
+      "loss": 0.7574,
+      "step": 11005
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.993421052631579e-06,
+      "loss": 0.2175,
+      "step": 11006
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.992872807017544e-06,
+      "loss": 0.4669,
+      "step": 11007
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.992324561403509e-06,
+      "loss": 0.9589,
+      "step": 11008
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.991776315789474e-06,
+      "loss": 0.7782,
+      "step": 11009
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.991228070175439e-06,
+      "loss": 0.7021,
+      "step": 11010
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.9906798245614035e-06,
+      "loss": 0.2371,
+      "step": 11011
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.990131578947369e-06,
+      "loss": 0.371,
+      "step": 11012
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.989583333333334e-06,
+      "loss": 0.7284,
+      "step": 11013
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.989035087719299e-06,
+      "loss": 0.5919,
+      "step": 11014
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.988486842105264e-06,
+      "loss": 0.9266,
+      "step": 11015
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.9879385964912285e-06,
+      "loss": 0.5891,
+      "step": 11016
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.987390350877193e-06,
+      "loss": 0.5092,
+      "step": 11017
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.986842105263158e-06,
+      "loss": 0.4743,
+      "step": 11018
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.986293859649123e-06,
+      "loss": 0.9971,
+      "step": 11019
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.985745614035088e-06,
+      "loss": 1.2872,
+      "step": 11020
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.9851973684210535e-06,
+      "loss": 0.4496,
+      "step": 11021
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.9846491228070175e-06,
+      "loss": 0.2961,
+      "step": 11022
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.984100877192983e-06,
+      "loss": 1.0595,
+      "step": 11023
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.983552631578947e-06,
+      "loss": 0.9049,
+      "step": 11024
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.983004385964913e-06,
+      "loss": 0.2679,
+      "step": 11025
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.982456140350878e-06,
+      "loss": 0.2938,
+      "step": 11026
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.981907894736842e-06,
+      "loss": 0.3034,
+      "step": 11027
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.981359649122807e-06,
+      "loss": 0.1277,
+      "step": 11028
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.980811403508772e-06,
+      "loss": 0.4916,
+      "step": 11029
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.980263157894737e-06,
+      "loss": 0.9484,
+      "step": 11030
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.9797149122807026e-06,
+      "loss": 0.7543,
+      "step": 11031
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.9791666666666665e-06,
+      "loss": 0.9028,
+      "step": 11032
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.978618421052632e-06,
+      "loss": 0.5548,
+      "step": 11033
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.978070175438597e-06,
+      "loss": 0.6621,
+      "step": 11034
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.977521929824562e-06,
+      "loss": 0.8176,
+      "step": 11035
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.976973684210527e-06,
+      "loss": 0.5657,
+      "step": 11036
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.9764254385964915e-06,
+      "loss": 0.9552,
+      "step": 11037
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.975877192982456e-06,
+      "loss": 0.3245,
+      "step": 11038
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.975328947368421e-06,
+      "loss": 0.432,
+      "step": 11039
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.974780701754386e-06,
+      "loss": 0.7374,
+      "step": 11040
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.974232456140351e-06,
+      "loss": 0.5698,
+      "step": 11041
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.9736842105263165e-06,
+      "loss": 0.5346,
+      "step": 11042
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.9731359649122805e-06,
+      "loss": 0.2038,
+      "step": 11043
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.972587719298246e-06,
+      "loss": 0.9221,
+      "step": 11044
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.972039473684211e-06,
+      "loss": 0.7745,
+      "step": 11045
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.971491228070176e-06,
+      "loss": 0.8945,
+      "step": 11046
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.970942982456141e-06,
+      "loss": 1.2481,
+      "step": 11047
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.970394736842106e-06,
+      "loss": 0.5418,
+      "step": 11048
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.96984649122807e-06,
+      "loss": 0.7864,
+      "step": 11049
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.969298245614036e-06,
+      "loss": 0.7104,
+      "step": 11050
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.96875e-06,
+      "loss": 0.1875,
+      "step": 11051
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.968201754385966e-06,
+      "loss": 0.491,
+      "step": 11052
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.96765350877193e-06,
+      "loss": 0.8111,
+      "step": 11053
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.967105263157895e-06,
+      "loss": 0.6083,
+      "step": 11054
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.96655701754386e-06,
+      "loss": 0.2122,
+      "step": 11055
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.966008771929825e-06,
+      "loss": 0.5611,
+      "step": 11056
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.96546052631579e-06,
+      "loss": 0.8745,
+      "step": 11057
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.9649122807017545e-06,
+      "loss": 0.2769,
+      "step": 11058
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.964364035087719e-06,
+      "loss": 0.2592,
+      "step": 11059
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.963815789473684e-06,
+      "loss": 0.6336,
+      "step": 11060
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.96326754385965e-06,
+      "loss": 0.7883,
+      "step": 11061
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.962719298245614e-06,
+      "loss": 0.5212,
+      "step": 11062
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.9621710526315795e-06,
+      "loss": 0.6344,
+      "step": 11063
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.961622807017544e-06,
+      "loss": 0.5952,
+      "step": 11064
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.961074561403509e-06,
+      "loss": 0.5663,
+      "step": 11065
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.960526315789474e-06,
+      "loss": 0.271,
+      "step": 11066
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.959978070175439e-06,
+      "loss": 0.7863,
+      "step": 11067
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.959429824561404e-06,
+      "loss": 1.1038,
+      "step": 11068
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.958881578947369e-06,
+      "loss": 0.2476,
+      "step": 11069
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.958333333333333e-06,
+      "loss": 0.8005,
+      "step": 11070
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.957785087719299e-06,
+      "loss": 0.2566,
+      "step": 11071
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.957236842105264e-06,
+      "loss": 0.9073,
+      "step": 11072
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.956688596491229e-06,
+      "loss": 0.2674,
+      "step": 11073
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.956140350877193e-06,
+      "loss": 0.4318,
+      "step": 11074
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.955592105263158e-06,
+      "loss": 1.0775,
+      "step": 11075
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.955043859649123e-06,
+      "loss": 0.3736,
+      "step": 11076
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.954495614035088e-06,
+      "loss": 0.8963,
+      "step": 11077
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.953947368421053e-06,
+      "loss": 0.3209,
+      "step": 11078
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.9533991228070175e-06,
+      "loss": 0.748,
+      "step": 11079
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.952850877192983e-06,
+      "loss": 0.6361,
+      "step": 11080
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.952302631578947e-06,
+      "loss": 0.0877,
+      "step": 11081
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.951754385964913e-06,
+      "loss": 0.6313,
+      "step": 11082
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.951206140350877e-06,
+      "loss": 0.9759,
+      "step": 11083
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.9506578947368425e-06,
+      "loss": 0.9125,
+      "step": 11084
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.950109649122807e-06,
+      "loss": 0.71,
+      "step": 11085
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.949561403508772e-06,
+      "loss": 0.3451,
+      "step": 11086
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.949013157894737e-06,
+      "loss": 0.5051,
+      "step": 11087
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.948464912280703e-06,
+      "loss": 0.7087,
+      "step": 11088
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.947916666666667e-06,
+      "loss": 0.1307,
+      "step": 11089
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.947368421052632e-06,
+      "loss": 0.8131,
+      "step": 11090
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.946820175438597e-06,
+      "loss": 0.7542,
+      "step": 11091
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.946271929824562e-06,
+      "loss": 0.1556,
+      "step": 11092
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.945723684210527e-06,
+      "loss": 1.0118,
+      "step": 11093
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.945175438596492e-06,
+      "loss": 0.2787,
+      "step": 11094
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.9446271929824564e-06,
+      "loss": 0.8447,
+      "step": 11095
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.944078947368421e-06,
+      "loss": 0.5677,
+      "step": 11096
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.943530701754386e-06,
+      "loss": 0.4867,
+      "step": 11097
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.942982456140351e-06,
+      "loss": 0.3417,
+      "step": 11098
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.9424342105263166e-06,
+      "loss": 0.6664,
+      "step": 11099
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.9418859649122805e-06,
+      "loss": 0.3238,
+      "step": 11100
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.941337719298246e-06,
+      "loss": 0.5383,
+      "step": 11101
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.94078947368421e-06,
+      "loss": 0.1733,
+      "step": 11102
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.940241228070176e-06,
+      "loss": 0.2097,
+      "step": 11103
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.939692982456141e-06,
+      "loss": 0.9879,
+      "step": 11104
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.9391447368421055e-06,
+      "loss": 0.6128,
+      "step": 11105
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.93859649122807e-06,
+      "loss": 0.5291,
+      "step": 11106
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.938048245614036e-06,
+      "loss": 0.8388,
+      "step": 11107
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.9375e-06,
+      "loss": 0.6997,
+      "step": 11108
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.936951754385966e-06,
+      "loss": 0.5826,
+      "step": 11109
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.93640350877193e-06,
+      "loss": 0.5643,
+      "step": 11110
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.935855263157895e-06,
+      "loss": 0.6353,
+      "step": 11111
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.93530701754386e-06,
+      "loss": 0.7334,
+      "step": 11112
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.934758771929825e-06,
+      "loss": 0.2998,
+      "step": 11113
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.93421052631579e-06,
+      "loss": 0.3861,
+      "step": 11114
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.933662280701755e-06,
+      "loss": 0.7033,
+      "step": 11115
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.9331140350877194e-06,
+      "loss": 0.1081,
+      "step": 11116
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.932565789473684e-06,
+      "loss": 0.7721,
+      "step": 11117
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.932017543859649e-06,
+      "loss": 0.1012,
+      "step": 11118
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.931469298245614e-06,
+      "loss": 0.4155,
+      "step": 11119
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.93092105263158e-06,
+      "loss": 0.3408,
+      "step": 11120
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.9303728070175436e-06,
+      "loss": 0.2716,
+      "step": 11121
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.929824561403509e-06,
+      "loss": 0.2551,
+      "step": 11122
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.929276315789474e-06,
+      "loss": 0.0798,
+      "step": 11123
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.928728070175439e-06,
+      "loss": 0.2077,
+      "step": 11124
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.928179824561404e-06,
+      "loss": 0.6359,
+      "step": 11125
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.927631578947369e-06,
+      "loss": 0.4207,
+      "step": 11126
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.927083333333333e-06,
+      "loss": 0.6568,
+      "step": 11127
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.926535087719299e-06,
+      "loss": 1.1265,
+      "step": 11128
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.925986842105263e-06,
+      "loss": 1.1492,
+      "step": 11129
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.925438596491229e-06,
+      "loss": 0.601,
+      "step": 11130
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.9248903508771935e-06,
+      "loss": 0.1404,
+      "step": 11131
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.924342105263158e-06,
+      "loss": 0.8011,
+      "step": 11132
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.923793859649123e-06,
+      "loss": 0.7479,
+      "step": 11133
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.923245614035088e-06,
+      "loss": 0.5998,
+      "step": 11134
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.922697368421053e-06,
+      "loss": 0.4322,
+      "step": 11135
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.922149122807018e-06,
+      "loss": 0.1981,
+      "step": 11136
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.9216008771929824e-06,
+      "loss": 0.086,
+      "step": 11137
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.921052631578947e-06,
+      "loss": 0.9421,
+      "step": 11138
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.920504385964913e-06,
+      "loss": 0.9625,
+      "step": 11139
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.919956140350878e-06,
+      "loss": 0.2589,
+      "step": 11140
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.919407894736843e-06,
+      "loss": 1.1275,
+      "step": 11141
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.918859649122807e-06,
+      "loss": 0.1868,
+      "step": 11142
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.918311403508772e-06,
+      "loss": 0.6712,
+      "step": 11143
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.917763157894737e-06,
+      "loss": 1.3777,
+      "step": 11144
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.917214912280702e-06,
+      "loss": 0.1915,
+      "step": 11145
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.916666666666667e-06,
+      "loss": 0.068,
+      "step": 11146
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.916118421052632e-06,
+      "loss": 0.2888,
+      "step": 11147
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.915570175438596e-06,
+      "loss": 0.3993,
+      "step": 11148
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.915021929824562e-06,
+      "loss": 0.1654,
+      "step": 11149
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.914473684210527e-06,
+      "loss": 0.1007,
+      "step": 11150
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.913925438596492e-06,
+      "loss": 0.5241,
+      "step": 11151
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.9133771929824565e-06,
+      "loss": 1.2913,
+      "step": 11152
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.912828947368421e-06,
+      "loss": 0.5399,
+      "step": 11153
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.912280701754386e-06,
+      "loss": 0.9706,
+      "step": 11154
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.911732456140351e-06,
+      "loss": 0.5571,
+      "step": 11155
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.911184210526316e-06,
+      "loss": 0.3233,
+      "step": 11156
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.910635964912281e-06,
+      "loss": 0.8241,
+      "step": 11157
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.910087719298246e-06,
+      "loss": 0.7574,
+      "step": 11158
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.909539473684211e-06,
+      "loss": 0.7137,
+      "step": 11159
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.908991228070176e-06,
+      "loss": 0.4645,
+      "step": 11160
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.908442982456141e-06,
+      "loss": 0.4232,
+      "step": 11161
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.907894736842106e-06,
+      "loss": 0.8667,
+      "step": 11162
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.9073464912280704e-06,
+      "loss": 0.2252,
+      "step": 11163
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.906798245614035e-06,
+      "loss": 1.18,
+      "step": 11164
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.90625e-06,
+      "loss": 0.9271,
+      "step": 11165
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.905701754385966e-06,
+      "loss": 0.9335,
+      "step": 11166
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.90515350877193e-06,
+      "loss": 0.8306,
+      "step": 11167
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.904605263157895e-06,
+      "loss": 0.5099,
+      "step": 11168
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.90405701754386e-06,
+      "loss": 1.7237,
+      "step": 11169
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.903508771929825e-06,
+      "loss": 0.4799,
+      "step": 11170
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.90296052631579e-06,
+      "loss": 0.9445,
+      "step": 11171
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.902412280701755e-06,
+      "loss": 0.7678,
+      "step": 11172
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.9018640350877195e-06,
+      "loss": 0.7111,
+      "step": 11173
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.901315789473684e-06,
+      "loss": 1.0275,
+      "step": 11174
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.900767543859649e-06,
+      "loss": 0.5631,
+      "step": 11175
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.900219298245614e-06,
+      "loss": 0.3082,
+      "step": 11176
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.89967105263158e-06,
+      "loss": 0.8979,
+      "step": 11177
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.8991228070175445e-06,
+      "loss": 0.3308,
+      "step": 11178
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.898574561403509e-06,
+      "loss": 0.3333,
+      "step": 11179
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.898026315789474e-06,
+      "loss": 0.5252,
+      "step": 11180
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.897478070175439e-06,
+      "loss": 0.941,
+      "step": 11181
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.896929824561404e-06,
+      "loss": 0.9565,
+      "step": 11182
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.896381578947369e-06,
+      "loss": 0.7972,
+      "step": 11183
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.8958333333333334e-06,
+      "loss": 0.5309,
+      "step": 11184
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.895285087719299e-06,
+      "loss": 0.0552,
+      "step": 11185
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.894736842105263e-06,
+      "loss": 0.5533,
+      "step": 11186
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.894188596491229e-06,
+      "loss": 0.6436,
+      "step": 11187
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.893640350877193e-06,
+      "loss": 0.1843,
+      "step": 11188
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.893092105263158e-06,
+      "loss": 0.3254,
+      "step": 11189
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.892543859649123e-06,
+      "loss": 1.2076,
+      "step": 11190
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.891995614035088e-06,
+      "loss": 0.5691,
+      "step": 11191
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.891447368421053e-06,
+      "loss": 1.0294,
+      "step": 11192
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.890899122807018e-06,
+      "loss": 0.1935,
+      "step": 11193
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.8903508771929825e-06,
+      "loss": 0.3531,
+      "step": 11194
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.889802631578947e-06,
+      "loss": 0.3993,
+      "step": 11195
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.889254385964912e-06,
+      "loss": 0.7739,
+      "step": 11196
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.888706140350878e-06,
+      "loss": 1.0519,
+      "step": 11197
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.888157894736843e-06,
+      "loss": 0.788,
+      "step": 11198
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.8876096491228075e-06,
+      "loss": 0.4089,
+      "step": 11199
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.887061403508772e-06,
+      "loss": 0.4423,
+      "step": 11200
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.886513157894737e-06,
+      "loss": 0.6628,
+      "step": 11201
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.885964912280702e-06,
+      "loss": 0.2096,
+      "step": 11202
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.885416666666667e-06,
+      "loss": 1.4461,
+      "step": 11203
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.884868421052632e-06,
+      "loss": 0.9692,
+      "step": 11204
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.8843201754385965e-06,
+      "loss": 0.4447,
+      "step": 11205
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.883771929824562e-06,
+      "loss": 1.3318,
+      "step": 11206
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.883223684210526e-06,
+      "loss": 0.1958,
+      "step": 11207
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.882675438596492e-06,
+      "loss": 0.3985,
+      "step": 11208
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.882127192982457e-06,
+      "loss": 1.0068,
+      "step": 11209
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.8815789473684214e-06,
+      "loss": 0.6101,
+      "step": 11210
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.881030701754386e-06,
+      "loss": 0.7292,
+      "step": 11211
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.880482456140351e-06,
+      "loss": 0.2432,
+      "step": 11212
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.879934210526316e-06,
+      "loss": 0.8627,
+      "step": 11213
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.879385964912281e-06,
+      "loss": 0.5538,
+      "step": 11214
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.8788377192982455e-06,
+      "loss": 0.6157,
+      "step": 11215
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.878289473684211e-06,
+      "loss": 0.3937,
+      "step": 11216
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.877741228070176e-06,
+      "loss": 0.7017,
+      "step": 11217
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.877192982456141e-06,
+      "loss": 0.5753,
+      "step": 11218
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.876644736842106e-06,
+      "loss": 0.198,
+      "step": 11219
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.8760964912280705e-06,
+      "loss": 0.927,
+      "step": 11220
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.875548245614035e-06,
+      "loss": 0.9528,
+      "step": 11221
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.875e-06,
+      "loss": 1.2689,
+      "step": 11222
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.874451754385965e-06,
+      "loss": 0.3517,
+      "step": 11223
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.87390350877193e-06,
+      "loss": 0.1198,
+      "step": 11224
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.8733552631578955e-06,
+      "loss": 0.2451,
+      "step": 11225
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.8728070175438595e-06,
+      "loss": 0.6577,
+      "step": 11226
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.872258771929825e-06,
+      "loss": 0.602,
+      "step": 11227
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.87171052631579e-06,
+      "loss": 1.0,
+      "step": 11228
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.871162280701755e-06,
+      "loss": 0.6513,
+      "step": 11229
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.87061403508772e-06,
+      "loss": 0.2529,
+      "step": 11230
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.8700657894736844e-06,
+      "loss": 0.2927,
+      "step": 11231
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.869517543859649e-06,
+      "loss": 0.5552,
+      "step": 11232
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.868969298245615e-06,
+      "loss": 0.6624,
+      "step": 11233
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.868421052631579e-06,
+      "loss": 0.2521,
+      "step": 11234
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.867872807017545e-06,
+      "loss": 0.881,
+      "step": 11235
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.867324561403509e-06,
+      "loss": 0.5657,
+      "step": 11236
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.866776315789474e-06,
+      "loss": 0.4317,
+      "step": 11237
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.866228070175439e-06,
+      "loss": 0.3964,
+      "step": 11238
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.865679824561404e-06,
+      "loss": 0.5793,
+      "step": 11239
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.865131578947369e-06,
+      "loss": 1.0088,
+      "step": 11240
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.8645833333333335e-06,
+      "loss": 0.2031,
+      "step": 11241
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.864035087719298e-06,
+      "loss": 0.6972,
+      "step": 11242
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.863486842105263e-06,
+      "loss": 0.3803,
+      "step": 11243
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.862938596491229e-06,
+      "loss": 0.6742,
+      "step": 11244
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.862390350877193e-06,
+      "loss": 1.5708,
+      "step": 11245
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.8618421052631585e-06,
+      "loss": 0.4145,
+      "step": 11246
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.861293859649123e-06,
+      "loss": 0.1776,
+      "step": 11247
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.860745614035088e-06,
+      "loss": 0.3123,
+      "step": 11248
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.860197368421053e-06,
+      "loss": 0.1583,
+      "step": 11249
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.859649122807018e-06,
+      "loss": 0.9921,
+      "step": 11250
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.859100877192983e-06,
+      "loss": 0.7885,
+      "step": 11251
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.858552631578948e-06,
+      "loss": 0.8453,
+      "step": 11252
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.858004385964912e-06,
+      "loss": 0.2751,
+      "step": 11253
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.857456140350878e-06,
+      "loss": 0.39,
+      "step": 11254
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.856907894736843e-06,
+      "loss": 0.3856,
+      "step": 11255
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.856359649122808e-06,
+      "loss": 0.1364,
+      "step": 11256
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.855811403508772e-06,
+      "loss": 0.5224,
+      "step": 11257
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.855263157894737e-06,
+      "loss": 0.9972,
+      "step": 11258
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.854714912280702e-06,
+      "loss": 0.2563,
+      "step": 11259
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.854166666666667e-06,
+      "loss": 0.6294,
+      "step": 11260
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.853618421052632e-06,
+      "loss": 0.4169,
+      "step": 11261
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.8530701754385965e-06,
+      "loss": 0.9346,
+      "step": 11262
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.852521929824562e-06,
+      "loss": 0.4031,
+      "step": 11263
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.851973684210526e-06,
+      "loss": 0.5155,
+      "step": 11264
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.851425438596492e-06,
+      "loss": 0.1343,
+      "step": 11265
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.850877192982456e-06,
+      "loss": 0.3706,
+      "step": 11266
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.8503289473684215e-06,
+      "loss": 0.6541,
+      "step": 11267
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.849780701754386e-06,
+      "loss": 0.9072,
+      "step": 11268
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.849232456140351e-06,
+      "loss": 0.2093,
+      "step": 11269
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.848684210526316e-06,
+      "loss": 0.3312,
+      "step": 11270
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.848135964912282e-06,
+      "loss": 0.6438,
+      "step": 11271
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.847587719298246e-06,
+      "loss": 0.9322,
+      "step": 11272
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.847039473684211e-06,
+      "loss": 0.7353,
+      "step": 11273
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.846491228070175e-06,
+      "loss": 1.6219,
+      "step": 11274
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.845942982456141e-06,
+      "loss": 0.1482,
+      "step": 11275
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.845394736842106e-06,
+      "loss": 0.8863,
+      "step": 11276
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.844846491228071e-06,
+      "loss": 0.2597,
+      "step": 11277
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.8442982456140354e-06,
+      "loss": 0.4808,
+      "step": 11278
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.84375e-06,
+      "loss": 0.6291,
+      "step": 11279
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.843201754385965e-06,
+      "loss": 0.5664,
+      "step": 11280
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.84265350877193e-06,
+      "loss": 0.2236,
+      "step": 11281
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.842105263157895e-06,
+      "loss": 0.5681,
+      "step": 11282
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.8415570175438596e-06,
+      "loss": 0.2817,
+      "step": 11283
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.841008771929825e-06,
+      "loss": 0.9928,
+      "step": 11284
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.840460526315789e-06,
+      "loss": 0.3829,
+      "step": 11285
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.839912280701755e-06,
+      "loss": 0.2648,
+      "step": 11286
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.83936403508772e-06,
+      "loss": 0.6,
+      "step": 11287
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.8388157894736845e-06,
+      "loss": 0.8149,
+      "step": 11288
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.838267543859649e-06,
+      "loss": 0.2529,
+      "step": 11289
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.837719298245615e-06,
+      "loss": 1.0812,
+      "step": 11290
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.837171052631579e-06,
+      "loss": 0.8075,
+      "step": 11291
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.836622807017545e-06,
+      "loss": 1.0197,
+      "step": 11292
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.836074561403509e-06,
+      "loss": 0.5473,
+      "step": 11293
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.835526315789474e-06,
+      "loss": 0.066,
+      "step": 11294
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.834978070175439e-06,
+      "loss": 1.1508,
+      "step": 11295
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.834429824561404e-06,
+      "loss": 0.5316,
+      "step": 11296
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.833881578947369e-06,
+      "loss": 0.1836,
+      "step": 11297
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.833333333333334e-06,
+      "loss": 0.4967,
+      "step": 11298
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.8327850877192984e-06,
+      "loss": 0.4064,
+      "step": 11299
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.832236842105263e-06,
+      "loss": 0.8381,
+      "step": 11300
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.831688596491228e-06,
+      "loss": 2.0989,
+      "step": 11301
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.831140350877193e-06,
+      "loss": 0.4426,
+      "step": 11302
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.830592105263159e-06,
+      "loss": 0.6905,
+      "step": 11303
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.8300438596491226e-06,
+      "loss": 0.5915,
+      "step": 11304
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.829495614035088e-06,
+      "loss": 0.9054,
+      "step": 11305
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.828947368421053e-06,
+      "loss": 0.5429,
+      "step": 11306
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.828399122807018e-06,
+      "loss": 0.9744,
+      "step": 11307
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.827850877192983e-06,
+      "loss": 0.52,
+      "step": 11308
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.8273026315789475e-06,
+      "loss": 0.4588,
+      "step": 11309
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.826754385964912e-06,
+      "loss": 0.8781,
+      "step": 11310
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.826206140350878e-06,
+      "loss": 1.8457,
+      "step": 11311
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.825657894736842e-06,
+      "loss": 0.99,
+      "step": 11312
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.825109649122808e-06,
+      "loss": 0.4016,
+      "step": 11313
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.8245614035087725e-06,
+      "loss": 1.0849,
+      "step": 11314
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.824013157894737e-06,
+      "loss": 0.4392,
+      "step": 11315
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.823464912280702e-06,
+      "loss": 0.2277,
+      "step": 11316
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.822916666666667e-06,
+      "loss": 0.6822,
+      "step": 11317
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.822368421052632e-06,
+      "loss": 0.6439,
+      "step": 11318
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.821820175438597e-06,
+      "loss": 0.7483,
+      "step": 11319
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.8212719298245615e-06,
+      "loss": 1.2024,
+      "step": 11320
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.820723684210526e-06,
+      "loss": 0.7823,
+      "step": 11321
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.820175438596492e-06,
+      "loss": 0.8991,
+      "step": 11322
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.819627192982456e-06,
+      "loss": 0.3017,
+      "step": 11323
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.819078947368422e-06,
+      "loss": 0.48,
+      "step": 11324
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.8185307017543864e-06,
+      "loss": 0.2674,
+      "step": 11325
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.817982456140351e-06,
+      "loss": 1.0045,
+      "step": 11326
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.817434210526316e-06,
+      "loss": 0.7838,
+      "step": 11327
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.816885964912281e-06,
+      "loss": 0.0781,
+      "step": 11328
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.816337719298246e-06,
+      "loss": 0.7374,
+      "step": 11329
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.815789473684211e-06,
+      "loss": 1.4537,
+      "step": 11330
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.815241228070175e-06,
+      "loss": 0.4581,
+      "step": 11331
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.8146929824561406e-06,
+      "loss": 0.6712,
+      "step": 11332
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.814144736842106e-06,
+      "loss": 0.5111,
+      "step": 11333
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.8135964912280703e-06,
+      "loss": 0.3049,
+      "step": 11334
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.8130482456140355e-06,
+      "loss": 0.7727,
+      "step": 11335
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.8125e-06,
+      "loss": 0.2359,
+      "step": 11336
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.811951754385965e-06,
+      "loss": 0.3547,
+      "step": 11337
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.8114035087719304e-06,
+      "loss": 0.282,
+      "step": 11338
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.810855263157895e-06,
+      "loss": 0.2178,
+      "step": 11339
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.81030701754386e-06,
+      "loss": 0.9251,
+      "step": 11340
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.8097587719298253e-06,
+      "loss": 0.3211,
+      "step": 11341
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.8092105263157897e-06,
+      "loss": 0.5301,
+      "step": 11342
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.808662280701755e-06,
+      "loss": 0.4077,
+      "step": 11343
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.8081140350877194e-06,
+      "loss": 0.4998,
+      "step": 11344
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.8075657894736846e-06,
+      "loss": 0.3661,
+      "step": 11345
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.8070175438596494e-06,
+      "loss": 0.5151,
+      "step": 11346
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.8064692982456143e-06,
+      "loss": 0.2265,
+      "step": 11347
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.805921052631579e-06,
+      "loss": 0.4469,
+      "step": 11348
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.8053728070175443e-06,
+      "loss": 0.8504,
+      "step": 11349
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.804824561403509e-06,
+      "loss": 0.3784,
+      "step": 11350
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.804276315789474e-06,
+      "loss": 0.4436,
+      "step": 11351
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.803728070175439e-06,
+      "loss": 0.3431,
+      "step": 11352
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.8031798245614036e-06,
+      "loss": 0.2832,
+      "step": 11353
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.802631578947369e-06,
+      "loss": 0.8499,
+      "step": 11354
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.8020833333333333e-06,
+      "loss": 0.6626,
+      "step": 11355
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.8015350877192985e-06,
+      "loss": 0.4017,
+      "step": 11356
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.8009868421052638e-06,
+      "loss": 0.5883,
+      "step": 11357
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.800438596491228e-06,
+      "loss": 0.3817,
+      "step": 11358
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.7998903508771934e-06,
+      "loss": 0.506,
+      "step": 11359
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.799342105263158e-06,
+      "loss": 0.9645,
+      "step": 11360
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.798793859649123e-06,
+      "loss": 0.4929,
+      "step": 11361
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.7982456140350883e-06,
+      "loss": 0.1759,
+      "step": 11362
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.7976973684210527e-06,
+      "loss": 0.2515,
+      "step": 11363
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.797149122807018e-06,
+      "loss": 0.3134,
+      "step": 11364
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.796600877192983e-06,
+      "loss": 0.5817,
+      "step": 11365
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.7960526315789476e-06,
+      "loss": 0.5299,
+      "step": 11366
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.795504385964913e-06,
+      "loss": 0.269,
+      "step": 11367
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.7949561403508773e-06,
+      "loss": 0.7067,
+      "step": 11368
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.7944078947368425e-06,
+      "loss": 0.704,
+      "step": 11369
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.7938596491228073e-06,
+      "loss": 0.8399,
+      "step": 11370
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.793311403508772e-06,
+      "loss": 0.2237,
+      "step": 11371
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.792763157894737e-06,
+      "loss": 0.1734,
+      "step": 11372
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.7922149122807022e-06,
+      "loss": 0.8282,
+      "step": 11373
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.7916666666666666e-06,
+      "loss": 0.622,
+      "step": 11374
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.791118421052632e-06,
+      "loss": 0.6994,
+      "step": 11375
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.790570175438597e-06,
+      "loss": 0.4901,
+      "step": 11376
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.7900219298245615e-06,
+      "loss": 0.4678,
+      "step": 11377
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.789473684210527e-06,
+      "loss": 1.1669,
+      "step": 11378
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.788925438596491e-06,
+      "loss": 0.8144,
+      "step": 11379
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.7883771929824564e-06,
+      "loss": 0.8276,
+      "step": 11380
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.7878289473684217e-06,
+      "loss": 0.9156,
+      "step": 11381
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.787280701754386e-06,
+      "loss": 0.0788,
+      "step": 11382
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.7867324561403513e-06,
+      "loss": 0.5627,
+      "step": 11383
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.786184210526316e-06,
+      "loss": 0.7755,
+      "step": 11384
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.785635964912281e-06,
+      "loss": 0.0665,
+      "step": 11385
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.7850877192982462e-06,
+      "loss": 0.4584,
+      "step": 11386
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.7845394736842106e-06,
+      "loss": 0.1738,
+      "step": 11387
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.783991228070176e-06,
+      "loss": 0.3421,
+      "step": 11388
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.7834429824561407e-06,
+      "loss": 0.7799,
+      "step": 11389
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.7828947368421055e-06,
+      "loss": 0.5262,
+      "step": 11390
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.7823464912280704e-06,
+      "loss": 0.6169,
+      "step": 11391
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.7817982456140356e-06,
+      "loss": 0.3205,
+      "step": 11392
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.78125e-06,
+      "loss": 0.3423,
+      "step": 11393
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.7807017543859653e-06,
+      "loss": 1.346,
+      "step": 11394
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.78015350877193e-06,
+      "loss": 0.8971,
+      "step": 11395
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.779605263157895e-06,
+      "loss": 0.8371,
+      "step": 11396
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.77905701754386e-06,
+      "loss": 0.428,
+      "step": 11397
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.7785087719298246e-06,
+      "loss": 0.7977,
+      "step": 11398
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.77796052631579e-06,
+      "loss": 0.732,
+      "step": 11399
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.777412280701755e-06,
+      "loss": 0.0874,
+      "step": 11400
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.7768640350877195e-06,
+      "loss": 0.4284,
+      "step": 11401
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.7763157894736847e-06,
+      "loss": 0.4547,
+      "step": 11402
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.775767543859649e-06,
+      "loss": 0.8272,
+      "step": 11403
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.7752192982456143e-06,
+      "loss": 0.8469,
+      "step": 11404
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.7746710526315796e-06,
+      "loss": 0.082,
+      "step": 11405
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.774122807017544e-06,
+      "loss": 0.8158,
+      "step": 11406
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.7735745614035092e-06,
+      "loss": 1.2106,
+      "step": 11407
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.773026315789474e-06,
+      "loss": 0.5477,
+      "step": 11408
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.772478070175439e-06,
+      "loss": 1.1756,
+      "step": 11409
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.7719298245614037e-06,
+      "loss": 0.1377,
+      "step": 11410
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.771381578947369e-06,
+      "loss": 0.6073,
+      "step": 11411
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.7708333333333334e-06,
+      "loss": 0.9354,
+      "step": 11412
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.7702850877192986e-06,
+      "loss": 0.0904,
+      "step": 11413
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.7697368421052634e-06,
+      "loss": 0.2912,
+      "step": 11414
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.7691885964912283e-06,
+      "loss": 0.5557,
+      "step": 11415
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.7686403508771935e-06,
+      "loss": 0.6424,
+      "step": 11416
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.768092105263158e-06,
+      "loss": 0.3047,
+      "step": 11417
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.767543859649123e-06,
+      "loss": 0.1423,
+      "step": 11418
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.7669956140350884e-06,
+      "loss": 0.087,
+      "step": 11419
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.766447368421053e-06,
+      "loss": 0.8317,
+      "step": 11420
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.765899122807018e-06,
+      "loss": 0.2367,
+      "step": 11421
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.7653508771929825e-06,
+      "loss": 0.5132,
+      "step": 11422
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.7648026315789477e-06,
+      "loss": 0.7667,
+      "step": 11423
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.764254385964913e-06,
+      "loss": 0.9293,
+      "step": 11424
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.7637061403508774e-06,
+      "loss": 0.9119,
+      "step": 11425
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.7631578947368426e-06,
+      "loss": 0.441,
+      "step": 11426
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.7626096491228074e-06,
+      "loss": 0.4359,
+      "step": 11427
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.7620614035087723e-06,
+      "loss": 0.8361,
+      "step": 11428
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.761513157894737e-06,
+      "loss": 0.3937,
+      "step": 11429
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.760964912280702e-06,
+      "loss": 0.5886,
+      "step": 11430
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.7604166666666667e-06,
+      "loss": 1.0942,
+      "step": 11431
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.759868421052632e-06,
+      "loss": 0.4057,
+      "step": 11432
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.759320175438597e-06,
+      "loss": 0.5162,
+      "step": 11433
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.7587719298245616e-06,
+      "loss": 0.6062,
+      "step": 11434
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.758223684210527e-06,
+      "loss": 0.3192,
+      "step": 11435
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.7576754385964913e-06,
+      "loss": 0.2968,
+      "step": 11436
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.7571271929824565e-06,
+      "loss": 0.7647,
+      "step": 11437
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.756578947368421e-06,
+      "loss": 0.5674,
+      "step": 11438
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.756030701754386e-06,
+      "loss": 0.5168,
+      "step": 11439
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.7554824561403514e-06,
+      "loss": 0.6641,
+      "step": 11440
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.754934210526316e-06,
+      "loss": 0.3098,
+      "step": 11441
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.754385964912281e-06,
+      "loss": 0.999,
+      "step": 11442
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.7538377192982463e-06,
+      "loss": 0.6528,
+      "step": 11443
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.7532894736842107e-06,
+      "loss": 0.3432,
+      "step": 11444
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.752741228070176e-06,
+      "loss": 0.2884,
+      "step": 11445
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.7521929824561404e-06,
+      "loss": 0.3926,
+      "step": 11446
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.7516447368421056e-06,
+      "loss": 0.4922,
+      "step": 11447
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.7510964912280704e-06,
+      "loss": 0.7226,
+      "step": 11448
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.7505482456140353e-06,
+      "loss": 0.213,
+      "step": 11449
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.7500000000000005e-06,
+      "loss": 0.5196,
+      "step": 11450
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.7494517543859653e-06,
+      "loss": 0.91,
+      "step": 11451
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.74890350877193e-06,
+      "loss": 1.0012,
+      "step": 11452
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.748355263157895e-06,
+      "loss": 0.3479,
+      "step": 11453
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.7478070175438602e-06,
+      "loss": 0.8682,
+      "step": 11454
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.7472587719298246e-06,
+      "loss": 0.4167,
+      "step": 11455
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.74671052631579e-06,
+      "loss": 0.5334,
+      "step": 11456
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.7461622807017543e-06,
+      "loss": 0.9891,
+      "step": 11457
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.7456140350877195e-06,
+      "loss": 0.4043,
+      "step": 11458
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.7450657894736848e-06,
+      "loss": 0.5908,
+      "step": 11459
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.744517543859649e-06,
+      "loss": 0.4542,
+      "step": 11460
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.7439692982456144e-06,
+      "loss": 0.4669,
+      "step": 11461
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.7434210526315797e-06,
+      "loss": 0.3807,
+      "step": 11462
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.742872807017544e-06,
+      "loss": 0.5094,
+      "step": 11463
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.7423245614035093e-06,
+      "loss": 0.4008,
+      "step": 11464
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.7417763157894737e-06,
+      "loss": 1.1534,
+      "step": 11465
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.741228070175439e-06,
+      "loss": 0.1478,
+      "step": 11466
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.740679824561404e-06,
+      "loss": 0.4691,
+      "step": 11467
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.7401315789473686e-06,
+      "loss": 0.7657,
+      "step": 11468
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.739583333333334e-06,
+      "loss": 0.4417,
+      "step": 11469
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.7390350877192987e-06,
+      "loss": 0.5137,
+      "step": 11470
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.7384868421052635e-06,
+      "loss": 0.6135,
+      "step": 11471
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.7379385964912284e-06,
+      "loss": 0.165,
+      "step": 11472
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.737390350877193e-06,
+      "loss": 0.9694,
+      "step": 11473
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.736842105263158e-06,
+      "loss": 0.477,
+      "step": 11474
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.7362938596491233e-06,
+      "loss": 0.7808,
+      "step": 11475
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.7357456140350877e-06,
+      "loss": 1.1181,
+      "step": 11476
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.735197368421053e-06,
+      "loss": 0.8055,
+      "step": 11477
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.734649122807018e-06,
+      "loss": 0.4755,
+      "step": 11478
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.7341008771929826e-06,
+      "loss": 0.6739,
+      "step": 11479
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.733552631578948e-06,
+      "loss": 0.6541,
+      "step": 11480
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.733004385964912e-06,
+      "loss": 0.5007,
+      "step": 11481
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.7324561403508774e-06,
+      "loss": 0.9112,
+      "step": 11482
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.7319078947368427e-06,
+      "loss": 0.5566,
+      "step": 11483
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.731359649122807e-06,
+      "loss": 0.6009,
+      "step": 11484
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.7308114035087723e-06,
+      "loss": 1.8272,
+      "step": 11485
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.730263157894737e-06,
+      "loss": 0.2381,
+      "step": 11486
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.729714912280702e-06,
+      "loss": 0.4972,
+      "step": 11487
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.7291666666666672e-06,
+      "loss": 0.2787,
+      "step": 11488
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.7286184210526316e-06,
+      "loss": 0.869,
+      "step": 11489
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.728070175438597e-06,
+      "loss": 0.4711,
+      "step": 11490
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.7275219298245617e-06,
+      "loss": 0.4717,
+      "step": 11491
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.7269736842105265e-06,
+      "loss": 0.5823,
+      "step": 11492
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.7264254385964914e-06,
+      "loss": 0.8116,
+      "step": 11493
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.7258771929824566e-06,
+      "loss": 0.5865,
+      "step": 11494
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.725328947368421e-06,
+      "loss": 0.3344,
+      "step": 11495
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.7247807017543863e-06,
+      "loss": 0.8611,
+      "step": 11496
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.7242324561403515e-06,
+      "loss": 0.1657,
+      "step": 11497
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.723684210526316e-06,
+      "loss": 0.4989,
+      "step": 11498
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.723135964912281e-06,
+      "loss": 0.6888,
+      "step": 11499
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.7225877192982456e-06,
+      "loss": 0.7784,
+      "step": 11500
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.722039473684211e-06,
+      "loss": 0.7009,
+      "step": 11501
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.721491228070176e-06,
+      "loss": 0.5907,
+      "step": 11502
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.7209429824561405e-06,
+      "loss": 0.3792,
+      "step": 11503
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.7203947368421057e-06,
+      "loss": 0.5036,
+      "step": 11504
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.7198464912280705e-06,
+      "loss": 0.3556,
+      "step": 11505
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.7192982456140354e-06,
+      "loss": 0.3224,
+      "step": 11506
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.7187500000000006e-06,
+      "loss": 0.7036,
+      "step": 11507
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.718201754385965e-06,
+      "loss": 0.5014,
+      "step": 11508
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.7176535087719303e-06,
+      "loss": 0.7002,
+      "step": 11509
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.717105263157895e-06,
+      "loss": 0.5167,
+      "step": 11510
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.71655701754386e-06,
+      "loss": 0.6621,
+      "step": 11511
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.7160087719298247e-06,
+      "loss": 0.4365,
+      "step": 11512
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.71546052631579e-06,
+      "loss": 0.5998,
+      "step": 11513
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.7149122807017544e-06,
+      "loss": 0.213,
+      "step": 11514
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.7143640350877196e-06,
+      "loss": 1.232,
+      "step": 11515
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.7138157894736845e-06,
+      "loss": 0.5732,
+      "step": 11516
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.7132675438596493e-06,
+      "loss": 1.3962,
+      "step": 11517
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.7127192982456145e-06,
+      "loss": 0.5183,
+      "step": 11518
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.712171052631579e-06,
+      "loss": 0.2405,
+      "step": 11519
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.711622807017544e-06,
+      "loss": 0.6273,
+      "step": 11520
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.7110745614035094e-06,
+      "loss": 0.3988,
+      "step": 11521
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.710526315789474e-06,
+      "loss": 0.4229,
+      "step": 11522
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.709978070175439e-06,
+      "loss": 0.3259,
+      "step": 11523
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.7094298245614035e-06,
+      "loss": 0.2188,
+      "step": 11524
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.7088815789473687e-06,
+      "loss": 1.028,
+      "step": 11525
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.708333333333334e-06,
+      "loss": 0.4235,
+      "step": 11526
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.7077850877192984e-06,
+      "loss": 0.7896,
+      "step": 11527
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.7072368421052636e-06,
+      "loss": 1.2767,
+      "step": 11528
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.7066885964912284e-06,
+      "loss": 0.0776,
+      "step": 11529
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.7061403508771933e-06,
+      "loss": 0.5303,
+      "step": 11530
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.705592105263158e-06,
+      "loss": 0.3382,
+      "step": 11531
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.7050438596491233e-06,
+      "loss": 0.1309,
+      "step": 11532
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.7044956140350877e-06,
+      "loss": 1.1249,
+      "step": 11533
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.703947368421053e-06,
+      "loss": 0.3409,
+      "step": 11534
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.703399122807018e-06,
+      "loss": 0.2567,
+      "step": 11535
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.7028508771929826e-06,
+      "loss": 0.9742,
+      "step": 11536
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.702302631578948e-06,
+      "loss": 0.577,
+      "step": 11537
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.7017543859649123e-06,
+      "loss": 0.383,
+      "step": 11538
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.7012061403508775e-06,
+      "loss": 0.5043,
+      "step": 11539
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.7006578947368428e-06,
+      "loss": 0.5021,
+      "step": 11540
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.700109649122807e-06,
+      "loss": 0.2759,
+      "step": 11541
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.6995614035087724e-06,
+      "loss": 0.4742,
+      "step": 11542
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.699013157894737e-06,
+      "loss": 0.3381,
+      "step": 11543
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.698464912280702e-06,
+      "loss": 1.0273,
+      "step": 11544
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.6979166666666673e-06,
+      "loss": 1.1423,
+      "step": 11545
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.6973684210526317e-06,
+      "loss": 0.5378,
+      "step": 11546
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.696820175438597e-06,
+      "loss": 0.7895,
+      "step": 11547
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.696271929824562e-06,
+      "loss": 0.3394,
+      "step": 11548
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.6957236842105266e-06,
+      "loss": 0.3952,
+      "step": 11549
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.6951754385964915e-06,
+      "loss": 1.0832,
+      "step": 11550
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.6946271929824563e-06,
+      "loss": 0.698,
+      "step": 11551
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.6940789473684215e-06,
+      "loss": 1.0965,
+      "step": 11552
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.6935307017543864e-06,
+      "loss": 0.9057,
+      "step": 11553
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.692982456140351e-06,
+      "loss": 0.3974,
+      "step": 11554
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.692434210526316e-06,
+      "loss": 0.5456,
+      "step": 11555
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.6918859649122812e-06,
+      "loss": 1.0814,
+      "step": 11556
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.6913377192982457e-06,
+      "loss": 0.3289,
+      "step": 11557
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.690789473684211e-06,
+      "loss": 0.5935,
+      "step": 11558
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.6902412280701753e-06,
+      "loss": 1.2783,
+      "step": 11559
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.6896929824561405e-06,
+      "loss": 0.2486,
+      "step": 11560
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.689144736842106e-06,
+      "loss": 0.6441,
+      "step": 11561
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.68859649122807e-06,
+      "loss": 0.4555,
+      "step": 11562
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.6880482456140354e-06,
+      "loss": 0.4002,
+      "step": 11563
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.6875000000000007e-06,
+      "loss": 0.3947,
+      "step": 11564
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.686951754385965e-06,
+      "loss": 0.2153,
+      "step": 11565
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.6864035087719303e-06,
+      "loss": 1.0447,
+      "step": 11566
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.6858552631578947e-06,
+      "loss": 0.779,
+      "step": 11567
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.68530701754386e-06,
+      "loss": 0.3382,
+      "step": 11568
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.684758771929825e-06,
+      "loss": 0.2253,
+      "step": 11569
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.6842105263157896e-06,
+      "loss": 1.8555,
+      "step": 11570
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.683662280701755e-06,
+      "loss": 0.6432,
+      "step": 11571
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.6831140350877197e-06,
+      "loss": 0.7635,
+      "step": 11572
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.6825657894736845e-06,
+      "loss": 0.4775,
+      "step": 11573
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.6820175438596494e-06,
+      "loss": 0.773,
+      "step": 11574
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.6814692982456146e-06,
+      "loss": 0.9208,
+      "step": 11575
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.680921052631579e-06,
+      "loss": 1.6548,
+      "step": 11576
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.6803728070175443e-06,
+      "loss": 0.4902,
+      "step": 11577
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.6798245614035087e-06,
+      "loss": 1.0379,
+      "step": 11578
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.679276315789474e-06,
+      "loss": 0.3545,
+      "step": 11579
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.678728070175439e-06,
+      "loss": 0.7529,
+      "step": 11580
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.6781798245614036e-06,
+      "loss": 1.1977,
+      "step": 11581
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.677631578947369e-06,
+      "loss": 0.536,
+      "step": 11582
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.677083333333334e-06,
+      "loss": 0.8873,
+      "step": 11583
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.6765350877192985e-06,
+      "loss": 0.2869,
+      "step": 11584
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.6759868421052637e-06,
+      "loss": 0.2695,
+      "step": 11585
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.675438596491228e-06,
+      "loss": 0.4148,
+      "step": 11586
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.6748903508771934e-06,
+      "loss": 0.5772,
+      "step": 11587
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.674342105263158e-06,
+      "loss": 0.576,
+      "step": 11588
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.673793859649123e-06,
+      "loss": 1.9874,
+      "step": 11589
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.6732456140350883e-06,
+      "loss": 0.622,
+      "step": 11590
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.672697368421053e-06,
+      "loss": 0.9255,
+      "step": 11591
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.672149122807018e-06,
+      "loss": 0.566,
+      "step": 11592
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.6716008771929827e-06,
+      "loss": 0.5901,
+      "step": 11593
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.6710526315789476e-06,
+      "loss": 0.1671,
+      "step": 11594
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.6705043859649124e-06,
+      "loss": 0.9816,
+      "step": 11595
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.6699561403508776e-06,
+      "loss": 1.3129,
+      "step": 11596
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.669407894736842e-06,
+      "loss": 0.412,
+      "step": 11597
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.6688596491228073e-06,
+      "loss": 0.8971,
+      "step": 11598
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.6683114035087725e-06,
+      "loss": 0.5647,
+      "step": 11599
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.667763157894737e-06,
+      "loss": 0.4323,
+      "step": 11600
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.667214912280702e-06,
+      "loss": 0.2845,
+      "step": 11601
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.6666666666666666e-06,
+      "loss": 0.9266,
+      "step": 11602
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.666118421052632e-06,
+      "loss": 1.164,
+      "step": 11603
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.665570175438597e-06,
+      "loss": 0.3318,
+      "step": 11604
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.6650219298245615e-06,
+      "loss": 0.4158,
+      "step": 11605
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.6644736842105267e-06,
+      "loss": 0.4952,
+      "step": 11606
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.6639254385964915e-06,
+      "loss": 0.7582,
+      "step": 11607
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.6633771929824564e-06,
+      "loss": 0.5961,
+      "step": 11608
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.6628289473684216e-06,
+      "loss": 0.5544,
+      "step": 11609
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.662280701754386e-06,
+      "loss": 0.9732,
+      "step": 11610
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.6617324561403513e-06,
+      "loss": 0.0996,
+      "step": 11611
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.661184210526316e-06,
+      "loss": 0.5244,
+      "step": 11612
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.660635964912281e-06,
+      "loss": 0.1421,
+      "step": 11613
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.6600877192982457e-06,
+      "loss": 2.0485,
+      "step": 11614
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.659539473684211e-06,
+      "loss": 0.7915,
+      "step": 11615
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.6589912280701754e-06,
+      "loss": 0.3448,
+      "step": 11616
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.6584429824561406e-06,
+      "loss": 0.6664,
+      "step": 11617
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.657894736842106e-06,
+      "loss": 0.7154,
+      "step": 11618
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.6573464912280703e-06,
+      "loss": 0.5795,
+      "step": 11619
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.6567982456140355e-06,
+      "loss": 0.4897,
+      "step": 11620
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.65625e-06,
+      "loss": 0.7208,
+      "step": 11621
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.655701754385965e-06,
+      "loss": 0.7583,
+      "step": 11622
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.6551535087719304e-06,
+      "loss": 0.3701,
+      "step": 11623
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.654605263157895e-06,
+      "loss": 0.7681,
+      "step": 11624
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.65405701754386e-06,
+      "loss": 0.6454,
+      "step": 11625
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.653508771929825e-06,
+      "loss": 0.6011,
+      "step": 11626
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.6529605263157897e-06,
+      "loss": 0.8595,
+      "step": 11627
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.652412280701755e-06,
+      "loss": 0.194,
+      "step": 11628
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.6518640350877194e-06,
+      "loss": 0.8983,
+      "step": 11629
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.6513157894736846e-06,
+      "loss": 0.6075,
+      "step": 11630
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.6507675438596495e-06,
+      "loss": 0.7802,
+      "step": 11631
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.6502192982456143e-06,
+      "loss": 0.4512,
+      "step": 11632
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.649671052631579e-06,
+      "loss": 0.6677,
+      "step": 11633
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.6491228070175443e-06,
+      "loss": 0.5788,
+      "step": 11634
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.648574561403509e-06,
+      "loss": 0.2357,
+      "step": 11635
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.648026315789474e-06,
+      "loss": 0.7725,
+      "step": 11636
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.647478070175439e-06,
+      "loss": 0.6321,
+      "step": 11637
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.6469298245614036e-06,
+      "loss": 0.4054,
+      "step": 11638
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.646381578947369e-06,
+      "loss": 0.2435,
+      "step": 11639
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.6458333333333333e-06,
+      "loss": 0.3374,
+      "step": 11640
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.6452850877192985e-06,
+      "loss": 0.2951,
+      "step": 11641
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.644736842105264e-06,
+      "loss": 0.4848,
+      "step": 11642
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.644188596491228e-06,
+      "loss": 0.7094,
+      "step": 11643
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.6436403508771934e-06,
+      "loss": 0.2918,
+      "step": 11644
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.643092105263158e-06,
+      "loss": 0.4004,
+      "step": 11645
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.642543859649123e-06,
+      "loss": 0.2884,
+      "step": 11646
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.6419956140350883e-06,
+      "loss": 0.6075,
+      "step": 11647
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.6414473684210527e-06,
+      "loss": 0.0624,
+      "step": 11648
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.640899122807018e-06,
+      "loss": 0.7344,
+      "step": 11649
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.640350877192983e-06,
+      "loss": 0.0774,
+      "step": 11650
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.6398026315789476e-06,
+      "loss": 0.8304,
+      "step": 11651
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.6392543859649125e-06,
+      "loss": 0.5236,
+      "step": 11652
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.6387061403508773e-06,
+      "loss": 0.559,
+      "step": 11653
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.6381578947368425e-06,
+      "loss": 0.2863,
+      "step": 11654
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.6376096491228074e-06,
+      "loss": 0.735,
+      "step": 11655
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.637061403508772e-06,
+      "loss": 0.2035,
+      "step": 11656
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.636513157894737e-06,
+      "loss": 0.8135,
+      "step": 11657
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.6359649122807023e-06,
+      "loss": 0.5986,
+      "step": 11658
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.6354166666666667e-06,
+      "loss": 0.1183,
+      "step": 11659
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.634868421052632e-06,
+      "loss": 0.2994,
+      "step": 11660
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.634320175438597e-06,
+      "loss": 1.4379,
+      "step": 11661
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.6337719298245616e-06,
+      "loss": 0.1961,
+      "step": 11662
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.633223684210527e-06,
+      "loss": 1.0845,
+      "step": 11663
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.632675438596491e-06,
+      "loss": 0.6496,
+      "step": 11664
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.6321271929824565e-06,
+      "loss": 0.7018,
+      "step": 11665
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.6315789473684217e-06,
+      "loss": 0.5675,
+      "step": 11666
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.631030701754386e-06,
+      "loss": 0.5644,
+      "step": 11667
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.6304824561403514e-06,
+      "loss": 0.5265,
+      "step": 11668
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.629934210526316e-06,
+      "loss": 0.3763,
+      "step": 11669
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.629385964912281e-06,
+      "loss": 0.3222,
+      "step": 11670
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.628837719298246e-06,
+      "loss": 0.21,
+      "step": 11671
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.6282894736842106e-06,
+      "loss": 0.3419,
+      "step": 11672
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.627741228070176e-06,
+      "loss": 0.2952,
+      "step": 11673
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.6271929824561407e-06,
+      "loss": 0.8858,
+      "step": 11674
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.6266447368421055e-06,
+      "loss": 1.2827,
+      "step": 11675
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.6260964912280704e-06,
+      "loss": 0.43,
+      "step": 11676
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.6255482456140356e-06,
+      "loss": 1.0533,
+      "step": 11677
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.625e-06,
+      "loss": 0.8688,
+      "step": 11678
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.6244517543859653e-06,
+      "loss": 0.1219,
+      "step": 11679
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.6239035087719297e-06,
+      "loss": 0.6495,
+      "step": 11680
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.623355263157895e-06,
+      "loss": 0.5221,
+      "step": 11681
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.62280701754386e-06,
+      "loss": 0.3063,
+      "step": 11682
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.6222587719298246e-06,
+      "loss": 0.8831,
+      "step": 11683
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.62171052631579e-06,
+      "loss": 0.366,
+      "step": 11684
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.621162280701755e-06,
+      "loss": 0.811,
+      "step": 11685
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.6206140350877195e-06,
+      "loss": 0.4248,
+      "step": 11686
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.6200657894736847e-06,
+      "loss": 0.2863,
+      "step": 11687
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.619517543859649e-06,
+      "loss": 0.9446,
+      "step": 11688
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.6189692982456144e-06,
+      "loss": 0.381,
+      "step": 11689
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.618421052631579e-06,
+      "loss": 0.7072,
+      "step": 11690
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.617872807017544e-06,
+      "loss": 1.2082,
+      "step": 11691
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.6173245614035093e-06,
+      "loss": 0.345,
+      "step": 11692
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.616776315789474e-06,
+      "loss": 0.0467,
+      "step": 11693
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.616228070175439e-06,
+      "loss": 0.9377,
+      "step": 11694
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.6156798245614037e-06,
+      "loss": 1.0101,
+      "step": 11695
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.615131578947369e-06,
+      "loss": 0.74,
+      "step": 11696
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.6145833333333334e-06,
+      "loss": 0.5847,
+      "step": 11697
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.6140350877192986e-06,
+      "loss": 0.7759,
+      "step": 11698
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.613486842105263e-06,
+      "loss": 1.1561,
+      "step": 11699
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.6129385964912283e-06,
+      "loss": 0.1672,
+      "step": 11700
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.6123903508771935e-06,
+      "loss": 0.5817,
+      "step": 11701
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.611842105263158e-06,
+      "loss": 1.1764,
+      "step": 11702
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.611293859649123e-06,
+      "loss": 0.8441,
+      "step": 11703
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.6107456140350884e-06,
+      "loss": 0.4126,
+      "step": 11704
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.610197368421053e-06,
+      "loss": 0.57,
+      "step": 11705
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.609649122807018e-06,
+      "loss": 0.6573,
+      "step": 11706
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.6091008771929825e-06,
+      "loss": 0.6375,
+      "step": 11707
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.6085526315789477e-06,
+      "loss": 1.2755,
+      "step": 11708
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.6080043859649126e-06,
+      "loss": 0.2062,
+      "step": 11709
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.6074561403508774e-06,
+      "loss": 0.2392,
+      "step": 11710
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.6069078947368426e-06,
+      "loss": 0.2343,
+      "step": 11711
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.6063596491228074e-06,
+      "loss": 0.3916,
+      "step": 11712
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.6058114035087723e-06,
+      "loss": 0.8977,
+      "step": 11713
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.605263157894737e-06,
+      "loss": 0.1592,
+      "step": 11714
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.604714912280702e-06,
+      "loss": 0.6123,
+      "step": 11715
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.6041666666666667e-06,
+      "loss": 0.2014,
+      "step": 11716
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.603618421052632e-06,
+      "loss": 0.366,
+      "step": 11717
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.6030701754385964e-06,
+      "loss": 0.3747,
+      "step": 11718
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.6025219298245616e-06,
+      "loss": 0.7994,
+      "step": 11719
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.601973684210527e-06,
+      "loss": 0.9729,
+      "step": 11720
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.6014254385964913e-06,
+      "loss": 0.0567,
+      "step": 11721
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.6008771929824565e-06,
+      "loss": 0.8972,
+      "step": 11722
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.600328947368421e-06,
+      "loss": 0.1562,
+      "step": 11723
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.599780701754386e-06,
+      "loss": 0.4003,
+      "step": 11724
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.5992324561403514e-06,
+      "loss": 0.7758,
+      "step": 11725
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.598684210526316e-06,
+      "loss": 0.9204,
+      "step": 11726
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.598135964912281e-06,
+      "loss": 0.8581,
+      "step": 11727
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.5975877192982463e-06,
+      "loss": 0.7087,
+      "step": 11728
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.5970394736842107e-06,
+      "loss": 0.4496,
+      "step": 11729
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.596491228070176e-06,
+      "loss": 0.406,
+      "step": 11730
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.5959429824561404e-06,
+      "loss": 0.259,
+      "step": 11731
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.5953947368421056e-06,
+      "loss": 0.5011,
+      "step": 11732
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.5948464912280705e-06,
+      "loss": 0.7281,
+      "step": 11733
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.5942982456140353e-06,
+      "loss": 1.3434,
+      "step": 11734
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.59375e-06,
+      "loss": 0.3776,
+      "step": 11735
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.5932017543859654e-06,
+      "loss": 0.4882,
+      "step": 11736
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.59265350877193e-06,
+      "loss": 0.52,
+      "step": 11737
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.592105263157895e-06,
+      "loss": 0.3254,
+      "step": 11738
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.5915570175438603e-06,
+      "loss": 0.4434,
+      "step": 11739
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.5910087719298247e-06,
+      "loss": 0.4435,
+      "step": 11740
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.59046052631579e-06,
+      "loss": 0.9895,
+      "step": 11741
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.5899122807017543e-06,
+      "loss": 0.3948,
+      "step": 11742
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.5893640350877196e-06,
+      "loss": 1.0909,
+      "step": 11743
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.588815789473685e-06,
+      "loss": 0.2193,
+      "step": 11744
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.588267543859649e-06,
+      "loss": 0.9014,
+      "step": 11745
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.5877192982456145e-06,
+      "loss": 0.3146,
+      "step": 11746
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.5871710526315797e-06,
+      "loss": 0.6833,
+      "step": 11747
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.586622807017544e-06,
+      "loss": 0.7762,
+      "step": 11748
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.5860745614035093e-06,
+      "loss": 0.2083,
+      "step": 11749
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.5855263157894737e-06,
+      "loss": 0.4276,
+      "step": 11750
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.584978070175439e-06,
+      "loss": 0.4868,
+      "step": 11751
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.584429824561404e-06,
+      "loss": 0.8173,
+      "step": 11752
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.5838815789473686e-06,
+      "loss": 0.5291,
+      "step": 11753
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.5833333333333335e-06,
+      "loss": 1.1158,
+      "step": 11754
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.5827850877192987e-06,
+      "loss": 0.7472,
+      "step": 11755
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.5822368421052635e-06,
+      "loss": 0.4811,
+      "step": 11756
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.5816885964912284e-06,
+      "loss": 0.4392,
+      "step": 11757
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.581140350877193e-06,
+      "loss": 0.1548,
+      "step": 11758
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.580592105263158e-06,
+      "loss": 0.3952,
+      "step": 11759
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.5800438596491233e-06,
+      "loss": 0.5015,
+      "step": 11760
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.5794956140350877e-06,
+      "loss": 0.6232,
+      "step": 11761
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.578947368421053e-06,
+      "loss": 0.4524,
+      "step": 11762
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.578399122807018e-06,
+      "loss": 0.3307,
+      "step": 11763
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.5778508771929826e-06,
+      "loss": 0.1829,
+      "step": 11764
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.577302631578948e-06,
+      "loss": 0.4042,
+      "step": 11765
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.5767543859649122e-06,
+      "loss": 0.4767,
+      "step": 11766
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.5762061403508775e-06,
+      "loss": 0.7812,
+      "step": 11767
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.5756578947368427e-06,
+      "loss": 0.3549,
+      "step": 11768
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.575109649122807e-06,
+      "loss": 0.574,
+      "step": 11769
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.5745614035087724e-06,
+      "loss": 1.553,
+      "step": 11770
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.574013157894737e-06,
+      "loss": 0.7854,
+      "step": 11771
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.573464912280702e-06,
+      "loss": 1.3569,
+      "step": 11772
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.572916666666667e-06,
+      "loss": 0.2612,
+      "step": 11773
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.5723684210526317e-06,
+      "loss": 0.3311,
+      "step": 11774
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.571820175438597e-06,
+      "loss": 0.8486,
+      "step": 11775
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.5712719298245617e-06,
+      "loss": 0.1735,
+      "step": 11776
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.5707236842105266e-06,
+      "loss": 0.2551,
+      "step": 11777
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.5701754385964914e-06,
+      "loss": 0.7705,
+      "step": 11778
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.5696271929824566e-06,
+      "loss": 0.2847,
+      "step": 11779
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.569078947368421e-06,
+      "loss": 0.8142,
+      "step": 11780
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.5685307017543863e-06,
+      "loss": 0.6523,
+      "step": 11781
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.5679824561403515e-06,
+      "loss": 0.2276,
+      "step": 11782
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.567434210526316e-06,
+      "loss": 0.5239,
+      "step": 11783
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.566885964912281e-06,
+      "loss": 0.6783,
+      "step": 11784
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.5663377192982456e-06,
+      "loss": 0.8581,
+      "step": 11785
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.565789473684211e-06,
+      "loss": 0.8972,
+      "step": 11786
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.565241228070176e-06,
+      "loss": 0.6308,
+      "step": 11787
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.5646929824561405e-06,
+      "loss": 0.378,
+      "step": 11788
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.5641447368421057e-06,
+      "loss": 0.368,
+      "step": 11789
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.5635964912280705e-06,
+      "loss": 0.5209,
+      "step": 11790
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.5630482456140354e-06,
+      "loss": 0.5278,
+      "step": 11791
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.5625e-06,
+      "loss": 0.2876,
+      "step": 11792
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.561951754385965e-06,
+      "loss": 0.1566,
+      "step": 11793
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.5614035087719303e-06,
+      "loss": 0.445,
+      "step": 11794
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.560855263157895e-06,
+      "loss": 0.0567,
+      "step": 11795
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.56030701754386e-06,
+      "loss": 0.1929,
+      "step": 11796
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.5597587719298247e-06,
+      "loss": 0.304,
+      "step": 11797
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.55921052631579e-06,
+      "loss": 0.722,
+      "step": 11798
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.5586622807017544e-06,
+      "loss": 0.838,
+      "step": 11799
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.5581140350877196e-06,
+      "loss": 1.0589,
+      "step": 11800
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.557565789473684e-06,
+      "loss": 0.5861,
+      "step": 11801
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.5570175438596493e-06,
+      "loss": 0.3208,
+      "step": 11802
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.5564692982456145e-06,
+      "loss": 0.1425,
+      "step": 11803
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.555921052631579e-06,
+      "loss": 0.1653,
+      "step": 11804
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.555372807017544e-06,
+      "loss": 0.7855,
+      "step": 11805
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.5548245614035094e-06,
+      "loss": 0.843,
+      "step": 11806
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.554276315789474e-06,
+      "loss": 0.9254,
+      "step": 11807
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.553728070175439e-06,
+      "loss": 0.4971,
+      "step": 11808
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.5531798245614035e-06,
+      "loss": 0.7828,
+      "step": 11809
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.5526315789473687e-06,
+      "loss": 0.4947,
+      "step": 11810
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.552083333333334e-06,
+      "loss": 0.8381,
+      "step": 11811
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.5515350877192984e-06,
+      "loss": 1.1216,
+      "step": 11812
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.5509868421052636e-06,
+      "loss": 0.6627,
+      "step": 11813
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.5504385964912285e-06,
+      "loss": 0.4305,
+      "step": 11814
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.5498903508771933e-06,
+      "loss": 0.534,
+      "step": 11815
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.549342105263158e-06,
+      "loss": 0.9739,
+      "step": 11816
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.5487938596491234e-06,
+      "loss": 0.7701,
+      "step": 11817
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.5482456140350878e-06,
+      "loss": 0.2331,
+      "step": 11818
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.547697368421053e-06,
+      "loss": 1.5412,
+      "step": 11819
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.547149122807018e-06,
+      "loss": 0.2203,
+      "step": 11820
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.5466008771929827e-06,
+      "loss": 0.4018,
+      "step": 11821
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.546052631578948e-06,
+      "loss": 1.1397,
+      "step": 11822
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.5455043859649123e-06,
+      "loss": 0.2134,
+      "step": 11823
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.5449561403508775e-06,
+      "loss": 0.3352,
+      "step": 11824
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.544407894736843e-06,
+      "loss": 1.5563,
+      "step": 11825
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.543859649122807e-06,
+      "loss": 0.8661,
+      "step": 11826
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.5433114035087724e-06,
+      "loss": 0.9518,
+      "step": 11827
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.542763157894737e-06,
+      "loss": 0.7602,
+      "step": 11828
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.542214912280702e-06,
+      "loss": 0.3561,
+      "step": 11829
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.5416666666666673e-06,
+      "loss": 0.1815,
+      "step": 11830
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.5411184210526317e-06,
+      "loss": 0.821,
+      "step": 11831
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.540570175438597e-06,
+      "loss": 0.3988,
+      "step": 11832
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.540021929824562e-06,
+      "loss": 1.4752,
+      "step": 11833
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.5394736842105266e-06,
+      "loss": 0.1577,
+      "step": 11834
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.5389254385964915e-06,
+      "loss": 0.3581,
+      "step": 11835
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.5383771929824563e-06,
+      "loss": 1.1599,
+      "step": 11836
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.537828947368421e-06,
+      "loss": 0.17,
+      "step": 11837
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.5372807017543864e-06,
+      "loss": 0.436,
+      "step": 11838
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.536732456140351e-06,
+      "loss": 0.662,
+      "step": 11839
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.536184210526316e-06,
+      "loss": 0.6246,
+      "step": 11840
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.5356359649122813e-06,
+      "loss": 0.5296,
+      "step": 11841
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.5350877192982457e-06,
+      "loss": 0.1813,
+      "step": 11842
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.534539473684211e-06,
+      "loss": 0.479,
+      "step": 11843
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.5339912280701753e-06,
+      "loss": 1.0246,
+      "step": 11844
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.5334429824561406e-06,
+      "loss": 0.711,
+      "step": 11845
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.532894736842106e-06,
+      "loss": 0.9132,
+      "step": 11846
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.53234649122807e-06,
+      "loss": 0.4908,
+      "step": 11847
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.5317982456140355e-06,
+      "loss": 0.4445,
+      "step": 11848
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.5312500000000007e-06,
+      "loss": 0.3505,
+      "step": 11849
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.530701754385965e-06,
+      "loss": 1.1333,
+      "step": 11850
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.5301535087719304e-06,
+      "loss": 0.5105,
+      "step": 11851
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.5296052631578948e-06,
+      "loss": 0.5515,
+      "step": 11852
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.52905701754386e-06,
+      "loss": 0.2683,
+      "step": 11853
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.528508771929825e-06,
+      "loss": 0.6251,
+      "step": 11854
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.5279605263157897e-06,
+      "loss": 0.4248,
+      "step": 11855
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.5274122807017545e-06,
+      "loss": 0.7669,
+      "step": 11856
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.5268640350877197e-06,
+      "loss": 1.5772,
+      "step": 11857
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.5263157894736846e-06,
+      "loss": 0.8973,
+      "step": 11858
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.5257675438596494e-06,
+      "loss": 0.8152,
+      "step": 11859
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.5252192982456146e-06,
+      "loss": 0.4008,
+      "step": 11860
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.524671052631579e-06,
+      "loss": 0.4937,
+      "step": 11861
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.5241228070175443e-06,
+      "loss": 0.5406,
+      "step": 11862
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.5235745614035087e-06,
+      "loss": 0.4252,
+      "step": 11863
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.523026315789474e-06,
+      "loss": 0.2785,
+      "step": 11864
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.522478070175439e-06,
+      "loss": 0.4834,
+      "step": 11865
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.5219298245614036e-06,
+      "loss": 1.3066,
+      "step": 11866
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.521381578947369e-06,
+      "loss": 0.4134,
+      "step": 11867
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.520833333333334e-06,
+      "loss": 0.5522,
+      "step": 11868
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.5202850877192985e-06,
+      "loss": 0.7217,
+      "step": 11869
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.5197368421052637e-06,
+      "loss": 0.7223,
+      "step": 11870
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.519188596491228e-06,
+      "loss": 0.4555,
+      "step": 11871
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.5186403508771934e-06,
+      "loss": 0.4896,
+      "step": 11872
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.518092105263158e-06,
+      "loss": 1.0885,
+      "step": 11873
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.517543859649123e-06,
+      "loss": 0.2174,
+      "step": 11874
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.516995614035088e-06,
+      "loss": 0.1394,
+      "step": 11875
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.516447368421053e-06,
+      "loss": 0.1557,
+      "step": 11876
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.515899122807018e-06,
+      "loss": 0.602,
+      "step": 11877
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.5153508771929827e-06,
+      "loss": 0.1127,
+      "step": 11878
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.5148026315789476e-06,
+      "loss": 0.7187,
+      "step": 11879
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.5142543859649124e-06,
+      "loss": 0.5066,
+      "step": 11880
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.5137061403508776e-06,
+      "loss": 1.0269,
+      "step": 11881
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.513157894736842e-06,
+      "loss": 0.6992,
+      "step": 11882
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.5126096491228073e-06,
+      "loss": 0.9685,
+      "step": 11883
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.5120614035087725e-06,
+      "loss": 0.6491,
+      "step": 11884
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.511513157894737e-06,
+      "loss": 0.4587,
+      "step": 11885
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.510964912280702e-06,
+      "loss": 0.3818,
+      "step": 11886
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.5104166666666666e-06,
+      "loss": 0.198,
+      "step": 11887
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.509868421052632e-06,
+      "loss": 0.3297,
+      "step": 11888
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.509320175438597e-06,
+      "loss": 0.2662,
+      "step": 11889
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.5087719298245615e-06,
+      "loss": 0.3689,
+      "step": 11890
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.5082236842105267e-06,
+      "loss": 1.0521,
+      "step": 11891
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.5076754385964916e-06,
+      "loss": 0.2285,
+      "step": 11892
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.5071271929824564e-06,
+      "loss": 0.4377,
+      "step": 11893
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.506578947368421e-06,
+      "loss": 0.8382,
+      "step": 11894
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.506030701754386e-06,
+      "loss": 0.6727,
+      "step": 11895
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.5054824561403513e-06,
+      "loss": 0.1969,
+      "step": 11896
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.504934210526316e-06,
+      "loss": 0.6257,
+      "step": 11897
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.504385964912281e-06,
+      "loss": 0.6882,
+      "step": 11898
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.5038377192982458e-06,
+      "loss": 0.3703,
+      "step": 11899
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.503289473684211e-06,
+      "loss": 1.1525,
+      "step": 11900
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.5027412280701754e-06,
+      "loss": 0.2458,
+      "step": 11901
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.5021929824561406e-06,
+      "loss": 0.0407,
+      "step": 11902
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.501644736842106e-06,
+      "loss": 0.3774,
+      "step": 11903
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.5010964912280703e-06,
+      "loss": 0.2515,
+      "step": 11904
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.5005482456140355e-06,
+      "loss": 0.2785,
+      "step": 11905
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.5e-06,
+      "loss": 0.3573,
+      "step": 11906
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.499451754385965e-06,
+      "loss": 0.4063,
+      "step": 11907
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.4989035087719304e-06,
+      "loss": 1.4138,
+      "step": 11908
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.498355263157895e-06,
+      "loss": 0.096,
+      "step": 11909
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.49780701754386e-06,
+      "loss": 0.4759,
+      "step": 11910
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.497258771929825e-06,
+      "loss": 1.1522,
+      "step": 11911
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.4967105263157897e-06,
+      "loss": 0.3155,
+      "step": 11912
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.496162280701755e-06,
+      "loss": 0.7003,
+      "step": 11913
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.4956140350877194e-06,
+      "loss": 0.1988,
+      "step": 11914
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.4950657894736846e-06,
+      "loss": 0.4009,
+      "step": 11915
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.4945175438596495e-06,
+      "loss": 0.3674,
+      "step": 11916
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.4939692982456143e-06,
+      "loss": 0.8425,
+      "step": 11917
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.493421052631579e-06,
+      "loss": 0.5205,
+      "step": 11918
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.4928728070175444e-06,
+      "loss": 0.0278,
+      "step": 11919
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.4923245614035088e-06,
+      "loss": 0.9306,
+      "step": 11920
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.491776315789474e-06,
+      "loss": 0.3513,
+      "step": 11921
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.491228070175439e-06,
+      "loss": 0.9346,
+      "step": 11922
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.4906798245614037e-06,
+      "loss": 0.9345,
+      "step": 11923
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.490131578947369e-06,
+      "loss": 0.6269,
+      "step": 11924
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.4895833333333333e-06,
+      "loss": 0.0887,
+      "step": 11925
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.4890350877192986e-06,
+      "loss": 0.94,
+      "step": 11926
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.488486842105264e-06,
+      "loss": 0.7409,
+      "step": 11927
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.487938596491228e-06,
+      "loss": 0.6277,
+      "step": 11928
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.4873903508771935e-06,
+      "loss": 0.3576,
+      "step": 11929
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.486842105263158e-06,
+      "loss": 0.4096,
+      "step": 11930
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.486293859649123e-06,
+      "loss": 0.8214,
+      "step": 11931
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.4857456140350884e-06,
+      "loss": 0.3196,
+      "step": 11932
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.4851973684210528e-06,
+      "loss": 0.1601,
+      "step": 11933
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.484649122807018e-06,
+      "loss": 0.8043,
+      "step": 11934
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.484100877192983e-06,
+      "loss": 0.9145,
+      "step": 11935
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.4835526315789477e-06,
+      "loss": 0.5594,
+      "step": 11936
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.4830043859649125e-06,
+      "loss": 0.4059,
+      "step": 11937
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.4824561403508773e-06,
+      "loss": 0.8662,
+      "step": 11938
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.481907894736842e-06,
+      "loss": 1.4149,
+      "step": 11939
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.4813596491228074e-06,
+      "loss": 0.2448,
+      "step": 11940
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.480811403508772e-06,
+      "loss": 0.6159,
+      "step": 11941
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.480263157894737e-06,
+      "loss": 1.2213,
+      "step": 11942
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.4797149122807023e-06,
+      "loss": 0.2995,
+      "step": 11943
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.4791666666666667e-06,
+      "loss": 0.3339,
+      "step": 11944
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.478618421052632e-06,
+      "loss": 0.7326,
+      "step": 11945
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.478070175438597e-06,
+      "loss": 0.3637,
+      "step": 11946
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.4775219298245616e-06,
+      "loss": 0.0526,
+      "step": 11947
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.476973684210527e-06,
+      "loss": 0.1785,
+      "step": 11948
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.4764254385964912e-06,
+      "loss": 0.1078,
+      "step": 11949
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.4758771929824565e-06,
+      "loss": 0.3723,
+      "step": 11950
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.4753289473684217e-06,
+      "loss": 0.3577,
+      "step": 11951
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.474780701754386e-06,
+      "loss": 1.1737,
+      "step": 11952
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.4742324561403514e-06,
+      "loss": 0.3543,
+      "step": 11953
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.473684210526316e-06,
+      "loss": 0.9318,
+      "step": 11954
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.473135964912281e-06,
+      "loss": 0.3059,
+      "step": 11955
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.472587719298246e-06,
+      "loss": 0.2584,
+      "step": 11956
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.4720394736842107e-06,
+      "loss": 1.0048,
+      "step": 11957
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.4714912280701755e-06,
+      "loss": 0.8758,
+      "step": 11958
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.4709429824561407e-06,
+      "loss": 1.6268,
+      "step": 11959
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.4703947368421056e-06,
+      "loss": 0.3512,
+      "step": 11960
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.4698464912280704e-06,
+      "loss": 0.5751,
+      "step": 11961
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.4692982456140356e-06,
+      "loss": 0.6811,
+      "step": 11962
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.46875e-06,
+      "loss": 1.3165,
+      "step": 11963
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.4682017543859653e-06,
+      "loss": 0.4272,
+      "step": 11964
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.4676535087719297e-06,
+      "loss": 0.2255,
+      "step": 11965
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.467105263157895e-06,
+      "loss": 0.7986,
+      "step": 11966
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.46655701754386e-06,
+      "loss": 0.9138,
+      "step": 11967
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.4660087719298246e-06,
+      "loss": 0.4528,
+      "step": 11968
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.46546052631579e-06,
+      "loss": 0.7562,
+      "step": 11969
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.464912280701755e-06,
+      "loss": 0.3702,
+      "step": 11970
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.4643640350877195e-06,
+      "loss": 0.4512,
+      "step": 11971
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.4638157894736847e-06,
+      "loss": 0.7042,
+      "step": 11972
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.463267543859649e-06,
+      "loss": 0.9595,
+      "step": 11973
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.4627192982456144e-06,
+      "loss": 0.5041,
+      "step": 11974
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.462171052631579e-06,
+      "loss": 0.1769,
+      "step": 11975
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.461622807017544e-06,
+      "loss": 0.3607,
+      "step": 11976
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.461074561403509e-06,
+      "loss": 1.057,
+      "step": 11977
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.460526315789474e-06,
+      "loss": 0.1518,
+      "step": 11978
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.459978070175439e-06,
+      "loss": 0.8202,
+      "step": 11979
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.4594298245614037e-06,
+      "loss": 0.6806,
+      "step": 11980
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.458881578947369e-06,
+      "loss": 0.7789,
+      "step": 11981
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.4583333333333334e-06,
+      "loss": 0.5853,
+      "step": 11982
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.4577850877192986e-06,
+      "loss": 0.2988,
+      "step": 11983
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.457236842105263e-06,
+      "loss": 0.5846,
+      "step": 11984
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.4566885964912283e-06,
+      "loss": 1.5111,
+      "step": 11985
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.4561403508771935e-06,
+      "loss": 0.5943,
+      "step": 11986
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.455592105263158e-06,
+      "loss": 0.7457,
+      "step": 11987
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.455043859649123e-06,
+      "loss": 0.5651,
+      "step": 11988
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.4544956140350884e-06,
+      "loss": 0.6437,
+      "step": 11989
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.453947368421053e-06,
+      "loss": 0.6915,
+      "step": 11990
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.453399122807018e-06,
+      "loss": 1.0287,
+      "step": 11991
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.4528508771929825e-06,
+      "loss": 0.5574,
+      "step": 11992
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.4523026315789477e-06,
+      "loss": 0.5771,
+      "step": 11993
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.4517543859649126e-06,
+      "loss": 0.5914,
+      "step": 11994
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.4512061403508774e-06,
+      "loss": 1.0152,
+      "step": 11995
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.4506578947368426e-06,
+      "loss": 0.2938,
+      "step": 11996
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.4501096491228075e-06,
+      "loss": 0.7605,
+      "step": 11997
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.4495614035087723e-06,
+      "loss": 0.6747,
+      "step": 11998
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.449013157894737e-06,
+      "loss": 0.562,
+      "step": 11999
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.448464912280702e-06,
+      "loss": 1.1326,
+      "step": 12000
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.4479166666666668e-06,
+      "loss": 0.6811,
+      "step": 12001
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.447368421052632e-06,
+      "loss": 0.8927,
+      "step": 12002
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.4468201754385964e-06,
+      "loss": 0.7208,
+      "step": 12003
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.4462719298245617e-06,
+      "loss": 1.3321,
+      "step": 12004
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.445723684210527e-06,
+      "loss": 0.2695,
+      "step": 12005
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.4451754385964913e-06,
+      "loss": 0.9851,
+      "step": 12006
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.4446271929824566e-06,
+      "loss": 0.2188,
+      "step": 12007
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.444078947368421e-06,
+      "loss": 0.7933,
+      "step": 12008
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.443530701754386e-06,
+      "loss": 0.573,
+      "step": 12009
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.4429824561403515e-06,
+      "loss": 0.5214,
+      "step": 12010
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.442434210526316e-06,
+      "loss": 0.3915,
+      "step": 12011
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.441885964912281e-06,
+      "loss": 0.9638,
+      "step": 12012
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.441337719298246e-06,
+      "loss": 0.2188,
+      "step": 12013
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.4407894736842108e-06,
+      "loss": 0.4344,
+      "step": 12014
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.440241228070176e-06,
+      "loss": 0.6558,
+      "step": 12015
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.4396929824561404e-06,
+      "loss": 0.3734,
+      "step": 12016
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.4391447368421056e-06,
+      "loss": 0.5656,
+      "step": 12017
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.4385964912280705e-06,
+      "loss": 0.6127,
+      "step": 12018
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.4380482456140353e-06,
+      "loss": 0.7891,
+      "step": 12019
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.4375e-06,
+      "loss": 0.3015,
+      "step": 12020
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.4369517543859654e-06,
+      "loss": 0.5216,
+      "step": 12021
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.4364035087719298e-06,
+      "loss": 0.3449,
+      "step": 12022
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.435855263157895e-06,
+      "loss": 0.8435,
+      "step": 12023
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.4353070175438603e-06,
+      "loss": 0.3168,
+      "step": 12024
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.4347587719298247e-06,
+      "loss": 0.5658,
+      "step": 12025
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.43421052631579e-06,
+      "loss": 1.7117,
+      "step": 12026
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.4336622807017543e-06,
+      "loss": 0.4186,
+      "step": 12027
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.4331140350877196e-06,
+      "loss": 0.6441,
+      "step": 12028
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.432565789473685e-06,
+      "loss": 0.3288,
+      "step": 12029
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.4320175438596492e-06,
+      "loss": 0.0896,
+      "step": 12030
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.4314692982456145e-06,
+      "loss": 0.4293,
+      "step": 12031
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.4309210526315793e-06,
+      "loss": 0.3143,
+      "step": 12032
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.430372807017544e-06,
+      "loss": 0.9275,
+      "step": 12033
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.4298245614035094e-06,
+      "loss": 0.8804,
+      "step": 12034
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.4292763157894738e-06,
+      "loss": 0.6031,
+      "step": 12035
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.428728070175439e-06,
+      "loss": 1.2827,
+      "step": 12036
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.428179824561404e-06,
+      "loss": 0.3344,
+      "step": 12037
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.4276315789473687e-06,
+      "loss": 0.5248,
+      "step": 12038
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.4270833333333335e-06,
+      "loss": 0.6034,
+      "step": 12039
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.4265350877192987e-06,
+      "loss": 0.5275,
+      "step": 12040
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.425986842105263e-06,
+      "loss": 0.5891,
+      "step": 12041
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.4254385964912284e-06,
+      "loss": 0.2883,
+      "step": 12042
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.424890350877193e-06,
+      "loss": 0.1046,
+      "step": 12043
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.424342105263158e-06,
+      "loss": 0.0534,
+      "step": 12044
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.4237938596491233e-06,
+      "loss": 0.2344,
+      "step": 12045
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.4232456140350877e-06,
+      "loss": 1.0731,
+      "step": 12046
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.422697368421053e-06,
+      "loss": 0.194,
+      "step": 12047
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.422149122807018e-06,
+      "loss": 0.2069,
+      "step": 12048
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.4216008771929826e-06,
+      "loss": 0.6517,
+      "step": 12049
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.421052631578948e-06,
+      "loss": 0.1463,
+      "step": 12050
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.4205043859649122e-06,
+      "loss": 0.0707,
+      "step": 12051
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.4199561403508775e-06,
+      "loss": 0.5948,
+      "step": 12052
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.4194078947368427e-06,
+      "loss": 0.8857,
+      "step": 12053
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.418859649122807e-06,
+      "loss": 0.6694,
+      "step": 12054
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.4183114035087724e-06,
+      "loss": 0.5788,
+      "step": 12055
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.417763157894737e-06,
+      "loss": 1.5523,
+      "step": 12056
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.417214912280702e-06,
+      "loss": 0.6618,
+      "step": 12057
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.416666666666667e-06,
+      "loss": 1.149,
+      "step": 12058
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.4161184210526317e-06,
+      "loss": 0.5298,
+      "step": 12059
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.4155701754385965e-06,
+      "loss": 0.4022,
+      "step": 12060
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.4150219298245617e-06,
+      "loss": 0.5734,
+      "step": 12061
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.4144736842105266e-06,
+      "loss": 0.4367,
+      "step": 12062
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.4139254385964914e-06,
+      "loss": 0.3111,
+      "step": 12063
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.4133771929824566e-06,
+      "loss": 0.8173,
+      "step": 12064
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.412828947368421e-06,
+      "loss": 0.3365,
+      "step": 12065
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.4122807017543863e-06,
+      "loss": 0.5414,
+      "step": 12066
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.4117324561403515e-06,
+      "loss": 0.6426,
+      "step": 12067
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.411184210526316e-06,
+      "loss": 0.2934,
+      "step": 12068
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.410635964912281e-06,
+      "loss": 0.4331,
+      "step": 12069
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.4100877192982456e-06,
+      "loss": 0.317,
+      "step": 12070
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.409539473684211e-06,
+      "loss": 0.4464,
+      "step": 12071
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.408991228070176e-06,
+      "loss": 2.1221,
+      "step": 12072
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.4084429824561405e-06,
+      "loss": 0.2267,
+      "step": 12073
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.4078947368421057e-06,
+      "loss": 0.95,
+      "step": 12074
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.4073464912280706e-06,
+      "loss": 0.3983,
+      "step": 12075
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.4067982456140354e-06,
+      "loss": 1.0685,
+      "step": 12076
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.40625e-06,
+      "loss": 1.0645,
+      "step": 12077
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.405701754385965e-06,
+      "loss": 0.3016,
+      "step": 12078
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.4051535087719303e-06,
+      "loss": 0.5747,
+      "step": 12079
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.404605263157895e-06,
+      "loss": 0.9189,
+      "step": 12080
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.40405701754386e-06,
+      "loss": 0.4615,
+      "step": 12081
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.4035087719298248e-06,
+      "loss": 1.7338,
+      "step": 12082
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.40296052631579e-06,
+      "loss": 0.133,
+      "step": 12083
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.4024122807017544e-06,
+      "loss": 0.9402,
+      "step": 12084
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.4018640350877197e-06,
+      "loss": 0.2994,
+      "step": 12085
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.401315789473684e-06,
+      "loss": 0.6694,
+      "step": 12086
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.4007675438596493e-06,
+      "loss": 0.6608,
+      "step": 12087
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.4002192982456146e-06,
+      "loss": 0.8728,
+      "step": 12088
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.399671052631579e-06,
+      "loss": 0.8276,
+      "step": 12089
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.399122807017544e-06,
+      "loss": 0.3104,
+      "step": 12090
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.3985745614035094e-06,
+      "loss": 0.5059,
+      "step": 12091
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.398026315789474e-06,
+      "loss": 0.4368,
+      "step": 12092
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.397478070175439e-06,
+      "loss": 0.6063,
+      "step": 12093
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.3969298245614035e-06,
+      "loss": 0.7184,
+      "step": 12094
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.3963815789473687e-06,
+      "loss": 0.729,
+      "step": 12095
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.3958333333333336e-06,
+      "loss": 0.5506,
+      "step": 12096
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.3952850877192984e-06,
+      "loss": 0.1677,
+      "step": 12097
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.3947368421052636e-06,
+      "loss": 1.0241,
+      "step": 12098
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3941885964912285e-06,
+      "loss": 0.5365,
+      "step": 12099
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3936403508771933e-06,
+      "loss": 0.6094,
+      "step": 12100
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.393092105263158e-06,
+      "loss": 0.2997,
+      "step": 12101
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3925438596491234e-06,
+      "loss": 0.6283,
+      "step": 12102
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3919956140350878e-06,
+      "loss": 0.9166,
+      "step": 12103
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.391447368421053e-06,
+      "loss": 0.3767,
+      "step": 12104
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3908991228070174e-06,
+      "loss": 0.3492,
+      "step": 12105
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3903508771929827e-06,
+      "loss": 0.7433,
+      "step": 12106
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.389802631578948e-06,
+      "loss": 0.4472,
+      "step": 12107
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3892543859649123e-06,
+      "loss": 0.4671,
+      "step": 12108
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3887061403508776e-06,
+      "loss": 0.7175,
+      "step": 12109
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.388157894736843e-06,
+      "loss": 0.572,
+      "step": 12110
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3876096491228072e-06,
+      "loss": 0.5113,
+      "step": 12111
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3870614035087725e-06,
+      "loss": 0.4195,
+      "step": 12112
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.386513157894737e-06,
+      "loss": 1.7124,
+      "step": 12113
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.385964912280702e-06,
+      "loss": 0.6511,
+      "step": 12114
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.385416666666667e-06,
+      "loss": 0.039,
+      "step": 12115
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3848684210526318e-06,
+      "loss": 0.5973,
+      "step": 12116
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.384320175438597e-06,
+      "loss": 0.8514,
+      "step": 12117
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.383771929824562e-06,
+      "loss": 0.1862,
+      "step": 12118
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3832236842105267e-06,
+      "loss": 0.3322,
+      "step": 12119
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3826754385964915e-06,
+      "loss": 1.2562,
+      "step": 12120
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3821271929824563e-06,
+      "loss": 0.3694,
+      "step": 12121
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.381578947368421e-06,
+      "loss": 0.708,
+      "step": 12122
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3810307017543864e-06,
+      "loss": 0.7752,
+      "step": 12123
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3804824561403508e-06,
+      "loss": 0.7043,
+      "step": 12124
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.379934210526316e-06,
+      "loss": 0.8532,
+      "step": 12125
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3793859649122813e-06,
+      "loss": 0.6568,
+      "step": 12126
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3788377192982457e-06,
+      "loss": 1.5389,
+      "step": 12127
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.378289473684211e-06,
+      "loss": 0.9344,
+      "step": 12128
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3777412280701753e-06,
+      "loss": 0.2462,
+      "step": 12129
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3771929824561406e-06,
+      "loss": 0.5844,
+      "step": 12130
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.376644736842106e-06,
+      "loss": 0.1817,
+      "step": 12131
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3760964912280702e-06,
+      "loss": 0.312,
+      "step": 12132
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3755482456140355e-06,
+      "loss": 0.6896,
+      "step": 12133
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3750000000000003e-06,
+      "loss": 0.4306,
+      "step": 12134
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.374451754385965e-06,
+      "loss": 1.0982,
+      "step": 12135
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.3739035087719304e-06,
+      "loss": 0.7514,
+      "step": 12136
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.3733552631578948e-06,
+      "loss": 1.2335,
+      "step": 12137
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.37280701754386e-06,
+      "loss": 0.9152,
+      "step": 12138
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.372258771929825e-06,
+      "loss": 0.864,
+      "step": 12139
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.3717105263157897e-06,
+      "loss": 0.2365,
+      "step": 12140
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.3711622807017545e-06,
+      "loss": 0.5835,
+      "step": 12141
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.3706140350877197e-06,
+      "loss": 0.6554,
+      "step": 12142
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.370065789473684e-06,
+      "loss": 0.5931,
+      "step": 12143
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.3695175438596494e-06,
+      "loss": 0.3922,
+      "step": 12144
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.3689692982456146e-06,
+      "loss": 0.2144,
+      "step": 12145
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.368421052631579e-06,
+      "loss": 0.9856,
+      "step": 12146
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.3678728070175443e-06,
+      "loss": 0.5705,
+      "step": 12147
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.3673245614035087e-06,
+      "loss": 0.9278,
+      "step": 12148
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.366776315789474e-06,
+      "loss": 0.3679,
+      "step": 12149
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.366228070175439e-06,
+      "loss": 0.8661,
+      "step": 12150
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.3656798245614036e-06,
+      "loss": 0.8391,
+      "step": 12151
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.365131578947369e-06,
+      "loss": 0.2184,
+      "step": 12152
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.3645833333333337e-06,
+      "loss": 0.5007,
+      "step": 12153
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3640350877192985e-06,
+      "loss": 0.5391,
+      "step": 12154
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3634868421052637e-06,
+      "loss": 0.6824,
+      "step": 12155
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.362938596491228e-06,
+      "loss": 1.1832,
+      "step": 12156
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3623903508771934e-06,
+      "loss": 0.1691,
+      "step": 12157
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.361842105263158e-06,
+      "loss": 0.4214,
+      "step": 12158
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.361293859649123e-06,
+      "loss": 0.3161,
+      "step": 12159
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.360745614035088e-06,
+      "loss": 0.4746,
+      "step": 12160
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.360197368421053e-06,
+      "loss": 0.3638,
+      "step": 12161
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3596491228070175e-06,
+      "loss": 1.0632,
+      "step": 12162
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3591008771929828e-06,
+      "loss": 0.3719,
+      "step": 12163
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3585526315789476e-06,
+      "loss": 0.3397,
+      "step": 12164
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3580043859649124e-06,
+      "loss": 0.1197,
+      "step": 12165
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3574561403508777e-06,
+      "loss": 0.7355,
+      "step": 12166
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.356907894736842e-06,
+      "loss": 0.5746,
+      "step": 12167
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3563596491228073e-06,
+      "loss": 0.1331,
+      "step": 12168
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3558114035087725e-06,
+      "loss": 0.4636,
+      "step": 12169
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.355263157894737e-06,
+      "loss": 0.9109,
+      "step": 12170
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.354714912280702e-06,
+      "loss": 0.6916,
+      "step": 12171
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3541666666666666e-06,
+      "loss": 0.2756,
+      "step": 12172
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.353618421052632e-06,
+      "loss": 0.4448,
+      "step": 12173
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.353070175438597e-06,
+      "loss": 0.4621,
+      "step": 12174
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3525219298245615e-06,
+      "loss": 0.428,
+      "step": 12175
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3519736842105267e-06,
+      "loss": 0.8763,
+      "step": 12176
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3514254385964916e-06,
+      "loss": 0.3565,
+      "step": 12177
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3508771929824564e-06,
+      "loss": 0.4346,
+      "step": 12178
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3503289473684212e-06,
+      "loss": 0.3449,
+      "step": 12179
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.349780701754386e-06,
+      "loss": 1.0276,
+      "step": 12180
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3492324561403513e-06,
+      "loss": 0.8959,
+      "step": 12181
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.348684210526316e-06,
+      "loss": 0.9478,
+      "step": 12182
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.348135964912281e-06,
+      "loss": 0.4207,
+      "step": 12183
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3475877192982458e-06,
+      "loss": 0.1612,
+      "step": 12184
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.347039473684211e-06,
+      "loss": 1.4031,
+      "step": 12185
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3464912280701754e-06,
+      "loss": 0.2878,
+      "step": 12186
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3459429824561407e-06,
+      "loss": 0.185,
+      "step": 12187
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.345394736842106e-06,
+      "loss": 0.61,
+      "step": 12188
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3448464912280703e-06,
+      "loss": 0.4695,
+      "step": 12189
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3442982456140356e-06,
+      "loss": 0.4029,
+      "step": 12190
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.34375e-06,
+      "loss": 0.4665,
+      "step": 12191
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.343201754385965e-06,
+      "loss": 0.7446,
+      "step": 12192
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3426535087719305e-06,
+      "loss": 0.4422,
+      "step": 12193
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.342105263157895e-06,
+      "loss": 0.5736,
+      "step": 12194
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.34155701754386e-06,
+      "loss": 0.2622,
+      "step": 12195
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.341008771929825e-06,
+      "loss": 1.1627,
+      "step": 12196
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3404605263157898e-06,
+      "loss": 0.9215,
+      "step": 12197
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3399122807017546e-06,
+      "loss": 0.3202,
+      "step": 12198
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3393640350877194e-06,
+      "loss": 1.6013,
+      "step": 12199
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3388157894736847e-06,
+      "loss": 1.0161,
+      "step": 12200
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3382675438596495e-06,
+      "loss": 0.6351,
+      "step": 12201
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3377192982456143e-06,
+      "loss": 0.7696,
+      "step": 12202
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.337171052631579e-06,
+      "loss": 0.3937,
+      "step": 12203
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3366228070175444e-06,
+      "loss": 0.721,
+      "step": 12204
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3360745614035088e-06,
+      "loss": 1.0644,
+      "step": 12205
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.335526315789474e-06,
+      "loss": 0.5527,
+      "step": 12206
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3349780701754384e-06,
+      "loss": 0.4738,
+      "step": 12207
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3344298245614037e-06,
+      "loss": 1.2883,
+      "step": 12208
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.333881578947369e-06,
+      "loss": 1.3129,
+      "step": 12209
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 0.8671,
+      "step": 12210
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.3327850877192986e-06,
+      "loss": 0.1741,
+      "step": 12211
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.332236842105264e-06,
+      "loss": 0.4614,
+      "step": 12212
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.3316885964912282e-06,
+      "loss": 0.8445,
+      "step": 12213
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.3311403508771935e-06,
+      "loss": 0.5001,
+      "step": 12214
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.330592105263158e-06,
+      "loss": 0.541,
+      "step": 12215
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.330043859649123e-06,
+      "loss": 1.1952,
+      "step": 12216
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.329495614035088e-06,
+      "loss": 0.4954,
+      "step": 12217
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.3289473684210528e-06,
+      "loss": 1.3827,
+      "step": 12218
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.328399122807018e-06,
+      "loss": 0.3783,
+      "step": 12219
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.327850877192983e-06,
+      "loss": 0.5836,
+      "step": 12220
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.3273026315789477e-06,
+      "loss": 0.6333,
+      "step": 12221
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.3267543859649125e-06,
+      "loss": 0.7725,
+      "step": 12222
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.3262061403508773e-06,
+      "loss": 0.1448,
+      "step": 12223
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.325657894736842e-06,
+      "loss": 0.186,
+      "step": 12224
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.3251096491228074e-06,
+      "loss": 1.4378,
+      "step": 12225
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.324561403508772e-06,
+      "loss": 0.6369,
+      "step": 12226
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.324013157894737e-06,
+      "loss": 0.7135,
+      "step": 12227
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.3234649122807023e-06,
+      "loss": 0.5613,
+      "step": 12228
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.3229166666666667e-06,
+      "loss": 0.6912,
+      "step": 12229
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.322368421052632e-06,
+      "loss": 0.9174,
+      "step": 12230
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.321820175438597e-06,
+      "loss": 0.0679,
+      "step": 12231
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.3212719298245616e-06,
+      "loss": 0.4734,
+      "step": 12232
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.320723684210527e-06,
+      "loss": 0.7569,
+      "step": 12233
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.3201754385964912e-06,
+      "loss": 0.3181,
+      "step": 12234
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.3196271929824565e-06,
+      "loss": 0.1838,
+      "step": 12235
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.3190789473684213e-06,
+      "loss": 0.8859,
+      "step": 12236
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.318530701754386e-06,
+      "loss": 0.6267,
+      "step": 12237
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.3179824561403514e-06,
+      "loss": 0.252,
+      "step": 12238
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.317434210526316e-06,
+      "loss": 1.0828,
+      "step": 12239
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.316885964912281e-06,
+      "loss": 0.8078,
+      "step": 12240
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.316337719298246e-06,
+      "loss": 0.5777,
+      "step": 12241
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.3157894736842107e-06,
+      "loss": 0.4094,
+      "step": 12242
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.3152412280701755e-06,
+      "loss": 0.1649,
+      "step": 12243
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.3146929824561408e-06,
+      "loss": 0.5375,
+      "step": 12244
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.314144736842105e-06,
+      "loss": 0.5549,
+      "step": 12245
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.3135964912280704e-06,
+      "loss": 0.1693,
+      "step": 12246
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.3130482456140356e-06,
+      "loss": 0.3666,
+      "step": 12247
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.3125e-06,
+      "loss": 0.6366,
+      "step": 12248
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.3119517543859653e-06,
+      "loss": 0.8563,
+      "step": 12249
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.3114035087719297e-06,
+      "loss": 1.2394,
+      "step": 12250
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.310855263157895e-06,
+      "loss": 0.7382,
+      "step": 12251
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.31030701754386e-06,
+      "loss": 0.5067,
+      "step": 12252
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.3097587719298246e-06,
+      "loss": 0.1673,
+      "step": 12253
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.30921052631579e-06,
+      "loss": 0.161,
+      "step": 12254
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.308662280701755e-06,
+      "loss": 0.9849,
+      "step": 12255
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.3081140350877195e-06,
+      "loss": 0.2286,
+      "step": 12256
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.3075657894736847e-06,
+      "loss": 0.5341,
+      "step": 12257
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.307017543859649e-06,
+      "loss": 0.4015,
+      "step": 12258
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.3064692982456144e-06,
+      "loss": 0.4859,
+      "step": 12259
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.3059210526315792e-06,
+      "loss": 0.8344,
+      "step": 12260
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.305372807017544e-06,
+      "loss": 0.3629,
+      "step": 12261
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.304824561403509e-06,
+      "loss": 1.1992,
+      "step": 12262
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.304276315789474e-06,
+      "loss": 0.5389,
+      "step": 12263
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.303728070175439e-06,
+      "loss": 0.6227,
+      "step": 12264
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.3031798245614038e-06,
+      "loss": 0.1539,
+      "step": 12265
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.302631578947369e-06,
+      "loss": 0.6795,
+      "step": 12266
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.3020833333333334e-06,
+      "loss": 0.2136,
+      "step": 12267
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.3015350877192987e-06,
+      "loss": 0.8858,
+      "step": 12268
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.300986842105263e-06,
+      "loss": 1.0178,
+      "step": 12269
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.3004385964912283e-06,
+      "loss": 0.3392,
+      "step": 12270
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.2998903508771936e-06,
+      "loss": 0.3541,
+      "step": 12271
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.299342105263158e-06,
+      "loss": 0.3378,
+      "step": 12272
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.298793859649123e-06,
+      "loss": 0.3718,
+      "step": 12273
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.2982456140350885e-06,
+      "loss": 0.8975,
+      "step": 12274
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.297697368421053e-06,
+      "loss": 0.4305,
+      "step": 12275
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.297149122807018e-06,
+      "loss": 0.5341,
+      "step": 12276
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.2966008771929825e-06,
+      "loss": 0.3733,
+      "step": 12277
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.2960526315789478e-06,
+      "loss": 0.3552,
+      "step": 12278
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.2955043859649126e-06,
+      "loss": 1.4454,
+      "step": 12279
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.2949561403508774e-06,
+      "loss": 0.3222,
+      "step": 12280
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.2944078947368422e-06,
+      "loss": 0.4985,
+      "step": 12281
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.2938596491228075e-06,
+      "loss": 0.691,
+      "step": 12282
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.2933114035087723e-06,
+      "loss": 0.8082,
+      "step": 12283
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.292763157894737e-06,
+      "loss": 0.9224,
+      "step": 12284
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.292214912280702e-06,
+      "loss": 0.2308,
+      "step": 12285
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.2916666666666668e-06,
+      "loss": 0.2561,
+      "step": 12286
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.291118421052632e-06,
+      "loss": 0.3851,
+      "step": 12287
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.2905701754385964e-06,
+      "loss": 0.5,
+      "step": 12288
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.2900219298245617e-06,
+      "loss": 0.8066,
+      "step": 12289
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.289473684210527e-06,
+      "loss": 0.5066,
+      "step": 12290
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.2889254385964913e-06,
+      "loss": 0.9241,
+      "step": 12291
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.2883771929824566e-06,
+      "loss": 0.6504,
+      "step": 12292
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.287828947368421e-06,
+      "loss": 0.3805,
+      "step": 12293
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.2872807017543862e-06,
+      "loss": 1.21,
+      "step": 12294
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.2867324561403515e-06,
+      "loss": 0.645,
+      "step": 12295
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.286184210526316e-06,
+      "loss": 0.9228,
+      "step": 12296
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.285635964912281e-06,
+      "loss": 0.2949,
+      "step": 12297
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.285087719298246e-06,
+      "loss": 1.3068,
+      "step": 12298
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.2845394736842108e-06,
+      "loss": 0.3916,
+      "step": 12299
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.2839912280701756e-06,
+      "loss": 0.3859,
+      "step": 12300
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.2834429824561404e-06,
+      "loss": 0.1086,
+      "step": 12301
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.2828947368421057e-06,
+      "loss": 0.2607,
+      "step": 12302
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.2823464912280705e-06,
+      "loss": 0.296,
+      "step": 12303
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.2817982456140353e-06,
+      "loss": 0.3537,
+      "step": 12304
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.28125e-06,
+      "loss": 0.2024,
+      "step": 12305
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.2807017543859654e-06,
+      "loss": 0.2075,
+      "step": 12306
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.2801535087719298e-06,
+      "loss": 0.8538,
+      "step": 12307
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.279605263157895e-06,
+      "loss": 0.123,
+      "step": 12308
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.2790570175438603e-06,
+      "loss": 0.8427,
+      "step": 12309
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.2785087719298247e-06,
+      "loss": 0.7016,
+      "step": 12310
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.27796052631579e-06,
+      "loss": 0.3963,
+      "step": 12311
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.2774122807017543e-06,
+      "loss": 0.2185,
+      "step": 12312
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.2768640350877196e-06,
+      "loss": 0.1215,
+      "step": 12313
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.276315789473685e-06,
+      "loss": 0.9959,
+      "step": 12314
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.2757675438596492e-06,
+      "loss": 0.5165,
+      "step": 12315
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.2752192982456145e-06,
+      "loss": 0.5623,
+      "step": 12316
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.2746710526315793e-06,
+      "loss": 0.9792,
+      "step": 12317
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.274122807017544e-06,
+      "loss": 0.8236,
+      "step": 12318
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.273574561403509e-06,
+      "loss": 0.5253,
+      "step": 12319
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.2730263157894738e-06,
+      "loss": 0.268,
+      "step": 12320
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.272478070175439e-06,
+      "loss": 0.2412,
+      "step": 12321
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.271929824561404e-06,
+      "loss": 0.9911,
+      "step": 12322
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.2713815789473687e-06,
+      "loss": 0.4926,
+      "step": 12323
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.2708333333333335e-06,
+      "loss": 0.3753,
+      "step": 12324
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.2702850877192987e-06,
+      "loss": 1.059,
+      "step": 12325
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.269736842105263e-06,
+      "loss": 0.352,
+      "step": 12326
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.2691885964912284e-06,
+      "loss": 0.6887,
+      "step": 12327
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.268640350877193e-06,
+      "loss": 0.2609,
+      "step": 12328
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.268092105263158e-06,
+      "loss": 0.2167,
+      "step": 12329
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.2675438596491233e-06,
+      "loss": 0.107,
+      "step": 12330
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.2669956140350877e-06,
+      "loss": 0.7988,
+      "step": 12331
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.266447368421053e-06,
+      "loss": 0.4573,
+      "step": 12332
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.265899122807018e-06,
+      "loss": 0.9912,
+      "step": 12333
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.2653508771929826e-06,
+      "loss": 0.428,
+      "step": 12334
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.264802631578948e-06,
+      "loss": 0.2611,
+      "step": 12335
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.2642543859649122e-06,
+      "loss": 1.8312,
+      "step": 12336
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.2637061403508775e-06,
+      "loss": 0.5316,
+      "step": 12337
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.2631578947368423e-06,
+      "loss": 0.5963,
+      "step": 12338
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.262609649122807e-06,
+      "loss": 0.6521,
+      "step": 12339
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.2620614035087724e-06,
+      "loss": 0.4679,
+      "step": 12340
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.261513157894737e-06,
+      "loss": 0.5534,
+      "step": 12341
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.260964912280702e-06,
+      "loss": 0.5903,
+      "step": 12342
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.260416666666667e-06,
+      "loss": 0.5121,
+      "step": 12343
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.2598684210526317e-06,
+      "loss": 0.3843,
+      "step": 12344
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.2593201754385965e-06,
+      "loss": 0.4137,
+      "step": 12345
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.2587719298245618e-06,
+      "loss": 0.6872,
+      "step": 12346
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.2582236842105266e-06,
+      "loss": 0.7285,
+      "step": 12347
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.2576754385964914e-06,
+      "loss": 0.1333,
+      "step": 12348
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.2571271929824567e-06,
+      "loss": 0.2273,
+      "step": 12349
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.256578947368421e-06,
+      "loss": 0.652,
+      "step": 12350
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.2560307017543863e-06,
+      "loss": 0.8531,
+      "step": 12351
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.2554824561403516e-06,
+      "loss": 1.5343,
+      "step": 12352
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.254934210526316e-06,
+      "loss": 0.7742,
+      "step": 12353
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.254385964912281e-06,
+      "loss": 0.4647,
+      "step": 12354
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.2538377192982456e-06,
+      "loss": 0.2501,
+      "step": 12355
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.253289473684211e-06,
+      "loss": 0.3492,
+      "step": 12356
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.252741228070176e-06,
+      "loss": 0.9699,
+      "step": 12357
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.2521929824561405e-06,
+      "loss": 0.6563,
+      "step": 12358
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.2516447368421057e-06,
+      "loss": 0.385,
+      "step": 12359
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.2510964912280706e-06,
+      "loss": 0.5473,
+      "step": 12360
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.2505482456140354e-06,
+      "loss": 0.3266,
+      "step": 12361
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.2500000000000002e-06,
+      "loss": 0.5125,
+      "step": 12362
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.249451754385965e-06,
+      "loss": 0.9986,
+      "step": 12363
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.24890350877193e-06,
+      "loss": 0.9749,
+      "step": 12364
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.248355263157895e-06,
+      "loss": 1.7248,
+      "step": 12365
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.24780701754386e-06,
+      "loss": 0.839,
+      "step": 12366
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.2472587719298248e-06,
+      "loss": 0.6427,
+      "step": 12367
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.24671052631579e-06,
+      "loss": 0.5292,
+      "step": 12368
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.2461622807017544e-06,
+      "loss": 0.179,
+      "step": 12369
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.2456140350877197e-06,
+      "loss": 0.7845,
+      "step": 12370
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.245065789473684e-06,
+      "loss": 0.5098,
+      "step": 12371
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.2445175438596493e-06,
+      "loss": 0.5849,
+      "step": 12372
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.2439692982456146e-06,
+      "loss": 0.8738,
+      "step": 12373
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.243421052631579e-06,
+      "loss": 0.7838,
+      "step": 12374
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.2428728070175442e-06,
+      "loss": 1.1669,
+      "step": 12375
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.2423245614035095e-06,
+      "loss": 0.7838,
+      "step": 12376
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.241776315789474e-06,
+      "loss": 0.2915,
+      "step": 12377
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.241228070175439e-06,
+      "loss": 0.3254,
+      "step": 12378
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.2406798245614035e-06,
+      "loss": 0.2397,
+      "step": 12379
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.2401315789473688e-06,
+      "loss": 0.4016,
+      "step": 12380
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.2395833333333336e-06,
+      "loss": 0.3316,
+      "step": 12381
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.2390350877192984e-06,
+      "loss": 0.5895,
+      "step": 12382
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.2384868421052632e-06,
+      "loss": 0.1207,
+      "step": 12383
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.2379385964912285e-06,
+      "loss": 0.1831,
+      "step": 12384
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.2373903508771933e-06,
+      "loss": 0.2098,
+      "step": 12385
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.236842105263158e-06,
+      "loss": 0.6153,
+      "step": 12386
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.2362938596491234e-06,
+      "loss": 0.4354,
+      "step": 12387
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.2357456140350878e-06,
+      "loss": 0.6055,
+      "step": 12388
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.235197368421053e-06,
+      "loss": 0.8433,
+      "step": 12389
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.2346491228070174e-06,
+      "loss": 0.6463,
+      "step": 12390
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.2341008771929827e-06,
+      "loss": 1.2937,
+      "step": 12391
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.233552631578948e-06,
+      "loss": 0.5442,
+      "step": 12392
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.2330043859649123e-06,
+      "loss": 0.2436,
+      "step": 12393
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.2324561403508776e-06,
+      "loss": 0.4525,
+      "step": 12394
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.231907894736843e-06,
+      "loss": 0.7957,
+      "step": 12395
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.2313596491228072e-06,
+      "loss": 0.3377,
+      "step": 12396
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.2308114035087725e-06,
+      "loss": 0.5857,
+      "step": 12397
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.230263157894737e-06,
+      "loss": 0.064,
+      "step": 12398
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.229714912280702e-06,
+      "loss": 0.96,
+      "step": 12399
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.229166666666667e-06,
+      "loss": 0.541,
+      "step": 12400
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.2286184210526318e-06,
+      "loss": 1.0328,
+      "step": 12401
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.2280701754385966e-06,
+      "loss": 0.5356,
+      "step": 12402
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.227521929824562e-06,
+      "loss": 0.2066,
+      "step": 12403
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.2269736842105267e-06,
+      "loss": 1.0358,
+      "step": 12404
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.2264254385964915e-06,
+      "loss": 0.5881,
+      "step": 12405
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.2258771929824563e-06,
+      "loss": 0.6528,
+      "step": 12406
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.225328947368421e-06,
+      "loss": 0.265,
+      "step": 12407
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.2247807017543864e-06,
+      "loss": 0.2138,
+      "step": 12408
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.224232456140351e-06,
+      "loss": 0.7097,
+      "step": 12409
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.223684210526316e-06,
+      "loss": 0.2422,
+      "step": 12410
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.2231359649122813e-06,
+      "loss": 0.0874,
+      "step": 12411
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.2225877192982457e-06,
+      "loss": 0.71,
+      "step": 12412
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.222039473684211e-06,
+      "loss": 0.8811,
+      "step": 12413
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.2214912280701753e-06,
+      "loss": 0.6074,
+      "step": 12414
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.2209429824561406e-06,
+      "loss": 0.682,
+      "step": 12415
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.220394736842106e-06,
+      "loss": 0.519,
+      "step": 12416
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.2198464912280702e-06,
+      "loss": 0.5528,
+      "step": 12417
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.2192982456140355e-06,
+      "loss": 1.3494,
+      "step": 12418
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.2187500000000003e-06,
+      "loss": 0.4242,
+      "step": 12419
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.218201754385965e-06,
+      "loss": 0.2919,
+      "step": 12420
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.21765350877193e-06,
+      "loss": 0.3404,
+      "step": 12421
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.2171052631578948e-06,
+      "loss": 0.8844,
+      "step": 12422
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.21655701754386e-06,
+      "loss": 0.5175,
+      "step": 12423
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.216008771929825e-06,
+      "loss": 1.0238,
+      "step": 12424
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.2154605263157897e-06,
+      "loss": 0.3265,
+      "step": 12425
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.2149122807017545e-06,
+      "loss": 0.3534,
+      "step": 12426
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.2143640350877198e-06,
+      "loss": 1.3783,
+      "step": 12427
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.213815789473684e-06,
+      "loss": 0.2992,
+      "step": 12428
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.2132675438596494e-06,
+      "loss": 0.2828,
+      "step": 12429
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.2127192982456147e-06,
+      "loss": 0.8342,
+      "step": 12430
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.212171052631579e-06,
+      "loss": 0.4643,
+      "step": 12431
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.2116228070175443e-06,
+      "loss": 0.659,
+      "step": 12432
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.2110745614035087e-06,
+      "loss": 0.4607,
+      "step": 12433
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.210526315789474e-06,
+      "loss": 0.2456,
+      "step": 12434
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.209978070175439e-06,
+      "loss": 0.1955,
+      "step": 12435
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.2094298245614036e-06,
+      "loss": 1.2514,
+      "step": 12436
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.208881578947369e-06,
+      "loss": 0.1187,
+      "step": 12437
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.2083333333333337e-06,
+      "loss": 0.4137,
+      "step": 12438
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.2077850877192985e-06,
+      "loss": 0.4059,
+      "step": 12439
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.2072368421052637e-06,
+      "loss": 0.2641,
+      "step": 12440
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.206688596491228e-06,
+      "loss": 0.5762,
+      "step": 12441
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.2061403508771934e-06,
+      "loss": 0.2536,
+      "step": 12442
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.2055921052631582e-06,
+      "loss": 0.8306,
+      "step": 12443
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.205043859649123e-06,
+      "loss": 0.1468,
+      "step": 12444
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.204495614035088e-06,
+      "loss": 0.6316,
+      "step": 12445
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.203947368421053e-06,
+      "loss": 0.3211,
+      "step": 12446
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.2033991228070175e-06,
+      "loss": 2.1076,
+      "step": 12447
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.2028508771929828e-06,
+      "loss": 0.7768,
+      "step": 12448
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.2023026315789476e-06,
+      "loss": 0.3474,
+      "step": 12449
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.2017543859649124e-06,
+      "loss": 0.1525,
+      "step": 12450
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.2012061403508777e-06,
+      "loss": 0.7067,
+      "step": 12451
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.200657894736842e-06,
+      "loss": 0.3537,
+      "step": 12452
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.2001096491228073e-06,
+      "loss": 0.3377,
+      "step": 12453
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.1995614035087726e-06,
+      "loss": 0.3827,
+      "step": 12454
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.199013157894737e-06,
+      "loss": 0.4272,
+      "step": 12455
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.198464912280702e-06,
+      "loss": 1.4868,
+      "step": 12456
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.1979166666666666e-06,
+      "loss": 0.0995,
+      "step": 12457
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.197368421052632e-06,
+      "loss": 0.6324,
+      "step": 12458
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.196820175438597e-06,
+      "loss": 0.2611,
+      "step": 12459
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.1962719298245615e-06,
+      "loss": 0.9049,
+      "step": 12460
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.1957236842105268e-06,
+      "loss": 0.5842,
+      "step": 12461
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.1951754385964916e-06,
+      "loss": 0.4274,
+      "step": 12462
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.1946271929824564e-06,
+      "loss": 1.1322,
+      "step": 12463
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.1940789473684212e-06,
+      "loss": 0.3354,
+      "step": 12464
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.193530701754386e-06,
+      "loss": 0.4395,
+      "step": 12465
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.192982456140351e-06,
+      "loss": 0.2649,
+      "step": 12466
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.192434210526316e-06,
+      "loss": 1.3203,
+      "step": 12467
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.191885964912281e-06,
+      "loss": 0.3127,
+      "step": 12468
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.1913377192982458e-06,
+      "loss": 0.2032,
+      "step": 12469
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.190789473684211e-06,
+      "loss": 0.1563,
+      "step": 12470
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.1902412280701754e-06,
+      "loss": 0.621,
+      "step": 12471
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.1896929824561407e-06,
+      "loss": 0.3513,
+      "step": 12472
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.189144736842106e-06,
+      "loss": 1.0056,
+      "step": 12473
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.1885964912280703e-06,
+      "loss": 0.0867,
+      "step": 12474
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.1880482456140356e-06,
+      "loss": 1.2102,
+      "step": 12475
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.1875e-06,
+      "loss": 0.5551,
+      "step": 12476
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.1869517543859652e-06,
+      "loss": 0.2201,
+      "step": 12477
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.1864035087719305e-06,
+      "loss": 0.3104,
+      "step": 12478
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.185855263157895e-06,
+      "loss": 0.9106,
+      "step": 12479
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.18530701754386e-06,
+      "loss": 0.2417,
+      "step": 12480
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.184758771929825e-06,
+      "loss": 1.1731,
+      "step": 12481
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.1842105263157898e-06,
+      "loss": 0.2572,
+      "step": 12482
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.1836622807017546e-06,
+      "loss": 0.3695,
+      "step": 12483
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.1831140350877194e-06,
+      "loss": 0.5257,
+      "step": 12484
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.1825657894736842e-06,
+      "loss": 0.7365,
+      "step": 12485
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.1820175438596495e-06,
+      "loss": 0.2862,
+      "step": 12486
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.1814692982456143e-06,
+      "loss": 0.6859,
+      "step": 12487
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.180921052631579e-06,
+      "loss": 0.7861,
+      "step": 12488
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.1803728070175444e-06,
+      "loss": 0.502,
+      "step": 12489
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.179824561403509e-06,
+      "loss": 0.4075,
+      "step": 12490
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.179276315789474e-06,
+      "loss": 0.9585,
+      "step": 12491
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.1787280701754384e-06,
+      "loss": 0.3373,
+      "step": 12492
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.1781798245614037e-06,
+      "loss": 0.3105,
+      "step": 12493
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.177631578947369e-06,
+      "loss": 0.7193,
+      "step": 12494
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.1770833333333333e-06,
+      "loss": 0.4908,
+      "step": 12495
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.1765350877192986e-06,
+      "loss": 0.7053,
+      "step": 12496
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.175986842105264e-06,
+      "loss": 0.6674,
+      "step": 12497
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.1754385964912282e-06,
+      "loss": 1.2341,
+      "step": 12498
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.1748903508771935e-06,
+      "loss": 0.2291,
+      "step": 12499
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.174342105263158e-06,
+      "loss": 0.4774,
+      "step": 12500
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.173793859649123e-06,
+      "loss": 0.3448,
+      "step": 12501
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.173245614035088e-06,
+      "loss": 0.4142,
+      "step": 12502
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.1726973684210528e-06,
+      "loss": 0.6684,
+      "step": 12503
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.1721491228070176e-06,
+      "loss": 0.6396,
+      "step": 12504
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.171600877192983e-06,
+      "loss": 1.3338,
+      "step": 12505
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.1710526315789477e-06,
+      "loss": 1.2095,
+      "step": 12506
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.1705043859649125e-06,
+      "loss": 0.0721,
+      "step": 12507
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.1699561403508773e-06,
+      "loss": 0.1225,
+      "step": 12508
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.169407894736842e-06,
+      "loss": 0.1496,
+      "step": 12509
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.1688596491228074e-06,
+      "loss": 0.9207,
+      "step": 12510
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.168311403508772e-06,
+      "loss": 0.9689,
+      "step": 12511
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.167763157894737e-06,
+      "loss": 1.4451,
+      "step": 12512
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.1672149122807023e-06,
+      "loss": 1.6441,
+      "step": 12513
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.1666666666666667e-06,
+      "loss": 0.6234,
+      "step": 12514
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.166118421052632e-06,
+      "loss": 1.1433,
+      "step": 12515
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.165570175438597e-06,
+      "loss": 0.5808,
+      "step": 12516
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.1650219298245616e-06,
+      "loss": 0.1592,
+      "step": 12517
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.164473684210527e-06,
+      "loss": 0.1174,
+      "step": 12518
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.1639254385964912e-06,
+      "loss": 0.735,
+      "step": 12519
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.1633771929824565e-06,
+      "loss": 0.3283,
+      "step": 12520
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.1628289473684213e-06,
+      "loss": 0.416,
+      "step": 12521
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.162280701754386e-06,
+      "loss": 1.262,
+      "step": 12522
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.1617324561403514e-06,
+      "loss": 0.4242,
+      "step": 12523
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.1611842105263162e-06,
+      "loss": 1.1458,
+      "step": 12524
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.160635964912281e-06,
+      "loss": 0.8439,
+      "step": 12525
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.160087719298246e-06,
+      "loss": 0.3539,
+      "step": 12526
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.1595394736842107e-06,
+      "loss": 0.4299,
+      "step": 12527
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.1589912280701755e-06,
+      "loss": 0.9139,
+      "step": 12528
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.1584429824561408e-06,
+      "loss": 0.5427,
+      "step": 12529
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.157894736842105e-06,
+      "loss": 0.8277,
+      "step": 12530
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.1573464912280704e-06,
+      "loss": 1.0743,
+      "step": 12531
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.1567982456140357e-06,
+      "loss": 0.2217,
+      "step": 12532
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.15625e-06,
+      "loss": 0.8888,
+      "step": 12533
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.1557017543859653e-06,
+      "loss": 0.7902,
+      "step": 12534
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.1551535087719297e-06,
+      "loss": 0.8719,
+      "step": 12535
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.154605263157895e-06,
+      "loss": 0.4987,
+      "step": 12536
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.15405701754386e-06,
+      "loss": 0.476,
+      "step": 12537
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1535087719298246e-06,
+      "loss": 0.991,
+      "step": 12538
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.15296052631579e-06,
+      "loss": 0.9865,
+      "step": 12539
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1524122807017547e-06,
+      "loss": 0.3572,
+      "step": 12540
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1518640350877195e-06,
+      "loss": 0.6006,
+      "step": 12541
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1513157894736848e-06,
+      "loss": 0.286,
+      "step": 12542
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.150767543859649e-06,
+      "loss": 0.5574,
+      "step": 12543
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1502192982456144e-06,
+      "loss": 0.8402,
+      "step": 12544
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1496710526315792e-06,
+      "loss": 1.059,
+      "step": 12545
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.149122807017544e-06,
+      "loss": 0.4865,
+      "step": 12546
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.148574561403509e-06,
+      "loss": 0.3114,
+      "step": 12547
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.148026315789474e-06,
+      "loss": 0.2965,
+      "step": 12548
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1474780701754385e-06,
+      "loss": 0.6579,
+      "step": 12549
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1469298245614038e-06,
+      "loss": 0.288,
+      "step": 12550
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.146381578947369e-06,
+      "loss": 1.0833,
+      "step": 12551
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1458333333333334e-06,
+      "loss": 0.1398,
+      "step": 12552
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1452850877192987e-06,
+      "loss": 1.3364,
+      "step": 12553
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.144736842105263e-06,
+      "loss": 0.5464,
+      "step": 12554
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1441885964912283e-06,
+      "loss": 0.9584,
+      "step": 12555
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1436403508771936e-06,
+      "loss": 1.0819,
+      "step": 12556
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.143092105263158e-06,
+      "loss": 0.1667,
+      "step": 12557
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.1425438596491232e-06,
+      "loss": 0.4089,
+      "step": 12558
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.141995614035088e-06,
+      "loss": 0.5068,
+      "step": 12559
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.141447368421053e-06,
+      "loss": 0.444,
+      "step": 12560
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.140899122807018e-06,
+      "loss": 0.2296,
+      "step": 12561
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.1403508771929825e-06,
+      "loss": 0.4181,
+      "step": 12562
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.1398026315789478e-06,
+      "loss": 0.2372,
+      "step": 12563
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.1392543859649126e-06,
+      "loss": 0.5241,
+      "step": 12564
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.1387061403508774e-06,
+      "loss": 0.5843,
+      "step": 12565
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.1381578947368422e-06,
+      "loss": 0.4925,
+      "step": 12566
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.1376096491228075e-06,
+      "loss": 0.3161,
+      "step": 12567
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.137061403508772e-06,
+      "loss": 0.8559,
+      "step": 12568
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.136513157894737e-06,
+      "loss": 0.1711,
+      "step": 12569
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.135964912280702e-06,
+      "loss": 0.1479,
+      "step": 12570
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.1354166666666668e-06,
+      "loss": 0.4959,
+      "step": 12571
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.134868421052632e-06,
+      "loss": 0.2848,
+      "step": 12572
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.1343201754385964e-06,
+      "loss": 0.8182,
+      "step": 12573
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.1337719298245617e-06,
+      "loss": 0.7406,
+      "step": 12574
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.133223684210527e-06,
+      "loss": 0.9403,
+      "step": 12575
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.1326754385964913e-06,
+      "loss": 0.43,
+      "step": 12576
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.1321271929824566e-06,
+      "loss": 0.5496,
+      "step": 12577
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.131578947368421e-06,
+      "loss": 0.3483,
+      "step": 12578
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.1310307017543862e-06,
+      "loss": 0.8701,
+      "step": 12579
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.1304824561403515e-06,
+      "loss": 0.5495,
+      "step": 12580
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.129934210526316e-06,
+      "loss": 0.2899,
+      "step": 12581
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.129385964912281e-06,
+      "loss": 0.0428,
+      "step": 12582
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.128837719298246e-06,
+      "loss": 1.4969,
+      "step": 12583
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.1282894736842108e-06,
+      "loss": 0.2661,
+      "step": 12584
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.1277412280701756e-06,
+      "loss": 0.4673,
+      "step": 12585
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.1271929824561404e-06,
+      "loss": 0.3752,
+      "step": 12586
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.1266447368421053e-06,
+      "loss": 0.9827,
+      "step": 12587
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.1260964912280705e-06,
+      "loss": 0.6048,
+      "step": 12588
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.1255482456140353e-06,
+      "loss": 0.3327,
+      "step": 12589
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.125e-06,
+      "loss": 0.4988,
+      "step": 12590
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.1244517543859654e-06,
+      "loss": 0.099,
+      "step": 12591
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.12390350877193e-06,
+      "loss": 0.9192,
+      "step": 12592
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.123355263157895e-06,
+      "loss": 0.2137,
+      "step": 12593
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.1228070175438603e-06,
+      "loss": 0.6108,
+      "step": 12594
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.1222587719298247e-06,
+      "loss": 0.776,
+      "step": 12595
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.12171052631579e-06,
+      "loss": 1.0127,
+      "step": 12596
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.1211622807017543e-06,
+      "loss": 0.3434,
+      "step": 12597
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.1206140350877196e-06,
+      "loss": 0.5014,
+      "step": 12598
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.120065789473685e-06,
+      "loss": 0.7616,
+      "step": 12599
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.1195175438596492e-06,
+      "loss": 0.7328,
+      "step": 12600
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.1189692982456145e-06,
+      "loss": 0.8878,
+      "step": 12601
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.1184210526315793e-06,
+      "loss": 0.5414,
+      "step": 12602
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.117872807017544e-06,
+      "loss": 0.5012,
+      "step": 12603
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.117324561403509e-06,
+      "loss": 0.2634,
+      "step": 12604
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.116776315789474e-06,
+      "loss": 0.6983,
+      "step": 12605
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.1162280701754386e-06,
+      "loss": 0.1623,
+      "step": 12606
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.115679824561404e-06,
+      "loss": 0.3753,
+      "step": 12607
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.1151315789473687e-06,
+      "loss": 0.4688,
+      "step": 12608
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.1145833333333335e-06,
+      "loss": 0.5591,
+      "step": 12609
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.1140350877192988e-06,
+      "loss": 0.2026,
+      "step": 12610
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.113486842105263e-06,
+      "loss": 0.5042,
+      "step": 12611
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.1129385964912284e-06,
+      "loss": 0.5369,
+      "step": 12612
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.112390350877193e-06,
+      "loss": 0.3146,
+      "step": 12613
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.111842105263158e-06,
+      "loss": 0.2536,
+      "step": 12614
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.1112938596491233e-06,
+      "loss": 0.6251,
+      "step": 12615
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.1107456140350877e-06,
+      "loss": 0.2563,
+      "step": 12616
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.110197368421053e-06,
+      "loss": 0.4517,
+      "step": 12617
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.109649122807018e-06,
+      "loss": 1.3491,
+      "step": 12618
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.1091008771929826e-06,
+      "loss": 1.1961,
+      "step": 12619
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.108552631578948e-06,
+      "loss": 0.1704,
+      "step": 12620
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.1080043859649123e-06,
+      "loss": 0.3023,
+      "step": 12621
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.1074561403508775e-06,
+      "loss": 0.5143,
+      "step": 12622
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.1069078947368423e-06,
+      "loss": 0.3515,
+      "step": 12623
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.106359649122807e-06,
+      "loss": 0.3467,
+      "step": 12624
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.1058114035087724e-06,
+      "loss": 0.7643,
+      "step": 12625
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.1052631578947372e-06,
+      "loss": 0.3827,
+      "step": 12626
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.104714912280702e-06,
+      "loss": 0.6987,
+      "step": 12627
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.104166666666667e-06,
+      "loss": 0.5807,
+      "step": 12628
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.1036184210526317e-06,
+      "loss": 0.7283,
+      "step": 12629
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.1030701754385965e-06,
+      "loss": 0.0583,
+      "step": 12630
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.1025219298245618e-06,
+      "loss": 0.6462,
+      "step": 12631
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.101973684210526e-06,
+      "loss": 0.2198,
+      "step": 12632
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.1014254385964914e-06,
+      "loss": 0.5789,
+      "step": 12633
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.1008771929824567e-06,
+      "loss": 0.7126,
+      "step": 12634
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.100328947368421e-06,
+      "loss": 0.7863,
+      "step": 12635
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.0997807017543863e-06,
+      "loss": 0.9397,
+      "step": 12636
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.0992324561403516e-06,
+      "loss": 0.4375,
+      "step": 12637
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.098684210526316e-06,
+      "loss": 0.1713,
+      "step": 12638
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.0981359649122812e-06,
+      "loss": 0.5814,
+      "step": 12639
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.0975877192982456e-06,
+      "loss": 1.1423,
+      "step": 12640
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.097039473684211e-06,
+      "loss": 0.2307,
+      "step": 12641
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.0964912280701757e-06,
+      "loss": 0.4957,
+      "step": 12642
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.0959429824561405e-06,
+      "loss": 0.3135,
+      "step": 12643
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.0953947368421058e-06,
+      "loss": 0.5351,
+      "step": 12644
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.0948464912280706e-06,
+      "loss": 0.1814,
+      "step": 12645
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.0942982456140354e-06,
+      "loss": 0.6744,
+      "step": 12646
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.0937500000000002e-06,
+      "loss": 0.8562,
+      "step": 12647
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.093201754385965e-06,
+      "loss": 0.5438,
+      "step": 12648
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.09265350877193e-06,
+      "loss": 0.3596,
+      "step": 12649
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.092105263157895e-06,
+      "loss": 0.7988,
+      "step": 12650
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.0915570175438595e-06,
+      "loss": 0.2787,
+      "step": 12651
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.0910087719298248e-06,
+      "loss": 0.2001,
+      "step": 12652
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.09046052631579e-06,
+      "loss": 0.4041,
+      "step": 12653
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.0899122807017544e-06,
+      "loss": 0.2126,
+      "step": 12654
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.0893640350877197e-06,
+      "loss": 0.8765,
+      "step": 12655
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.088815789473684e-06,
+      "loss": 0.1157,
+      "step": 12656
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.0882675438596493e-06,
+      "loss": 0.3157,
+      "step": 12657
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.0877192982456146e-06,
+      "loss": 0.1871,
+      "step": 12658
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.087171052631579e-06,
+      "loss": 0.2386,
+      "step": 12659
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.0866228070175442e-06,
+      "loss": 0.1028,
+      "step": 12660
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.086074561403509e-06,
+      "loss": 0.4281,
+      "step": 12661
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.085526315789474e-06,
+      "loss": 0.8082,
+      "step": 12662
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.084978070175439e-06,
+      "loss": 0.7712,
+      "step": 12663
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.0844298245614035e-06,
+      "loss": 0.8242,
+      "step": 12664
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.0838815789473688e-06,
+      "loss": 0.4421,
+      "step": 12665
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.0833333333333336e-06,
+      "loss": 1.1995,
+      "step": 12666
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.0827850877192984e-06,
+      "loss": 0.4398,
+      "step": 12667
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.0822368421052632e-06,
+      "loss": 1.1052,
+      "step": 12668
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.0816885964912285e-06,
+      "loss": 0.356,
+      "step": 12669
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.081140350877193e-06,
+      "loss": 0.7503,
+      "step": 12670
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.080592105263158e-06,
+      "loss": 0.4054,
+      "step": 12671
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.0800438596491234e-06,
+      "loss": 1.3022,
+      "step": 12672
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.079495614035088e-06,
+      "loss": 0.6903,
+      "step": 12673
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.078947368421053e-06,
+      "loss": 0.3601,
+      "step": 12674
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.0783991228070174e-06,
+      "loss": 0.8861,
+      "step": 12675
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.0778508771929827e-06,
+      "loss": 0.0811,
+      "step": 12676
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.077302631578948e-06,
+      "loss": 0.6878,
+      "step": 12677
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.0767543859649123e-06,
+      "loss": 0.4811,
+      "step": 12678
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.0762061403508776e-06,
+      "loss": 0.6721,
+      "step": 12679
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.0756578947368424e-06,
+      "loss": 1.4681,
+      "step": 12680
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.0751096491228072e-06,
+      "loss": 0.4294,
+      "step": 12681
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.0745614035087725e-06,
+      "loss": 0.6585,
+      "step": 12682
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.074013157894737e-06,
+      "loss": 0.1982,
+      "step": 12683
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.073464912280702e-06,
+      "loss": 0.1932,
+      "step": 12684
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.072916666666667e-06,
+      "loss": 0.4971,
+      "step": 12685
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.0723684210526318e-06,
+      "loss": 0.2528,
+      "step": 12686
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.0718201754385966e-06,
+      "loss": 0.5013,
+      "step": 12687
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.071271929824562e-06,
+      "loss": 0.4538,
+      "step": 12688
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.0707236842105263e-06,
+      "loss": 1.0214,
+      "step": 12689
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.0701754385964915e-06,
+      "loss": 0.8577,
+      "step": 12690
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.0696271929824563e-06,
+      "loss": 0.2884,
+      "step": 12691
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.069078947368421e-06,
+      "loss": 0.1949,
+      "step": 12692
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.0685307017543864e-06,
+      "loss": 0.8948,
+      "step": 12693
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.067982456140351e-06,
+      "loss": 0.2967,
+      "step": 12694
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.067434210526316e-06,
+      "loss": 0.8077,
+      "step": 12695
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.0668859649122813e-06,
+      "loss": 0.9998,
+      "step": 12696
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.0663377192982457e-06,
+      "loss": 0.3052,
+      "step": 12697
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.065789473684211e-06,
+      "loss": 0.2973,
+      "step": 12698
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.0652412280701754e-06,
+      "loss": 0.8605,
+      "step": 12699
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.0646929824561406e-06,
+      "loss": 0.4257,
+      "step": 12700
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.064144736842106e-06,
+      "loss": 0.7899,
+      "step": 12701
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.0635964912280703e-06,
+      "loss": 1.0513,
+      "step": 12702
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.0630482456140355e-06,
+      "loss": 0.4797,
+      "step": 12703
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.0625000000000003e-06,
+      "loss": 0.4634,
+      "step": 12704
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.061951754385965e-06,
+      "loss": 0.7541,
+      "step": 12705
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.06140350877193e-06,
+      "loss": 0.6723,
+      "step": 12706
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.060855263157895e-06,
+      "loss": 0.6691,
+      "step": 12707
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.06030701754386e-06,
+      "loss": 0.2078,
+      "step": 12708
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.059758771929825e-06,
+      "loss": 0.9957,
+      "step": 12709
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.0592105263157897e-06,
+      "loss": 0.1303,
+      "step": 12710
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.0586622807017545e-06,
+      "loss": 0.0588,
+      "step": 12711
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.0581140350877198e-06,
+      "loss": 0.8416,
+      "step": 12712
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.057565789473684e-06,
+      "loss": 0.4238,
+      "step": 12713
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.0570175438596494e-06,
+      "loss": 0.9779,
+      "step": 12714
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.0564692982456147e-06,
+      "loss": 0.2249,
+      "step": 12715
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.055921052631579e-06,
+      "loss": 0.0937,
+      "step": 12716
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.0553728070175443e-06,
+      "loss": 0.4865,
+      "step": 12717
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.0548245614035087e-06,
+      "loss": 0.5203,
+      "step": 12718
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.054276315789474e-06,
+      "loss": 0.4687,
+      "step": 12719
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.0537280701754392e-06,
+      "loss": 0.2619,
+      "step": 12720
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.0531798245614036e-06,
+      "loss": 0.5079,
+      "step": 12721
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.052631578947369e-06,
+      "loss": 0.2324,
+      "step": 12722
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.0520833333333337e-06,
+      "loss": 0.3728,
+      "step": 12723
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.0515350877192985e-06,
+      "loss": 0.4955,
+      "step": 12724
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.0509868421052633e-06,
+      "loss": 0.269,
+      "step": 12725
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.050438596491228e-06,
+      "loss": 1.1867,
+      "step": 12726
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.0498903508771934e-06,
+      "loss": 0.2769,
+      "step": 12727
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.0493421052631582e-06,
+      "loss": 0.729,
+      "step": 12728
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.048793859649123e-06,
+      "loss": 0.2215,
+      "step": 12729
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.048245614035088e-06,
+      "loss": 0.5171,
+      "step": 12730
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.047697368421053e-06,
+      "loss": 0.3116,
+      "step": 12731
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.0471491228070175e-06,
+      "loss": 0.8725,
+      "step": 12732
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.0466008771929828e-06,
+      "loss": 0.6245,
+      "step": 12733
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.046052631578947e-06,
+      "loss": 0.805,
+      "step": 12734
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.0455043859649124e-06,
+      "loss": 0.3424,
+      "step": 12735
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.0449561403508777e-06,
+      "loss": 1.2018,
+      "step": 12736
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.044407894736842e-06,
+      "loss": 0.1803,
+      "step": 12737
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.0438596491228073e-06,
+      "loss": 0.4477,
+      "step": 12738
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.0433114035087726e-06,
+      "loss": 0.465,
+      "step": 12739
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.042763157894737e-06,
+      "loss": 0.2379,
+      "step": 12740
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.0422149122807022e-06,
+      "loss": 1.1688,
+      "step": 12741
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.0416666666666666e-06,
+      "loss": 0.7698,
+      "step": 12742
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.041118421052632e-06,
+      "loss": 1.0891,
+      "step": 12743
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.0405701754385967e-06,
+      "loss": 0.3007,
+      "step": 12744
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.0400219298245615e-06,
+      "loss": 0.3317,
+      "step": 12745
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.0394736842105268e-06,
+      "loss": 0.4196,
+      "step": 12746
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.0389254385964916e-06,
+      "loss": 0.5261,
+      "step": 12747
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.0383771929824564e-06,
+      "loss": 0.6815,
+      "step": 12748
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.0378289473684212e-06,
+      "loss": 0.5051,
+      "step": 12749
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.037280701754386e-06,
+      "loss": 0.5665,
+      "step": 12750
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.036732456140351e-06,
+      "loss": 0.5661,
+      "step": 12751
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.036184210526316e-06,
+      "loss": 0.8237,
+      "step": 12752
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.0356359649122805e-06,
+      "loss": 0.2882,
+      "step": 12753
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.035087719298246e-06,
+      "loss": 0.6676,
+      "step": 12754
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.034539473684211e-06,
+      "loss": 0.5408,
+      "step": 12755
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.0339912280701754e-06,
+      "loss": 0.1749,
+      "step": 12756
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.0334429824561407e-06,
+      "loss": 0.1754,
+      "step": 12757
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.032894736842106e-06,
+      "loss": 0.7566,
+      "step": 12758
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.0323464912280703e-06,
+      "loss": 0.0776,
+      "step": 12759
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.0317982456140356e-06,
+      "loss": 1.3883,
+      "step": 12760
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.03125e-06,
+      "loss": 0.5091,
+      "step": 12761
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.0307017543859652e-06,
+      "loss": 0.7126,
+      "step": 12762
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.03015350877193e-06,
+      "loss": 0.1868,
+      "step": 12763
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.029605263157895e-06,
+      "loss": 0.5744,
+      "step": 12764
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.02905701754386e-06,
+      "loss": 0.9983,
+      "step": 12765
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.028508771929825e-06,
+      "loss": 0.346,
+      "step": 12766
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.0279605263157898e-06,
+      "loss": 0.3465,
+      "step": 12767
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.0274122807017546e-06,
+      "loss": 0.5118,
+      "step": 12768
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.0268640350877194e-06,
+      "loss": 0.1904,
+      "step": 12769
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.0263157894736843e-06,
+      "loss": 0.7039,
+      "step": 12770
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.0257675438596495e-06,
+      "loss": 0.2563,
+      "step": 12771
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.025219298245614e-06,
+      "loss": 0.6301,
+      "step": 12772
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.024671052631579e-06,
+      "loss": 0.3039,
+      "step": 12773
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.0241228070175444e-06,
+      "loss": 0.4308,
+      "step": 12774
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.023574561403509e-06,
+      "loss": 1.7779,
+      "step": 12775
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.023026315789474e-06,
+      "loss": 0.2827,
+      "step": 12776
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.0224780701754385e-06,
+      "loss": 0.4154,
+      "step": 12777
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.0219298245614037e-06,
+      "loss": 0.2935,
+      "step": 12778
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.021381578947369e-06,
+      "loss": 0.4526,
+      "step": 12779
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.0208333333333334e-06,
+      "loss": 0.5902,
+      "step": 12780
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.0202850877192986e-06,
+      "loss": 0.1781,
+      "step": 12781
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.0197368421052634e-06,
+      "loss": 0.3395,
+      "step": 12782
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.0191885964912282e-06,
+      "loss": 0.1676,
+      "step": 12783
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.0186403508771935e-06,
+      "loss": 0.2553,
+      "step": 12784
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.018092105263158e-06,
+      "loss": 0.5825,
+      "step": 12785
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.017543859649123e-06,
+      "loss": 0.7472,
+      "step": 12786
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.016995614035088e-06,
+      "loss": 0.666,
+      "step": 12787
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.016447368421053e-06,
+      "loss": 0.4963,
+      "step": 12788
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.0158991228070176e-06,
+      "loss": 0.5343,
+      "step": 12789
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.015350877192983e-06,
+      "loss": 0.6457,
+      "step": 12790
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.0148026315789473e-06,
+      "loss": 0.3188,
+      "step": 12791
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.0142543859649125e-06,
+      "loss": 0.2965,
+      "step": 12792
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.0137061403508773e-06,
+      "loss": 0.1048,
+      "step": 12793
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.013157894736842e-06,
+      "loss": 0.2547,
+      "step": 12794
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.0126096491228074e-06,
+      "loss": 0.1885,
+      "step": 12795
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.012061403508772e-06,
+      "loss": 0.9609,
+      "step": 12796
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.011513157894737e-06,
+      "loss": 0.0879,
+      "step": 12797
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.0109649122807023e-06,
+      "loss": 0.6084,
+      "step": 12798
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.0104166666666667e-06,
+      "loss": 0.2595,
+      "step": 12799
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.009868421052632e-06,
+      "loss": 0.5414,
+      "step": 12800
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.009320175438597e-06,
+      "loss": 0.3306,
+      "step": 12801
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.0087719298245616e-06,
+      "loss": 0.4783,
+      "step": 12802
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.008223684210527e-06,
+      "loss": 0.0267,
+      "step": 12803
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.7639942587656346,
+      "eval_loss": 0.7222694158554077,
+      "eval_roc_auc": 0.9396730480495818,
+      "eval_runtime": 184.3533,
+      "eval_samples_per_second": 26.455,
+      "eval_steps_per_second": 3.309,
+      "step": 12803
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.0076754385964913e-06,
+      "loss": 0.3626,
+      "step": 12804
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.0071271929824565e-06,
+      "loss": 0.6312,
+      "step": 12805
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.0065789473684213e-06,
+      "loss": 0.3408,
+      "step": 12806
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.006030701754386e-06,
+      "loss": 0.1203,
+      "step": 12807
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.005482456140351e-06,
+      "loss": 0.4317,
+      "step": 12808
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.0049342105263162e-06,
+      "loss": 0.4335,
+      "step": 12809
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.004385964912281e-06,
+      "loss": 0.1852,
+      "step": 12810
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.003837719298246e-06,
+      "loss": 0.2608,
+      "step": 12811
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.0032894736842107e-06,
+      "loss": 0.9023,
+      "step": 12812
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3.0027412280701755e-06,
+      "loss": 0.784,
+      "step": 12813
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3.0021929824561408e-06,
+      "loss": 1.5447,
+      "step": 12814
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3.001644736842105e-06,
+      "loss": 0.2406,
+      "step": 12815
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3.0010964912280704e-06,
+      "loss": 0.4441,
+      "step": 12816
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3.0005482456140357e-06,
+      "loss": 0.63,
+      "step": 12817
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3e-06,
+      "loss": 0.5473,
+      "step": 12818
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.9994517543859653e-06,
+      "loss": 0.6219,
+      "step": 12819
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.9989035087719297e-06,
+      "loss": 0.1336,
+      "step": 12820
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.998355263157895e-06,
+      "loss": 1.503,
+      "step": 12821
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.9978070175438602e-06,
+      "loss": 0.5864,
+      "step": 12822
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.9972587719298246e-06,
+      "loss": 0.5348,
+      "step": 12823
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.99671052631579e-06,
+      "loss": 0.7685,
+      "step": 12824
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.9961622807017547e-06,
+      "loss": 0.8249,
+      "step": 12825
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.9956140350877195e-06,
+      "loss": 0.749,
+      "step": 12826
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.9950657894736843e-06,
+      "loss": 0.8991,
+      "step": 12827
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.994517543859649e-06,
+      "loss": 1.025,
+      "step": 12828
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.9939692982456144e-06,
+      "loss": 0.7694,
+      "step": 12829
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.9934210526315792e-06,
+      "loss": 1.2762,
+      "step": 12830
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.992872807017544e-06,
+      "loss": 0.3749,
+      "step": 12831
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.992324561403509e-06,
+      "loss": 0.0918,
+      "step": 12832
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.991776315789474e-06,
+      "loss": 0.7941,
+      "step": 12833
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.9912280701754385e-06,
+      "loss": 0.8638,
+      "step": 12834
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.990679824561404e-06,
+      "loss": 0.5594,
+      "step": 12835
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.990131578947369e-06,
+      "loss": 0.902,
+      "step": 12836
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.9895833333333334e-06,
+      "loss": 0.7134,
+      "step": 12837
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.9890350877192987e-06,
+      "loss": 0.4067,
+      "step": 12838
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.988486842105263e-06,
+      "loss": 0.3227,
+      "step": 12839
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.9879385964912283e-06,
+      "loss": 0.7931,
+      "step": 12840
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.9873903508771936e-06,
+      "loss": 0.2163,
+      "step": 12841
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.986842105263158e-06,
+      "loss": 0.6145,
+      "step": 12842
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.9862938596491232e-06,
+      "loss": 0.8846,
+      "step": 12843
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.985745614035088e-06,
+      "loss": 0.4287,
+      "step": 12844
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.985197368421053e-06,
+      "loss": 0.1764,
+      "step": 12845
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.9846491228070177e-06,
+      "loss": 0.3167,
+      "step": 12846
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.9841008771929825e-06,
+      "loss": 0.3027,
+      "step": 12847
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.9835526315789478e-06,
+      "loss": 0.4214,
+      "step": 12848
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.9830043859649126e-06,
+      "loss": 0.0928,
+      "step": 12849
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.9824561403508774e-06,
+      "loss": 0.6877,
+      "step": 12850
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.9819078947368423e-06,
+      "loss": 1.0283,
+      "step": 12851
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.9813596491228075e-06,
+      "loss": 2.0255,
+      "step": 12852
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.980811403508772e-06,
+      "loss": 0.3994,
+      "step": 12853
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.980263157894737e-06,
+      "loss": 0.1969,
+      "step": 12854
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.9797149122807016e-06,
+      "loss": 0.1178,
+      "step": 12855
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.979166666666667e-06,
+      "loss": 0.7572,
+      "step": 12856
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.978618421052632e-06,
+      "loss": 0.6679,
+      "step": 12857
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.9780701754385965e-06,
+      "loss": 0.2983,
+      "step": 12858
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.9775219298245617e-06,
+      "loss": 0.312,
+      "step": 12859
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.976973684210527e-06,
+      "loss": 0.3069,
+      "step": 12860
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.9764254385964913e-06,
+      "loss": 0.2913,
+      "step": 12861
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.9758771929824566e-06,
+      "loss": 0.4083,
+      "step": 12862
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.975328947368421e-06,
+      "loss": 0.9407,
+      "step": 12863
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.9747807017543862e-06,
+      "loss": 0.2554,
+      "step": 12864
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.974232456140351e-06,
+      "loss": 0.7942,
+      "step": 12865
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.973684210526316e-06,
+      "loss": 0.5219,
+      "step": 12866
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.973135964912281e-06,
+      "loss": 1.1946,
+      "step": 12867
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.972587719298246e-06,
+      "loss": 0.4456,
+      "step": 12868
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.972039473684211e-06,
+      "loss": 0.8141,
+      "step": 12869
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.9714912280701756e-06,
+      "loss": 0.4927,
+      "step": 12870
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.9709429824561404e-06,
+      "loss": 0.2341,
+      "step": 12871
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.9703947368421053e-06,
+      "loss": 0.9406,
+      "step": 12872
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.9698464912280705e-06,
+      "loss": 0.7073,
+      "step": 12873
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.969298245614035e-06,
+      "loss": 1.053,
+      "step": 12874
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.96875e-06,
+      "loss": 0.3005,
+      "step": 12875
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.9682017543859654e-06,
+      "loss": 0.7491,
+      "step": 12876
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.96765350877193e-06,
+      "loss": 0.4384,
+      "step": 12877
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.967105263157895e-06,
+      "loss": 0.86,
+      "step": 12878
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.9665570175438603e-06,
+      "loss": 0.1577,
+      "step": 12879
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.9660087719298247e-06,
+      "loss": 1.2503,
+      "step": 12880
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.96546052631579e-06,
+      "loss": 0.801,
+      "step": 12881
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.9649122807017544e-06,
+      "loss": 0.1509,
+      "step": 12882
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.9643640350877196e-06,
+      "loss": 1.0389,
+      "step": 12883
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.963815789473685e-06,
+      "loss": 0.5212,
+      "step": 12884
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.9632675438596493e-06,
+      "loss": 0.5092,
+      "step": 12885
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.9627192982456145e-06,
+      "loss": 0.1998,
+      "step": 12886
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.9621710526315793e-06,
+      "loss": 0.5114,
+      "step": 12887
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.961622807017544e-06,
+      "loss": 0.4141,
+      "step": 12888
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.961074561403509e-06,
+      "loss": 0.0886,
+      "step": 12889
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.960526315789474e-06,
+      "loss": 0.1692,
+      "step": 12890
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.9599780701754386e-06,
+      "loss": 0.5894,
+      "step": 12891
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.959429824561404e-06,
+      "loss": 0.5351,
+      "step": 12892
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.9588815789473687e-06,
+      "loss": 0.3672,
+      "step": 12893
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.9583333333333335e-06,
+      "loss": 0.7834,
+      "step": 12894
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.9577850877192988e-06,
+      "loss": 1.0192,
+      "step": 12895
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.957236842105263e-06,
+      "loss": 0.0562,
+      "step": 12896
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.9566885964912284e-06,
+      "loss": 0.5708,
+      "step": 12897
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.956140350877193e-06,
+      "loss": 0.717,
+      "step": 12898
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.955592105263158e-06,
+      "loss": 0.6549,
+      "step": 12899
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.9550438596491233e-06,
+      "loss": 0.5462,
+      "step": 12900
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.9544956140350877e-06,
+      "loss": 0.544,
+      "step": 12901
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.953947368421053e-06,
+      "loss": 0.3975,
+      "step": 12902
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.9533991228070182e-06,
+      "loss": 1.2393,
+      "step": 12903
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.9528508771929826e-06,
+      "loss": 0.469,
+      "step": 12904
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.952302631578948e-06,
+      "loss": 0.4187,
+      "step": 12905
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.9517543859649123e-06,
+      "loss": 2.089,
+      "step": 12906
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.9512061403508775e-06,
+      "loss": 1.1602,
+      "step": 12907
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.9506578947368423e-06,
+      "loss": 0.7069,
+      "step": 12908
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.950109649122807e-06,
+      "loss": 0.1911,
+      "step": 12909
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.949561403508772e-06,
+      "loss": 1.0917,
+      "step": 12910
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.9490131578947372e-06,
+      "loss": 0.8185,
+      "step": 12911
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.948464912280702e-06,
+      "loss": 0.4279,
+      "step": 12912
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.947916666666667e-06,
+      "loss": 0.2595,
+      "step": 12913
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.9473684210526317e-06,
+      "loss": 0.1555,
+      "step": 12914
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.9468201754385965e-06,
+      "loss": 1.293,
+      "step": 12915
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.9462719298245618e-06,
+      "loss": 0.6464,
+      "step": 12916
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.945723684210526e-06,
+      "loss": 0.2366,
+      "step": 12917
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.9451754385964914e-06,
+      "loss": 0.2144,
+      "step": 12918
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.9446271929824567e-06,
+      "loss": 0.2034,
+      "step": 12919
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.944078947368421e-06,
+      "loss": 0.665,
+      "step": 12920
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.9435307017543863e-06,
+      "loss": 1.2179,
+      "step": 12921
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.9429824561403516e-06,
+      "loss": 0.3375,
+      "step": 12922
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.942434210526316e-06,
+      "loss": 0.2428,
+      "step": 12923
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.9418859649122812e-06,
+      "loss": 1.1038,
+      "step": 12924
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.9413377192982456e-06,
+      "loss": 1.0591,
+      "step": 12925
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.940789473684211e-06,
+      "loss": 0.4981,
+      "step": 12926
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.9402412280701757e-06,
+      "loss": 0.2334,
+      "step": 12927
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.9396929824561405e-06,
+      "loss": 0.8854,
+      "step": 12928
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.9391447368421054e-06,
+      "loss": 0.5387,
+      "step": 12929
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.9385964912280706e-06,
+      "loss": 0.2925,
+      "step": 12930
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.9380482456140354e-06,
+      "loss": 1.1241,
+      "step": 12931
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.9375000000000003e-06,
+      "loss": 0.3595,
+      "step": 12932
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.936951754385965e-06,
+      "loss": 0.573,
+      "step": 12933
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.93640350877193e-06,
+      "loss": 0.7617,
+      "step": 12934
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.935855263157895e-06,
+      "loss": 0.2271,
+      "step": 12935
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.9353070175438595e-06,
+      "loss": 0.539,
+      "step": 12936
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.934758771929825e-06,
+      "loss": 0.2888,
+      "step": 12937
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.93421052631579e-06,
+      "loss": 0.8114,
+      "step": 12938
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.9336622807017544e-06,
+      "loss": 0.7806,
+      "step": 12939
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.9331140350877197e-06,
+      "loss": 0.5932,
+      "step": 12940
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.932565789473684e-06,
+      "loss": 0.2385,
+      "step": 12941
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.9320175438596493e-06,
+      "loss": 1.2045,
+      "step": 12942
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.9314692982456146e-06,
+      "loss": 0.3583,
+      "step": 12943
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.930921052631579e-06,
+      "loss": 0.3521,
+      "step": 12944
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.9303728070175442e-06,
+      "loss": 0.1566,
+      "step": 12945
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.929824561403509e-06,
+      "loss": 0.4464,
+      "step": 12946
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.929276315789474e-06,
+      "loss": 0.2597,
+      "step": 12947
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.9287280701754387e-06,
+      "loss": 0.7734,
+      "step": 12948
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.9281798245614035e-06,
+      "loss": 0.5478,
+      "step": 12949
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.927631578947369e-06,
+      "loss": 0.1133,
+      "step": 12950
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.9270833333333336e-06,
+      "loss": 1.2413,
+      "step": 12951
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.9265350877192984e-06,
+      "loss": 0.5115,
+      "step": 12952
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.9259868421052633e-06,
+      "loss": 0.1513,
+      "step": 12953
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.9254385964912285e-06,
+      "loss": 0.3804,
+      "step": 12954
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.924890350877193e-06,
+      "loss": 0.3927,
+      "step": 12955
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.924342105263158e-06,
+      "loss": 0.7251,
+      "step": 12956
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.9237938596491234e-06,
+      "loss": 0.1974,
+      "step": 12957
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.923245614035088e-06,
+      "loss": 0.757,
+      "step": 12958
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.922697368421053e-06,
+      "loss": 0.1583,
+      "step": 12959
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.9221491228070175e-06,
+      "loss": 0.622,
+      "step": 12960
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.9216008771929827e-06,
+      "loss": 0.2683,
+      "step": 12961
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.921052631578948e-06,
+      "loss": 0.2976,
+      "step": 12962
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.9205043859649124e-06,
+      "loss": 0.3596,
+      "step": 12963
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.9199561403508776e-06,
+      "loss": 0.3077,
+      "step": 12964
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.9194078947368424e-06,
+      "loss": 0.122,
+      "step": 12965
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.9188596491228073e-06,
+      "loss": 0.7065,
+      "step": 12966
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.918311403508772e-06,
+      "loss": 0.7065,
+      "step": 12967
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.917763157894737e-06,
+      "loss": 0.3576,
+      "step": 12968
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.917214912280702e-06,
+      "loss": 0.8594,
+      "step": 12969
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.916666666666667e-06,
+      "loss": 0.51,
+      "step": 12970
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.916118421052632e-06,
+      "loss": 0.3697,
+      "step": 12971
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.9155701754385966e-06,
+      "loss": 0.3659,
+      "step": 12972
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.915021929824562e-06,
+      "loss": 0.7353,
+      "step": 12973
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.9144736842105263e-06,
+      "loss": 0.7039,
+      "step": 12974
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.9139254385964915e-06,
+      "loss": 0.487,
+      "step": 12975
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.9133771929824563e-06,
+      "loss": 0.1548,
+      "step": 12976
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.912828947368421e-06,
+      "loss": 0.504,
+      "step": 12977
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.9122807017543864e-06,
+      "loss": 0.6685,
+      "step": 12978
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.911732456140351e-06,
+      "loss": 0.4131,
+      "step": 12979
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.911184210526316e-06,
+      "loss": 0.8763,
+      "step": 12980
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.9106359649122813e-06,
+      "loss": 0.7788,
+      "step": 12981
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.9100877192982457e-06,
+      "loss": 0.1903,
+      "step": 12982
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.909539473684211e-06,
+      "loss": 0.4794,
+      "step": 12983
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.9089912280701754e-06,
+      "loss": 0.1115,
+      "step": 12984
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.9084429824561406e-06,
+      "loss": 0.4457,
+      "step": 12985
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.907894736842106e-06,
+      "loss": 0.6134,
+      "step": 12986
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.9073464912280703e-06,
+      "loss": 1.0415,
+      "step": 12987
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.9067982456140355e-06,
+      "loss": 0.1246,
+      "step": 12988
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.9062500000000003e-06,
+      "loss": 1.1548,
+      "step": 12989
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.905701754385965e-06,
+      "loss": 0.2271,
+      "step": 12990
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.90515350877193e-06,
+      "loss": 0.8085,
+      "step": 12991
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.904605263157895e-06,
+      "loss": 0.315,
+      "step": 12992
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.9040570175438596e-06,
+      "loss": 0.5564,
+      "step": 12993
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.903508771929825e-06,
+      "loss": 1.1102,
+      "step": 12994
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.9029605263157897e-06,
+      "loss": 0.157,
+      "step": 12995
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.9024122807017545e-06,
+      "loss": 0.6657,
+      "step": 12996
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.9018640350877198e-06,
+      "loss": 0.4771,
+      "step": 12997
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.901315789473684e-06,
+      "loss": 0.2491,
+      "step": 12998
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.9007675438596494e-06,
+      "loss": 0.1067,
+      "step": 12999
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.9002192982456147e-06,
+      "loss": 1.0778,
+      "step": 13000
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.899671052631579e-06,
+      "loss": 0.5871,
+      "step": 13001
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.8991228070175443e-06,
+      "loss": 0.855,
+      "step": 13002
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.8985745614035087e-06,
+      "loss": 0.5523,
+      "step": 13003
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.898026315789474e-06,
+      "loss": 0.2926,
+      "step": 13004
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.8974780701754392e-06,
+      "loss": 0.5952,
+      "step": 13005
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.8969298245614036e-06,
+      "loss": 0.5511,
+      "step": 13006
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.896381578947369e-06,
+      "loss": 0.561,
+      "step": 13007
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.8958333333333337e-06,
+      "loss": 0.7692,
+      "step": 13008
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.8952850877192985e-06,
+      "loss": 0.3551,
+      "step": 13009
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.8947368421052634e-06,
+      "loss": 0.54,
+      "step": 13010
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.894188596491228e-06,
+      "loss": 0.3727,
+      "step": 13011
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.893640350877193e-06,
+      "loss": 0.6316,
+      "step": 13012
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.8930921052631582e-06,
+      "loss": 0.7445,
+      "step": 13013
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.892543859649123e-06,
+      "loss": 0.3044,
+      "step": 13014
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.891995614035088e-06,
+      "loss": 0.5646,
+      "step": 13015
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.891447368421053e-06,
+      "loss": 0.8665,
+      "step": 13016
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.8908991228070175e-06,
+      "loss": 0.3188,
+      "step": 13017
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.890350877192983e-06,
+      "loss": 0.1415,
+      "step": 13018
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.889802631578947e-06,
+      "loss": 0.5484,
+      "step": 13019
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.8892543859649124e-06,
+      "loss": 0.9662,
+      "step": 13020
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.8887061403508777e-06,
+      "loss": 0.6983,
+      "step": 13021
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.888157894736842e-06,
+      "loss": 0.6097,
+      "step": 13022
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.8876096491228073e-06,
+      "loss": 0.4839,
+      "step": 13023
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.8870614035087726e-06,
+      "loss": 0.3765,
+      "step": 13024
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.886513157894737e-06,
+      "loss": 0.1951,
+      "step": 13025
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.8859649122807022e-06,
+      "loss": 0.8539,
+      "step": 13026
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.8854166666666666e-06,
+      "loss": 0.3958,
+      "step": 13027
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.884868421052632e-06,
+      "loss": 0.2485,
+      "step": 13028
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.8843201754385967e-06,
+      "loss": 0.3863,
+      "step": 13029
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.8837719298245615e-06,
+      "loss": 0.7034,
+      "step": 13030
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.8832236842105264e-06,
+      "loss": 0.8556,
+      "step": 13031
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.8826754385964916e-06,
+      "loss": 0.2403,
+      "step": 13032
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.8821271929824564e-06,
+      "loss": 0.1899,
+      "step": 13033
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.8815789473684213e-06,
+      "loss": 0.1567,
+      "step": 13034
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.881030701754386e-06,
+      "loss": 0.3926,
+      "step": 13035
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.880482456140351e-06,
+      "loss": 0.3705,
+      "step": 13036
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.879934210526316e-06,
+      "loss": 0.3323,
+      "step": 13037
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.8793859649122806e-06,
+      "loss": 0.7337,
+      "step": 13038
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.878837719298246e-06,
+      "loss": 0.2629,
+      "step": 13039
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.878289473684211e-06,
+      "loss": 0.2304,
+      "step": 13040
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.8777412280701755e-06,
+      "loss": 0.375,
+      "step": 13041
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.8771929824561407e-06,
+      "loss": 0.6382,
+      "step": 13042
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.876644736842106e-06,
+      "loss": 0.3126,
+      "step": 13043
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.8760964912280704e-06,
+      "loss": 0.1465,
+      "step": 13044
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.8755482456140356e-06,
+      "loss": 0.3995,
+      "step": 13045
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.875e-06,
+      "loss": 0.1157,
+      "step": 13046
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.8744517543859653e-06,
+      "loss": 0.712,
+      "step": 13047
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.87390350877193e-06,
+      "loss": 1.0884,
+      "step": 13048
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.873355263157895e-06,
+      "loss": 0.1623,
+      "step": 13049
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.8728070175438597e-06,
+      "loss": 0.5199,
+      "step": 13050
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.872258771929825e-06,
+      "loss": 0.3425,
+      "step": 13051
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.87171052631579e-06,
+      "loss": 1.5692,
+      "step": 13052
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.8711622807017546e-06,
+      "loss": 0.5353,
+      "step": 13053
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.8706140350877194e-06,
+      "loss": 0.9885,
+      "step": 13054
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.8700657894736843e-06,
+      "loss": 0.13,
+      "step": 13055
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.8695175438596495e-06,
+      "loss": 0.2612,
+      "step": 13056
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.868969298245614e-06,
+      "loss": 0.5302,
+      "step": 13057
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.868421052631579e-06,
+      "loss": 0.995,
+      "step": 13058
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.8678728070175444e-06,
+      "loss": 0.1485,
+      "step": 13059
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.867324561403509e-06,
+      "loss": 0.3171,
+      "step": 13060
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.866776315789474e-06,
+      "loss": 0.6849,
+      "step": 13061
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.8662280701754385e-06,
+      "loss": 0.7653,
+      "step": 13062
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.8656798245614037e-06,
+      "loss": 0.4803,
+      "step": 13063
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.865131578947369e-06,
+      "loss": 0.1134,
+      "step": 13064
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.8645833333333334e-06,
+      "loss": 0.2246,
+      "step": 13065
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.8640350877192986e-06,
+      "loss": 0.7994,
+      "step": 13066
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.8634868421052634e-06,
+      "loss": 0.2952,
+      "step": 13067
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.8629385964912283e-06,
+      "loss": 0.5034,
+      "step": 13068
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.8623903508771935e-06,
+      "loss": 0.1626,
+      "step": 13069
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.861842105263158e-06,
+      "loss": 0.5958,
+      "step": 13070
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.861293859649123e-06,
+      "loss": 0.5865,
+      "step": 13071
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.860745614035088e-06,
+      "loss": 0.173,
+      "step": 13072
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.860197368421053e-06,
+      "loss": 0.4643,
+      "step": 13073
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.8596491228070176e-06,
+      "loss": 0.7703,
+      "step": 13074
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.859100877192983e-06,
+      "loss": 0.2986,
+      "step": 13075
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.8585526315789473e-06,
+      "loss": 0.2682,
+      "step": 13076
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.8580043859649125e-06,
+      "loss": 0.5674,
+      "step": 13077
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.8574561403508774e-06,
+      "loss": 0.119,
+      "step": 13078
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.856907894736842e-06,
+      "loss": 0.5618,
+      "step": 13079
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.8563596491228074e-06,
+      "loss": 0.27,
+      "step": 13080
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.855811403508772e-06,
+      "loss": 0.415,
+      "step": 13081
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.855263157894737e-06,
+      "loss": 0.5363,
+      "step": 13082
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.8547149122807023e-06,
+      "loss": 0.8996,
+      "step": 13083
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.8541666666666667e-06,
+      "loss": 0.7508,
+      "step": 13084
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.853618421052632e-06,
+      "loss": 0.1807,
+      "step": 13085
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.853070175438597e-06,
+      "loss": 0.7706,
+      "step": 13086
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.8525219298245616e-06,
+      "loss": 0.7143,
+      "step": 13087
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.851973684210527e-06,
+      "loss": 0.9312,
+      "step": 13088
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.8514254385964913e-06,
+      "loss": 0.6252,
+      "step": 13089
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.8508771929824565e-06,
+      "loss": 0.564,
+      "step": 13090
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.8503289473684213e-06,
+      "loss": 0.4966,
+      "step": 13091
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.849780701754386e-06,
+      "loss": 0.9047,
+      "step": 13092
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.849232456140351e-06,
+      "loss": 0.1188,
+      "step": 13093
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.8486842105263162e-06,
+      "loss": 0.6546,
+      "step": 13094
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.8481359649122806e-06,
+      "loss": 1.2699,
+      "step": 13095
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.847587719298246e-06,
+      "loss": 1.122,
+      "step": 13096
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.8470394736842107e-06,
+      "loss": 0.248,
+      "step": 13097
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.8464912280701755e-06,
+      "loss": 0.3023,
+      "step": 13098
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.845942982456141e-06,
+      "loss": 0.6153,
+      "step": 13099
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.845394736842105e-06,
+      "loss": 0.244,
+      "step": 13100
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.8448464912280704e-06,
+      "loss": 0.3934,
+      "step": 13101
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.8442982456140357e-06,
+      "loss": 0.0993,
+      "step": 13102
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.84375e-06,
+      "loss": 0.4507,
+      "step": 13103
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.8432017543859653e-06,
+      "loss": 0.5353,
+      "step": 13104
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.8426535087719297e-06,
+      "loss": 1.2201,
+      "step": 13105
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.842105263157895e-06,
+      "loss": 0.529,
+      "step": 13106
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.8415570175438602e-06,
+      "loss": 0.3976,
+      "step": 13107
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.8410087719298246e-06,
+      "loss": 0.5426,
+      "step": 13108
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.84046052631579e-06,
+      "loss": 0.9907,
+      "step": 13109
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.8399122807017547e-06,
+      "loss": 0.2966,
+      "step": 13110
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.8393640350877195e-06,
+      "loss": 0.9899,
+      "step": 13111
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.8388157894736844e-06,
+      "loss": 0.2277,
+      "step": 13112
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.838267543859649e-06,
+      "loss": 0.2053,
+      "step": 13113
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.837719298245614e-06,
+      "loss": 0.3096,
+      "step": 13114
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.8371710526315793e-06,
+      "loss": 0.7931,
+      "step": 13115
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.836622807017544e-06,
+      "loss": 0.4242,
+      "step": 13116
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.836074561403509e-06,
+      "loss": 1.0972,
+      "step": 13117
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.835526315789474e-06,
+      "loss": 1.2574,
+      "step": 13118
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.8349780701754386e-06,
+      "loss": 0.8253,
+      "step": 13119
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.834429824561404e-06,
+      "loss": 0.6105,
+      "step": 13120
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.833881578947369e-06,
+      "loss": 0.3703,
+      "step": 13121
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.8333333333333335e-06,
+      "loss": 0.3703,
+      "step": 13122
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.8327850877192987e-06,
+      "loss": 0.1175,
+      "step": 13123
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.832236842105263e-06,
+      "loss": 0.3216,
+      "step": 13124
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.8316885964912283e-06,
+      "loss": 0.3026,
+      "step": 13125
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.8311403508771936e-06,
+      "loss": 0.4224,
+      "step": 13126
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.830592105263158e-06,
+      "loss": 1.2969,
+      "step": 13127
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.8300438596491232e-06,
+      "loss": 0.3285,
+      "step": 13128
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.829495614035088e-06,
+      "loss": 0.7994,
+      "step": 13129
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.828947368421053e-06,
+      "loss": 0.8895,
+      "step": 13130
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.8283991228070177e-06,
+      "loss": 0.7147,
+      "step": 13131
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.8278508771929825e-06,
+      "loss": 0.3025,
+      "step": 13132
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.8273026315789474e-06,
+      "loss": 0.2528,
+      "step": 13133
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.8267543859649126e-06,
+      "loss": 0.9198,
+      "step": 13134
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.8262061403508774e-06,
+      "loss": 0.585,
+      "step": 13135
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.8256578947368423e-06,
+      "loss": 0.521,
+      "step": 13136
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.8251096491228075e-06,
+      "loss": 0.5466,
+      "step": 13137
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.824561403508772e-06,
+      "loss": 1.9597,
+      "step": 13138
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.824013157894737e-06,
+      "loss": 0.2407,
+      "step": 13139
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.8234649122807016e-06,
+      "loss": 1.0686,
+      "step": 13140
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.822916666666667e-06,
+      "loss": 0.2727,
+      "step": 13141
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.822368421052632e-06,
+      "loss": 0.2815,
+      "step": 13142
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.8218201754385965e-06,
+      "loss": 0.7868,
+      "step": 13143
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.8212719298245617e-06,
+      "loss": 0.6689,
+      "step": 13144
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.820723684210527e-06,
+      "loss": 0.9231,
+      "step": 13145
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.8201754385964914e-06,
+      "loss": 0.9137,
+      "step": 13146
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.8196271929824566e-06,
+      "loss": 0.6256,
+      "step": 13147
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.819078947368421e-06,
+      "loss": 0.5249,
+      "step": 13148
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.8185307017543863e-06,
+      "loss": 0.4966,
+      "step": 13149
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.817982456140351e-06,
+      "loss": 0.5642,
+      "step": 13150
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.817434210526316e-06,
+      "loss": 0.6659,
+      "step": 13151
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.816885964912281e-06,
+      "loss": 0.1004,
+      "step": 13152
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.816337719298246e-06,
+      "loss": 0.7532,
+      "step": 13153
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.815789473684211e-06,
+      "loss": 0.7933,
+      "step": 13154
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.8152412280701756e-06,
+      "loss": 0.874,
+      "step": 13155
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.8146929824561405e-06,
+      "loss": 0.5705,
+      "step": 13156
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.8141447368421053e-06,
+      "loss": 0.5502,
+      "step": 13157
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.8135964912280705e-06,
+      "loss": 0.1484,
+      "step": 13158
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.813048245614035e-06,
+      "loss": 0.809,
+      "step": 13159
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.8125e-06,
+      "loss": 0.2513,
+      "step": 13160
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.8119517543859654e-06,
+      "loss": 0.1878,
+      "step": 13161
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.81140350877193e-06,
+      "loss": 0.313,
+      "step": 13162
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.810855263157895e-06,
+      "loss": 0.2376,
+      "step": 13163
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.8103070175438603e-06,
+      "loss": 0.4025,
+      "step": 13164
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.8097587719298247e-06,
+      "loss": 0.386,
+      "step": 13165
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.80921052631579e-06,
+      "loss": 0.9402,
+      "step": 13166
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.8086622807017544e-06,
+      "loss": 0.505,
+      "step": 13167
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.8081140350877196e-06,
+      "loss": 0.1818,
+      "step": 13168
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.8075657894736844e-06,
+      "loss": 0.6634,
+      "step": 13169
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.8070175438596493e-06,
+      "loss": 0.4828,
+      "step": 13170
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.8064692982456145e-06,
+      "loss": 0.3778,
+      "step": 13171
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.8059210526315793e-06,
+      "loss": 0.6488,
+      "step": 13172
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.805372807017544e-06,
+      "loss": 0.319,
+      "step": 13173
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.804824561403509e-06,
+      "loss": 0.7049,
+      "step": 13174
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.804276315789474e-06,
+      "loss": 0.9632,
+      "step": 13175
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.8037280701754386e-06,
+      "loss": 0.6364,
+      "step": 13176
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.803179824561404e-06,
+      "loss": 0.3134,
+      "step": 13177
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.8026315789473683e-06,
+      "loss": 1.0034,
+      "step": 13178
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.8020833333333335e-06,
+      "loss": 0.3418,
+      "step": 13179
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.8015350877192988e-06,
+      "loss": 0.6556,
+      "step": 13180
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.800986842105263e-06,
+      "loss": 0.7124,
+      "step": 13181
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.8004385964912284e-06,
+      "loss": 0.1297,
+      "step": 13182
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.799890350877193e-06,
+      "loss": 0.1576,
+      "step": 13183
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.799342105263158e-06,
+      "loss": 0.1373,
+      "step": 13184
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.7987938596491233e-06,
+      "loss": 0.2181,
+      "step": 13185
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.7982456140350877e-06,
+      "loss": 0.2355,
+      "step": 13186
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.797697368421053e-06,
+      "loss": 0.4076,
+      "step": 13187
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.797149122807018e-06,
+      "loss": 0.9302,
+      "step": 13188
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.7966008771929826e-06,
+      "loss": 0.7358,
+      "step": 13189
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.796052631578948e-06,
+      "loss": 1.1921,
+      "step": 13190
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.7955043859649123e-06,
+      "loss": 0.3598,
+      "step": 13191
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.7949561403508775e-06,
+      "loss": 0.2853,
+      "step": 13192
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.7944078947368424e-06,
+      "loss": 0.2663,
+      "step": 13193
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.793859649122807e-06,
+      "loss": 0.6134,
+      "step": 13194
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.793311403508772e-06,
+      "loss": 0.3455,
+      "step": 13195
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.7927631578947373e-06,
+      "loss": 0.3549,
+      "step": 13196
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.7922149122807017e-06,
+      "loss": 0.7865,
+      "step": 13197
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.791666666666667e-06,
+      "loss": 1.0581,
+      "step": 13198
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.7911184210526317e-06,
+      "loss": 0.6105,
+      "step": 13199
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.7905701754385966e-06,
+      "loss": 0.3757,
+      "step": 13200
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.790021929824562e-06,
+      "loss": 0.8515,
+      "step": 13201
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.789473684210526e-06,
+      "loss": 0.21,
+      "step": 13202
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.7889254385964914e-06,
+      "loss": 0.6998,
+      "step": 13203
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.7883771929824567e-06,
+      "loss": 0.7779,
+      "step": 13204
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.787828947368421e-06,
+      "loss": 0.3829,
+      "step": 13205
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.7872807017543863e-06,
+      "loss": 0.7004,
+      "step": 13206
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.786732456140351e-06,
+      "loss": 0.8138,
+      "step": 13207
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.786184210526316e-06,
+      "loss": 0.7981,
+      "step": 13208
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.7856359649122812e-06,
+      "loss": 0.4599,
+      "step": 13209
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.7850877192982456e-06,
+      "loss": 0.7846,
+      "step": 13210
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.784539473684211e-06,
+      "loss": 0.1817,
+      "step": 13211
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.7839912280701757e-06,
+      "loss": 0.5293,
+      "step": 13212
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.7834429824561405e-06,
+      "loss": 0.0507,
+      "step": 13213
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.7828947368421054e-06,
+      "loss": 0.1034,
+      "step": 13214
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.7823464912280706e-06,
+      "loss": 0.0491,
+      "step": 13215
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.781798245614035e-06,
+      "loss": 0.4418,
+      "step": 13216
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.7812500000000003e-06,
+      "loss": 0.6622,
+      "step": 13217
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.780701754385965e-06,
+      "loss": 0.3172,
+      "step": 13218
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.78015350877193e-06,
+      "loss": 0.2805,
+      "step": 13219
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.779605263157895e-06,
+      "loss": 0.67,
+      "step": 13220
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.7790570175438596e-06,
+      "loss": 1.0462,
+      "step": 13221
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.778508771929825e-06,
+      "loss": 0.7702,
+      "step": 13222
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.77796052631579e-06,
+      "loss": 0.5874,
+      "step": 13223
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.7774122807017545e-06,
+      "loss": 1.0477,
+      "step": 13224
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.7768640350877197e-06,
+      "loss": 0.7339,
+      "step": 13225
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.776315789473684e-06,
+      "loss": 0.7762,
+      "step": 13226
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.7757675438596494e-06,
+      "loss": 0.5937,
+      "step": 13227
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.7752192982456146e-06,
+      "loss": 0.2257,
+      "step": 13228
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.774671052631579e-06,
+      "loss": 1.1469,
+      "step": 13229
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.7741228070175443e-06,
+      "loss": 0.705,
+      "step": 13230
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.773574561403509e-06,
+      "loss": 0.8724,
+      "step": 13231
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.773026315789474e-06,
+      "loss": 1.3155,
+      "step": 13232
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.7724780701754387e-06,
+      "loss": 0.1302,
+      "step": 13233
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.7719298245614036e-06,
+      "loss": 0.3835,
+      "step": 13234
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.7713815789473684e-06,
+      "loss": 0.348,
+      "step": 13235
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.7708333333333336e-06,
+      "loss": 0.6745,
+      "step": 13236
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.7702850877192985e-06,
+      "loss": 0.9402,
+      "step": 13237
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.7697368421052633e-06,
+      "loss": 0.4478,
+      "step": 13238
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.7691885964912285e-06,
+      "loss": 0.3261,
+      "step": 13239
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.768640350877193e-06,
+      "loss": 0.5134,
+      "step": 13240
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.768092105263158e-06,
+      "loss": 0.5411,
+      "step": 13241
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.7675438596491234e-06,
+      "loss": 0.1035,
+      "step": 13242
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.766995614035088e-06,
+      "loss": 0.5823,
+      "step": 13243
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.766447368421053e-06,
+      "loss": 1.0685,
+      "step": 13244
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.7658991228070175e-06,
+      "loss": 0.1323,
+      "step": 13245
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.7653508771929827e-06,
+      "loss": 0.464,
+      "step": 13246
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.764802631578948e-06,
+      "loss": 1.185,
+      "step": 13247
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.7642543859649124e-06,
+      "loss": 0.127,
+      "step": 13248
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.7637061403508776e-06,
+      "loss": 0.5392,
+      "step": 13249
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.7631578947368424e-06,
+      "loss": 1.1258,
+      "step": 13250
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.7626096491228073e-06,
+      "loss": 0.5693,
+      "step": 13251
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.762061403508772e-06,
+      "loss": 0.6037,
+      "step": 13252
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.761513157894737e-06,
+      "loss": 0.7151,
+      "step": 13253
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.760964912280702e-06,
+      "loss": 0.841,
+      "step": 13254
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.760416666666667e-06,
+      "loss": 0.5402,
+      "step": 13255
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.759868421052632e-06,
+      "loss": 0.9377,
+      "step": 13256
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.7593201754385966e-06,
+      "loss": 0.3075,
+      "step": 13257
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.758771929824562e-06,
+      "loss": 0.1839,
+      "step": 13258
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.7582236842105263e-06,
+      "loss": 0.7324,
+      "step": 13259
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.7576754385964915e-06,
+      "loss": 0.724,
+      "step": 13260
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.757127192982456e-06,
+      "loss": 0.3822,
+      "step": 13261
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.756578947368421e-06,
+      "loss": 0.2382,
+      "step": 13262
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.7560307017543864e-06,
+      "loss": 0.2972,
+      "step": 13263
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.755482456140351e-06,
+      "loss": 0.124,
+      "step": 13264
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.754934210526316e-06,
+      "loss": 0.831,
+      "step": 13265
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.7543859649122813e-06,
+      "loss": 1.0257,
+      "step": 13266
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.7538377192982457e-06,
+      "loss": 0.0598,
+      "step": 13267
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.753289473684211e-06,
+      "loss": 0.9744,
+      "step": 13268
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.7527412280701754e-06,
+      "loss": 0.547,
+      "step": 13269
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.7521929824561406e-06,
+      "loss": 0.3313,
+      "step": 13270
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.7516447368421055e-06,
+      "loss": 0.7521,
+      "step": 13271
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.7510964912280703e-06,
+      "loss": 0.5795,
+      "step": 13272
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.7505482456140355e-06,
+      "loss": 0.4598,
+      "step": 13273
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.7500000000000004e-06,
+      "loss": 0.6359,
+      "step": 13274
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.749451754385965e-06,
+      "loss": 1.0866,
+      "step": 13275
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.74890350877193e-06,
+      "loss": 0.9759,
+      "step": 13276
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.748355263157895e-06,
+      "loss": 0.5667,
+      "step": 13277
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.7478070175438597e-06,
+      "loss": 0.5917,
+      "step": 13278
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.747258771929825e-06,
+      "loss": 0.8647,
+      "step": 13279
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.7467105263157893e-06,
+      "loss": 0.6192,
+      "step": 13280
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.7461622807017545e-06,
+      "loss": 0.3193,
+      "step": 13281
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.74561403508772e-06,
+      "loss": 0.6997,
+      "step": 13282
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.745065789473684e-06,
+      "loss": 0.848,
+      "step": 13283
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.7445175438596494e-06,
+      "loss": 0.1448,
+      "step": 13284
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.7439692982456147e-06,
+      "loss": 0.1631,
+      "step": 13285
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.743421052631579e-06,
+      "loss": 0.2119,
+      "step": 13286
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.7428728070175443e-06,
+      "loss": 0.576,
+      "step": 13287
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.7423245614035087e-06,
+      "loss": 0.6061,
+      "step": 13288
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.741776315789474e-06,
+      "loss": 0.9966,
+      "step": 13289
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.741228070175439e-06,
+      "loss": 0.1799,
+      "step": 13290
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.7406798245614036e-06,
+      "loss": 0.1654,
+      "step": 13291
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.740131578947369e-06,
+      "loss": 0.3173,
+      "step": 13292
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.7395833333333337e-06,
+      "loss": 0.9022,
+      "step": 13293
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.7390350877192985e-06,
+      "loss": 0.2181,
+      "step": 13294
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.7384868421052634e-06,
+      "loss": 0.3855,
+      "step": 13295
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.737938596491228e-06,
+      "loss": 0.8395,
+      "step": 13296
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.737390350877193e-06,
+      "loss": 0.63,
+      "step": 13297
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.7368421052631583e-06,
+      "loss": 0.4599,
+      "step": 13298
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.7362938596491227e-06,
+      "loss": 0.2063,
+      "step": 13299
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.735745614035088e-06,
+      "loss": 0.5889,
+      "step": 13300
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.735197368421053e-06,
+      "loss": 0.4049,
+      "step": 13301
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.7346491228070176e-06,
+      "loss": 0.6784,
+      "step": 13302
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.734100877192983e-06,
+      "loss": 0.14,
+      "step": 13303
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.733552631578947e-06,
+      "loss": 0.2955,
+      "step": 13304
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.7330043859649125e-06,
+      "loss": 0.487,
+      "step": 13305
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.7324561403508777e-06,
+      "loss": 0.2959,
+      "step": 13306
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.731907894736842e-06,
+      "loss": 0.5362,
+      "step": 13307
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.7313596491228074e-06,
+      "loss": 0.2965,
+      "step": 13308
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.730811403508772e-06,
+      "loss": 0.2233,
+      "step": 13309
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.730263157894737e-06,
+      "loss": 0.4006,
+      "step": 13310
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.7297149122807023e-06,
+      "loss": 0.067,
+      "step": 13311
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.7291666666666667e-06,
+      "loss": 0.1692,
+      "step": 13312
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.728618421052632e-06,
+      "loss": 0.1354,
+      "step": 13313
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.7280701754385967e-06,
+      "loss": 0.532,
+      "step": 13314
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.7275219298245616e-06,
+      "loss": 0.9738,
+      "step": 13315
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.7269736842105264e-06,
+      "loss": 0.3703,
+      "step": 13316
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.7264254385964916e-06,
+      "loss": 1.1434,
+      "step": 13317
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.725877192982456e-06,
+      "loss": 0.6254,
+      "step": 13318
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.7253289473684213e-06,
+      "loss": 1.4856,
+      "step": 13319
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.724780701754386e-06,
+      "loss": 0.5443,
+      "step": 13320
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.724232456140351e-06,
+      "loss": 0.6579,
+      "step": 13321
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.723684210526316e-06,
+      "loss": 0.5564,
+      "step": 13322
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.7231359649122806e-06,
+      "loss": 1.718,
+      "step": 13323
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.722587719298246e-06,
+      "loss": 0.9133,
+      "step": 13324
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.722039473684211e-06,
+      "loss": 0.6065,
+      "step": 13325
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.7214912280701755e-06,
+      "loss": 0.4331,
+      "step": 13326
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.7209429824561407e-06,
+      "loss": 0.5519,
+      "step": 13327
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.720394736842106e-06,
+      "loss": 0.1418,
+      "step": 13328
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.7198464912280704e-06,
+      "loss": 0.8151,
+      "step": 13329
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.7192982456140356e-06,
+      "loss": 0.3871,
+      "step": 13330
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.71875e-06,
+      "loss": 0.4395,
+      "step": 13331
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.7182017543859653e-06,
+      "loss": 1.2623,
+      "step": 13332
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.71765350877193e-06,
+      "loss": 0.4934,
+      "step": 13333
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.717105263157895e-06,
+      "loss": 0.4459,
+      "step": 13334
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.7165570175438597e-06,
+      "loss": 0.5748,
+      "step": 13335
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.716008771929825e-06,
+      "loss": 0.2216,
+      "step": 13336
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.71546052631579e-06,
+      "loss": 0.153,
+      "step": 13337
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.7149122807017546e-06,
+      "loss": 1.0461,
+      "step": 13338
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.7143640350877195e-06,
+      "loss": 0.3434,
+      "step": 13339
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.7138157894736843e-06,
+      "loss": 0.1378,
+      "step": 13340
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.7132675438596495e-06,
+      "loss": 0.8942,
+      "step": 13341
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.712719298245614e-06,
+      "loss": 0.5305,
+      "step": 13342
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.712171052631579e-06,
+      "loss": 0.1385,
+      "step": 13343
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.7116228070175444e-06,
+      "loss": 0.1288,
+      "step": 13344
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.711074561403509e-06,
+      "loss": 0.4988,
+      "step": 13345
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.710526315789474e-06,
+      "loss": 0.4765,
+      "step": 13346
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.7099780701754385e-06,
+      "loss": 0.4129,
+      "step": 13347
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.7094298245614037e-06,
+      "loss": 0.1291,
+      "step": 13348
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.708881578947369e-06,
+      "loss": 0.4424,
+      "step": 13349
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.7083333333333334e-06,
+      "loss": 0.1174,
+      "step": 13350
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.7077850877192986e-06,
+      "loss": 0.3541,
+      "step": 13351
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.7072368421052635e-06,
+      "loss": 0.6594,
+      "step": 13352
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.7066885964912283e-06,
+      "loss": 0.6984,
+      "step": 13353
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.706140350877193e-06,
+      "loss": 0.5182,
+      "step": 13354
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.705592105263158e-06,
+      "loss": 0.2701,
+      "step": 13355
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.705043859649123e-06,
+      "loss": 0.6809,
+      "step": 13356
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.704495614035088e-06,
+      "loss": 0.8411,
+      "step": 13357
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.703947368421053e-06,
+      "loss": 0.4102,
+      "step": 13358
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.7033991228070176e-06,
+      "loss": 0.7087,
+      "step": 13359
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.702850877192983e-06,
+      "loss": 0.4937,
+      "step": 13360
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.7023026315789473e-06,
+      "loss": 0.9762,
+      "step": 13361
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.7017543859649125e-06,
+      "loss": 0.581,
+      "step": 13362
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.701206140350877e-06,
+      "loss": 1.0646,
+      "step": 13363
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.700657894736842e-06,
+      "loss": 0.4925,
+      "step": 13364
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.7001096491228074e-06,
+      "loss": 0.5995,
+      "step": 13365
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.699561403508772e-06,
+      "loss": 0.2974,
+      "step": 13366
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.699013157894737e-06,
+      "loss": 0.2215,
+      "step": 13367
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.6984649122807023e-06,
+      "loss": 0.2649,
+      "step": 13368
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.6979166666666667e-06,
+      "loss": 0.203,
+      "step": 13369
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.697368421052632e-06,
+      "loss": 0.5717,
+      "step": 13370
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.696820175438597e-06,
+      "loss": 0.6387,
+      "step": 13371
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.6962719298245616e-06,
+      "loss": 0.8916,
+      "step": 13372
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.6957236842105265e-06,
+      "loss": 0.8629,
+      "step": 13373
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.6951754385964913e-06,
+      "loss": 0.6491,
+      "step": 13374
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.6946271929824565e-06,
+      "loss": 0.8418,
+      "step": 13375
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.6940789473684214e-06,
+      "loss": 0.3316,
+      "step": 13376
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.693530701754386e-06,
+      "loss": 0.7332,
+      "step": 13377
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.692982456140351e-06,
+      "loss": 0.3575,
+      "step": 13378
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.6924342105263163e-06,
+      "loss": 0.2912,
+      "step": 13379
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.6918859649122807e-06,
+      "loss": 0.2645,
+      "step": 13380
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.691337719298246e-06,
+      "loss": 0.7101,
+      "step": 13381
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.6907894736842103e-06,
+      "loss": 0.3354,
+      "step": 13382
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.6902412280701756e-06,
+      "loss": 0.5431,
+      "step": 13383
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.689692982456141e-06,
+      "loss": 0.2314,
+      "step": 13384
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.689144736842105e-06,
+      "loss": 1.1462,
+      "step": 13385
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.6885964912280705e-06,
+      "loss": 0.3453,
+      "step": 13386
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.6880482456140357e-06,
+      "loss": 1.6952,
+      "step": 13387
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.6875e-06,
+      "loss": 0.3541,
+      "step": 13388
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.6869517543859654e-06,
+      "loss": 0.9618,
+      "step": 13389
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.6864035087719298e-06,
+      "loss": 0.9187,
+      "step": 13390
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.685855263157895e-06,
+      "loss": 0.1607,
+      "step": 13391
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.68530701754386e-06,
+      "loss": 0.145,
+      "step": 13392
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.6847587719298247e-06,
+      "loss": 1.0828,
+      "step": 13393
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.68421052631579e-06,
+      "loss": 0.1109,
+      "step": 13394
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.6836622807017547e-06,
+      "loss": 0.4376,
+      "step": 13395
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.6831140350877195e-06,
+      "loss": 0.2473,
+      "step": 13396
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.6825657894736844e-06,
+      "loss": 0.3886,
+      "step": 13397
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.682017543859649e-06,
+      "loss": 0.3529,
+      "step": 13398
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.681469298245614e-06,
+      "loss": 0.1057,
+      "step": 13399
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.6809210526315793e-06,
+      "loss": 0.5042,
+      "step": 13400
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.6803728070175437e-06,
+      "loss": 0.5439,
+      "step": 13401
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.679824561403509e-06,
+      "loss": 0.6877,
+      "step": 13402
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.679276315789474e-06,
+      "loss": 0.3575,
+      "step": 13403
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.6787280701754386e-06,
+      "loss": 0.8839,
+      "step": 13404
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.678179824561404e-06,
+      "loss": 0.7362,
+      "step": 13405
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.677631578947369e-06,
+      "loss": 0.5811,
+      "step": 13406
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.6770833333333335e-06,
+      "loss": 0.5211,
+      "step": 13407
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.6765350877192987e-06,
+      "loss": 1.1149,
+      "step": 13408
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.675986842105263e-06,
+      "loss": 0.4007,
+      "step": 13409
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.6754385964912284e-06,
+      "loss": 0.7072,
+      "step": 13410
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.674890350877193e-06,
+      "loss": 0.1504,
+      "step": 13411
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.674342105263158e-06,
+      "loss": 0.7836,
+      "step": 13412
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.6737938596491233e-06,
+      "loss": 0.6175,
+      "step": 13413
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.673245614035088e-06,
+      "loss": 0.5636,
+      "step": 13414
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.672697368421053e-06,
+      "loss": 0.7006,
+      "step": 13415
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.6721491228070177e-06,
+      "loss": 1.2585,
+      "step": 13416
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.6716008771929826e-06,
+      "loss": 0.3751,
+      "step": 13417
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.6710526315789474e-06,
+      "loss": 0.7613,
+      "step": 13418
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.6705043859649126e-06,
+      "loss": 0.7592,
+      "step": 13419
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.6699561403508775e-06,
+      "loss": 0.5408,
+      "step": 13420
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.6694078947368423e-06,
+      "loss": 0.4494,
+      "step": 13421
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.6688596491228075e-06,
+      "loss": 0.7869,
+      "step": 13422
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.668311403508772e-06,
+      "loss": 0.2451,
+      "step": 13423
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.667763157894737e-06,
+      "loss": 0.5504,
+      "step": 13424
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.6672149122807016e-06,
+      "loss": 0.8284,
+      "step": 13425
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.666666666666667e-06,
+      "loss": 0.6757,
+      "step": 13426
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.666118421052632e-06,
+      "loss": 0.1212,
+      "step": 13427
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.6655701754385965e-06,
+      "loss": 0.3468,
+      "step": 13428
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.6650219298245617e-06,
+      "loss": 0.1386,
+      "step": 13429
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.664473684210527e-06,
+      "loss": 0.9727,
+      "step": 13430
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.6639254385964914e-06,
+      "loss": 0.1125,
+      "step": 13431
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.6633771929824566e-06,
+      "loss": 1.1092,
+      "step": 13432
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.662828947368421e-06,
+      "loss": 0.8068,
+      "step": 13433
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.6622807017543863e-06,
+      "loss": 0.9944,
+      "step": 13434
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.661732456140351e-06,
+      "loss": 1.0383,
+      "step": 13435
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.661184210526316e-06,
+      "loss": 0.3771,
+      "step": 13436
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.6606359649122807e-06,
+      "loss": 0.205,
+      "step": 13437
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.660087719298246e-06,
+      "loss": 0.6975,
+      "step": 13438
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.659539473684211e-06,
+      "loss": 0.7462,
+      "step": 13439
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.6589912280701756e-06,
+      "loss": 0.2999,
+      "step": 13440
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.6584429824561405e-06,
+      "loss": 0.7757,
+      "step": 13441
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.6578947368421053e-06,
+      "loss": 1.0342,
+      "step": 13442
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.6573464912280705e-06,
+      "loss": 0.1203,
+      "step": 13443
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.656798245614035e-06,
+      "loss": 0.5957,
+      "step": 13444
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.65625e-06,
+      "loss": 1.8214,
+      "step": 13445
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.6557017543859654e-06,
+      "loss": 0.4613,
+      "step": 13446
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.65515350877193e-06,
+      "loss": 0.2361,
+      "step": 13447
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.654605263157895e-06,
+      "loss": 1.3606,
+      "step": 13448
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.6540570175438603e-06,
+      "loss": 0.8657,
+      "step": 13449
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.6535087719298247e-06,
+      "loss": 0.9941,
+      "step": 13450
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.65296052631579e-06,
+      "loss": 0.6065,
+      "step": 13451
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.6524122807017544e-06,
+      "loss": 0.4028,
+      "step": 13452
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.6518640350877196e-06,
+      "loss": 0.672,
+      "step": 13453
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.6513157894736845e-06,
+      "loss": 1.0053,
+      "step": 13454
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.6507675438596493e-06,
+      "loss": 0.6371,
+      "step": 13455
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.650219298245614e-06,
+      "loss": 1.2069,
+      "step": 13456
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.6496710526315794e-06,
+      "loss": 0.4688,
+      "step": 13457
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.649122807017544e-06,
+      "loss": 0.8453,
+      "step": 13458
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.648574561403509e-06,
+      "loss": 0.7343,
+      "step": 13459
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.648026315789474e-06,
+      "loss": 0.1938,
+      "step": 13460
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.6474780701754387e-06,
+      "loss": 0.6059,
+      "step": 13461
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.646929824561404e-06,
+      "loss": 0.7777,
+      "step": 13462
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.6463815789473683e-06,
+      "loss": 0.1166,
+      "step": 13463
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.6458333333333336e-06,
+      "loss": 0.3407,
+      "step": 13464
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.645285087719299e-06,
+      "loss": 0.9726,
+      "step": 13465
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.644736842105263e-06,
+      "loss": 1.3969,
+      "step": 13466
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.6441885964912285e-06,
+      "loss": 0.3551,
+      "step": 13467
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.643640350877193e-06,
+      "loss": 0.9407,
+      "step": 13468
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.643092105263158e-06,
+      "loss": 0.9456,
+      "step": 13469
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.6425438596491233e-06,
+      "loss": 0.1731,
+      "step": 13470
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.6419956140350877e-06,
+      "loss": 0.6067,
+      "step": 13471
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.641447368421053e-06,
+      "loss": 0.5752,
+      "step": 13472
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.640899122807018e-06,
+      "loss": 0.4429,
+      "step": 13473
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.6403508771929826e-06,
+      "loss": 0.8073,
+      "step": 13474
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.6398026315789475e-06,
+      "loss": 0.6309,
+      "step": 13475
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.6392543859649123e-06,
+      "loss": 0.4065,
+      "step": 13476
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.6387061403508775e-06,
+      "loss": 0.0911,
+      "step": 13477
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.6381578947368424e-06,
+      "loss": 0.8589,
+      "step": 13478
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.637609649122807e-06,
+      "loss": 0.3531,
+      "step": 13479
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.637061403508772e-06,
+      "loss": 0.2272,
+      "step": 13480
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.6365131578947373e-06,
+      "loss": 0.2639,
+      "step": 13481
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.6359649122807017e-06,
+      "loss": 0.4671,
+      "step": 13482
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.635416666666667e-06,
+      "loss": 0.4237,
+      "step": 13483
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.6348684210526313e-06,
+      "loss": 0.0833,
+      "step": 13484
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.6343201754385966e-06,
+      "loss": 0.2782,
+      "step": 13485
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.633771929824562e-06,
+      "loss": 0.4103,
+      "step": 13486
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.6332236842105262e-06,
+      "loss": 0.5785,
+      "step": 13487
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.6326754385964915e-06,
+      "loss": 0.9314,
+      "step": 13488
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.6321271929824567e-06,
+      "loss": 0.854,
+      "step": 13489
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.631578947368421e-06,
+      "loss": 0.4178,
+      "step": 13490
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.6310307017543864e-06,
+      "loss": 0.5552,
+      "step": 13491
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.630482456140351e-06,
+      "loss": 0.3755,
+      "step": 13492
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.629934210526316e-06,
+      "loss": 0.5865,
+      "step": 13493
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.629385964912281e-06,
+      "loss": 0.7455,
+      "step": 13494
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.6288377192982457e-06,
+      "loss": 0.0237,
+      "step": 13495
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.628289473684211e-06,
+      "loss": 0.4745,
+      "step": 13496
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.6277412280701757e-06,
+      "loss": 1.4402,
+      "step": 13497
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.6271929824561406e-06,
+      "loss": 0.4721,
+      "step": 13498
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.6266447368421054e-06,
+      "loss": 1.0829,
+      "step": 13499
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.6260964912280706e-06,
+      "loss": 0.8364,
+      "step": 13500
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.625548245614035e-06,
+      "loss": 0.0971,
+      "step": 13501
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.6250000000000003e-06,
+      "loss": 0.7487,
+      "step": 13502
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.6244517543859647e-06,
+      "loss": 1.0599,
+      "step": 13503
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.62390350877193e-06,
+      "loss": 0.5638,
+      "step": 13504
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.623355263157895e-06,
+      "loss": 0.4552,
+      "step": 13505
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.6228070175438596e-06,
+      "loss": 0.7014,
+      "step": 13506
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.622258771929825e-06,
+      "loss": 1.0351,
+      "step": 13507
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.62171052631579e-06,
+      "loss": 0.1625,
+      "step": 13508
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.6211622807017545e-06,
+      "loss": 0.7694,
+      "step": 13509
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.6206140350877197e-06,
+      "loss": 0.1203,
+      "step": 13510
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.620065789473684e-06,
+      "loss": 0.2805,
+      "step": 13511
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.6195175438596494e-06,
+      "loss": 0.6981,
+      "step": 13512
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.6189692982456146e-06,
+      "loss": 0.2952,
+      "step": 13513
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.618421052631579e-06,
+      "loss": 0.2459,
+      "step": 13514
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.6178728070175443e-06,
+      "loss": 0.2269,
+      "step": 13515
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.617324561403509e-06,
+      "loss": 0.0949,
+      "step": 13516
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.616776315789474e-06,
+      "loss": 0.5824,
+      "step": 13517
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.6162280701754387e-06,
+      "loss": 0.0691,
+      "step": 13518
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.6156798245614036e-06,
+      "loss": 0.1958,
+      "step": 13519
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.6151315789473684e-06,
+      "loss": 0.6365,
+      "step": 13520
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.6145833333333336e-06,
+      "loss": 0.6672,
+      "step": 13521
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.6140350877192985e-06,
+      "loss": 0.7102,
+      "step": 13522
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.6134868421052633e-06,
+      "loss": 0.0415,
+      "step": 13523
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.6129385964912285e-06,
+      "loss": 0.593,
+      "step": 13524
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.612390350877193e-06,
+      "loss": 0.2122,
+      "step": 13525
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.611842105263158e-06,
+      "loss": 0.7382,
+      "step": 13526
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.6112938596491234e-06,
+      "loss": 0.4359,
+      "step": 13527
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.610745614035088e-06,
+      "loss": 0.349,
+      "step": 13528
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.610197368421053e-06,
+      "loss": 0.8843,
+      "step": 13529
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.6096491228070175e-06,
+      "loss": 0.1955,
+      "step": 13530
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.6091008771929827e-06,
+      "loss": 0.5833,
+      "step": 13531
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.608552631578948e-06,
+      "loss": 1.2349,
+      "step": 13532
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.6080043859649124e-06,
+      "loss": 0.495,
+      "step": 13533
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.6074561403508776e-06,
+      "loss": 1.0356,
+      "step": 13534
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.6069078947368425e-06,
+      "loss": 0.7423,
+      "step": 13535
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.6063596491228073e-06,
+      "loss": 0.2832,
+      "step": 13536
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.605811403508772e-06,
+      "loss": 0.8985,
+      "step": 13537
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.605263157894737e-06,
+      "loss": 0.1694,
+      "step": 13538
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.6047149122807018e-06,
+      "loss": 0.5444,
+      "step": 13539
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.604166666666667e-06,
+      "loss": 0.316,
+      "step": 13540
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.603618421052632e-06,
+      "loss": 0.6309,
+      "step": 13541
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.6030701754385967e-06,
+      "loss": 0.5681,
+      "step": 13542
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.602521929824562e-06,
+      "loss": 0.3423,
+      "step": 13543
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.6019736842105263e-06,
+      "loss": 0.4167,
+      "step": 13544
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.6014254385964916e-06,
+      "loss": 0.1246,
+      "step": 13545
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.600877192982456e-06,
+      "loss": 0.4497,
+      "step": 13546
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.600328947368421e-06,
+      "loss": 0.6082,
+      "step": 13547
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.5997807017543864e-06,
+      "loss": 0.331,
+      "step": 13548
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.599232456140351e-06,
+      "loss": 0.0881,
+      "step": 13549
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.598684210526316e-06,
+      "loss": 0.1381,
+      "step": 13550
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.5981359649122813e-06,
+      "loss": 0.4756,
+      "step": 13551
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.5975877192982457e-06,
+      "loss": 0.2614,
+      "step": 13552
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.597039473684211e-06,
+      "loss": 0.7262,
+      "step": 13553
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.5964912280701754e-06,
+      "loss": 0.5696,
+      "step": 13554
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.5959429824561406e-06,
+      "loss": 1.1508,
+      "step": 13555
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.5953947368421055e-06,
+      "loss": 0.0922,
+      "step": 13556
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.5948464912280703e-06,
+      "loss": 0.522,
+      "step": 13557
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.594298245614035e-06,
+      "loss": 0.423,
+      "step": 13558
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.5937500000000004e-06,
+      "loss": 0.4188,
+      "step": 13559
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.593201754385965e-06,
+      "loss": 0.1154,
+      "step": 13560
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.59265350877193e-06,
+      "loss": 0.4846,
+      "step": 13561
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.592105263157895e-06,
+      "loss": 0.3959,
+      "step": 13562
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.5915570175438597e-06,
+      "loss": 0.5189,
+      "step": 13563
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.591008771929825e-06,
+      "loss": 0.8235,
+      "step": 13564
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.5904605263157893e-06,
+      "loss": 0.5757,
+      "step": 13565
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.5899122807017546e-06,
+      "loss": 0.0525,
+      "step": 13566
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.58936403508772e-06,
+      "loss": 0.1737,
+      "step": 13567
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.588815789473684e-06,
+      "loss": 0.6847,
+      "step": 13568
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.5882675438596495e-06,
+      "loss": 0.0797,
+      "step": 13569
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.5877192982456147e-06,
+      "loss": 0.9182,
+      "step": 13570
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.587171052631579e-06,
+      "loss": 1.1728,
+      "step": 13571
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.5866228070175444e-06,
+      "loss": 0.5562,
+      "step": 13572
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.5860745614035088e-06,
+      "loss": 0.6101,
+      "step": 13573
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.585526315789474e-06,
+      "loss": 0.1712,
+      "step": 13574
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.584978070175439e-06,
+      "loss": 0.5566,
+      "step": 13575
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.5844298245614037e-06,
+      "loss": 0.6939,
+      "step": 13576
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.5838815789473685e-06,
+      "loss": 0.6265,
+      "step": 13577
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.5833333333333337e-06,
+      "loss": 0.4418,
+      "step": 13578
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.5827850877192986e-06,
+      "loss": 0.601,
+      "step": 13579
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.5822368421052634e-06,
+      "loss": 0.1923,
+      "step": 13580
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.581688596491228e-06,
+      "loss": 0.1178,
+      "step": 13581
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.581140350877193e-06,
+      "loss": 0.4237,
+      "step": 13582
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.5805921052631583e-06,
+      "loss": 0.2937,
+      "step": 13583
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.5800438596491227e-06,
+      "loss": 0.2509,
+      "step": 13584
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.579495614035088e-06,
+      "loss": 0.3985,
+      "step": 13585
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.578947368421053e-06,
+      "loss": 0.7172,
+      "step": 13586
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.5783991228070176e-06,
+      "loss": 0.9902,
+      "step": 13587
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.577850877192983e-06,
+      "loss": 0.5565,
+      "step": 13588
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.5773026315789472e-06,
+      "loss": 0.3052,
+      "step": 13589
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.5767543859649125e-06,
+      "loss": 1.2329,
+      "step": 13590
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.5762061403508777e-06,
+      "loss": 0.2745,
+      "step": 13591
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.575657894736842e-06,
+      "loss": 0.7427,
+      "step": 13592
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.5751096491228074e-06,
+      "loss": 0.9763,
+      "step": 13593
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.574561403508772e-06,
+      "loss": 0.1974,
+      "step": 13594
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.574013157894737e-06,
+      "loss": 0.1589,
+      "step": 13595
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.5734649122807023e-06,
+      "loss": 0.502,
+      "step": 13596
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.5729166666666667e-06,
+      "loss": 0.6917,
+      "step": 13597
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.572368421052632e-06,
+      "loss": 0.5374,
+      "step": 13598
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.5718201754385967e-06,
+      "loss": 0.2828,
+      "step": 13599
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.5712719298245616e-06,
+      "loss": 0.3508,
+      "step": 13600
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.5707236842105264e-06,
+      "loss": 0.6513,
+      "step": 13601
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.5701754385964916e-06,
+      "loss": 0.6614,
+      "step": 13602
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.569627192982456e-06,
+      "loss": 0.5015,
+      "step": 13603
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.5690789473684213e-06,
+      "loss": 0.5812,
+      "step": 13604
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.568530701754386e-06,
+      "loss": 1.01,
+      "step": 13605
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.567982456140351e-06,
+      "loss": 0.4409,
+      "step": 13606
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.567434210526316e-06,
+      "loss": 1.0291,
+      "step": 13607
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.5668859649122806e-06,
+      "loss": 0.5503,
+      "step": 13608
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.566337719298246e-06,
+      "loss": 0.7011,
+      "step": 13609
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.565789473684211e-06,
+      "loss": 0.4495,
+      "step": 13610
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.5652412280701755e-06,
+      "loss": 0.102,
+      "step": 13611
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.5646929824561407e-06,
+      "loss": 0.4557,
+      "step": 13612
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.5641447368421056e-06,
+      "loss": 0.3742,
+      "step": 13613
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.5635964912280704e-06,
+      "loss": 1.429,
+      "step": 13614
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.5630482456140356e-06,
+      "loss": 0.5719,
+      "step": 13615
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.5625e-06,
+      "loss": 0.5191,
+      "step": 13616
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.5619517543859653e-06,
+      "loss": 0.6785,
+      "step": 13617
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.56140350877193e-06,
+      "loss": 0.3667,
+      "step": 13618
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.560855263157895e-06,
+      "loss": 0.2466,
+      "step": 13619
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.5603070175438598e-06,
+      "loss": 0.9431,
+      "step": 13620
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.559758771929825e-06,
+      "loss": 0.2692,
+      "step": 13621
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.5592105263157894e-06,
+      "loss": 0.5317,
+      "step": 13622
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.5586622807017546e-06,
+      "loss": 0.5529,
+      "step": 13623
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.5581140350877195e-06,
+      "loss": 0.6818,
+      "step": 13624
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.5575657894736843e-06,
+      "loss": 0.3409,
+      "step": 13625
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.5570175438596495e-06,
+      "loss": 0.4229,
+      "step": 13626
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.556469298245614e-06,
+      "loss": 0.657,
+      "step": 13627
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.555921052631579e-06,
+      "loss": 0.3127,
+      "step": 13628
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.5553728070175444e-06,
+      "loss": 0.4754,
+      "step": 13629
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.554824561403509e-06,
+      "loss": 0.8708,
+      "step": 13630
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.554276315789474e-06,
+      "loss": 0.2655,
+      "step": 13631
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.5537280701754385e-06,
+      "loss": 0.7189,
+      "step": 13632
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.5531798245614037e-06,
+      "loss": 0.6115,
+      "step": 13633
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.552631578947369e-06,
+      "loss": 0.2813,
+      "step": 13634
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.5520833333333334e-06,
+      "loss": 0.5571,
+      "step": 13635
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.5515350877192986e-06,
+      "loss": 0.3523,
+      "step": 13636
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.5509868421052635e-06,
+      "loss": 0.5569,
+      "step": 13637
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.5504385964912283e-06,
+      "loss": 0.6053,
+      "step": 13638
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.549890350877193e-06,
+      "loss": 0.3944,
+      "step": 13639
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.549342105263158e-06,
+      "loss": 0.7113,
+      "step": 13640
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.5487938596491228e-06,
+      "loss": 1.2227,
+      "step": 13641
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.548245614035088e-06,
+      "loss": 0.114,
+      "step": 13642
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.547697368421053e-06,
+      "loss": 1.3777,
+      "step": 13643
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.5471491228070177e-06,
+      "loss": 0.4864,
+      "step": 13644
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.546600877192983e-06,
+      "loss": 0.1414,
+      "step": 13645
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.5460526315789473e-06,
+      "loss": 0.4405,
+      "step": 13646
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.5455043859649126e-06,
+      "loss": 0.6869,
+      "step": 13647
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.544956140350877e-06,
+      "loss": 1.2529,
+      "step": 13648
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.544407894736842e-06,
+      "loss": 0.5893,
+      "step": 13649
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.5438596491228075e-06,
+      "loss": 0.3123,
+      "step": 13650
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.543311403508772e-06,
+      "loss": 0.65,
+      "step": 13651
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.542763157894737e-06,
+      "loss": 0.3316,
+      "step": 13652
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.5422149122807024e-06,
+      "loss": 0.6144,
+      "step": 13653
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.5416666666666668e-06,
+      "loss": 0.5321,
+      "step": 13654
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.541118421052632e-06,
+      "loss": 0.176,
+      "step": 13655
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.540570175438597e-06,
+      "loss": 0.0749,
+      "step": 13656
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.5400219298245617e-06,
+      "loss": 0.6966,
+      "step": 13657
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.5394736842105265e-06,
+      "loss": 0.1171,
+      "step": 13658
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.5389254385964913e-06,
+      "loss": 0.2737,
+      "step": 13659
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.538377192982456e-06,
+      "loss": 0.1278,
+      "step": 13660
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.5378289473684214e-06,
+      "loss": 0.2654,
+      "step": 13661
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.537280701754386e-06,
+      "loss": 0.3688,
+      "step": 13662
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.536732456140351e-06,
+      "loss": 0.9628,
+      "step": 13663
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.5361842105263163e-06,
+      "loss": 1.125,
+      "step": 13664
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.5356359649122807e-06,
+      "loss": 0.8658,
+      "step": 13665
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.535087719298246e-06,
+      "loss": 0.176,
+      "step": 13666
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.5345394736842103e-06,
+      "loss": 0.3456,
+      "step": 13667
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.5339912280701756e-06,
+      "loss": 0.2302,
+      "step": 13668
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.533442982456141e-06,
+      "loss": 0.1925,
+      "step": 13669
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.5328947368421052e-06,
+      "loss": 0.3052,
+      "step": 13670
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.5323464912280705e-06,
+      "loss": 1.4525,
+      "step": 13671
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.5317982456140357e-06,
+      "loss": 0.6249,
+      "step": 13672
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.53125e-06,
+      "loss": 0.5071,
+      "step": 13673
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.5307017543859654e-06,
+      "loss": 0.6411,
+      "step": 13674
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.5301535087719298e-06,
+      "loss": 0.2104,
+      "step": 13675
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.529605263157895e-06,
+      "loss": 0.4164,
+      "step": 13676
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.52905701754386e-06,
+      "loss": 0.5987,
+      "step": 13677
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.5285087719298247e-06,
+      "loss": 0.422,
+      "step": 13678
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.5279605263157895e-06,
+      "loss": 0.764,
+      "step": 13679
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.5274122807017547e-06,
+      "loss": 0.2625,
+      "step": 13680
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.5268640350877196e-06,
+      "loss": 0.1691,
+      "step": 13681
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.5263157894736844e-06,
+      "loss": 0.4943,
+      "step": 13682
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.525767543859649e-06,
+      "loss": 0.5129,
+      "step": 13683
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.525219298245614e-06,
+      "loss": 1.1482,
+      "step": 13684
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.5246710526315793e-06,
+      "loss": 0.3113,
+      "step": 13685
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.5241228070175437e-06,
+      "loss": 1.6816,
+      "step": 13686
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.523574561403509e-06,
+      "loss": 0.1973,
+      "step": 13687
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.523026315789474e-06,
+      "loss": 0.7971,
+      "step": 13688
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.5224780701754386e-06,
+      "loss": 0.2131,
+      "step": 13689
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.521929824561404e-06,
+      "loss": 1.2158,
+      "step": 13690
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.521381578947369e-06,
+      "loss": 0.5928,
+      "step": 13691
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.5208333333333335e-06,
+      "loss": 0.2024,
+      "step": 13692
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.5202850877192987e-06,
+      "loss": 0.2687,
+      "step": 13693
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.519736842105263e-06,
+      "loss": 0.2814,
+      "step": 13694
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.5191885964912284e-06,
+      "loss": 0.4268,
+      "step": 13695
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.518640350877193e-06,
+      "loss": 1.1216,
+      "step": 13696
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.518092105263158e-06,
+      "loss": 1.1345,
+      "step": 13697
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.5175438596491233e-06,
+      "loss": 0.5447,
+      "step": 13698
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.516995614035088e-06,
+      "loss": 0.0785,
+      "step": 13699
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.516447368421053e-06,
+      "loss": 0.4237,
+      "step": 13700
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.5158991228070177e-06,
+      "loss": 0.6655,
+      "step": 13701
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.5153508771929826e-06,
+      "loss": 0.9819,
+      "step": 13702
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.5148026315789474e-06,
+      "loss": 0.4553,
+      "step": 13703
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.5142543859649126e-06,
+      "loss": 0.8305,
+      "step": 13704
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.513706140350877e-06,
+      "loss": 0.1644,
+      "step": 13705
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.5131578947368423e-06,
+      "loss": 0.2499,
+      "step": 13706
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.5126096491228075e-06,
+      "loss": 1.4127,
+      "step": 13707
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.512061403508772e-06,
+      "loss": 0.4887,
+      "step": 13708
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.511513157894737e-06,
+      "loss": 0.2377,
+      "step": 13709
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.5109649122807016e-06,
+      "loss": 0.6404,
+      "step": 13710
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.510416666666667e-06,
+      "loss": 0.5293,
+      "step": 13711
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.509868421052632e-06,
+      "loss": 0.3847,
+      "step": 13712
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.5093201754385965e-06,
+      "loss": 0.1582,
+      "step": 13713
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.5087719298245617e-06,
+      "loss": 0.2164,
+      "step": 13714
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.5082236842105266e-06,
+      "loss": 0.3777,
+      "step": 13715
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.5076754385964914e-06,
+      "loss": 0.4555,
+      "step": 13716
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.5071271929824566e-06,
+      "loss": 0.8619,
+      "step": 13717
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.506578947368421e-06,
+      "loss": 0.3849,
+      "step": 13718
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.5060307017543863e-06,
+      "loss": 0.2272,
+      "step": 13719
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.505482456140351e-06,
+      "loss": 0.9936,
+      "step": 13720
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.504934210526316e-06,
+      "loss": 0.2468,
+      "step": 13721
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.5043859649122808e-06,
+      "loss": 0.6048,
+      "step": 13722
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.503837719298246e-06,
+      "loss": 0.7678,
+      "step": 13723
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.5032894736842104e-06,
+      "loss": 0.9613,
+      "step": 13724
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.5027412280701757e-06,
+      "loss": 0.546,
+      "step": 13725
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.5021929824561405e-06,
+      "loss": 1.1319,
+      "step": 13726
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.5016447368421053e-06,
+      "loss": 0.703,
+      "step": 13727
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.5010964912280706e-06,
+      "loss": 0.4585,
+      "step": 13728
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.500548245614035e-06,
+      "loss": 0.1089,
+      "step": 13729
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.5e-06,
+      "loss": 0.6602,
+      "step": 13730
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.499451754385965e-06,
+      "loss": 0.5897,
+      "step": 13731
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.4989035087719303e-06,
+      "loss": 0.1866,
+      "step": 13732
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.498355263157895e-06,
+      "loss": 0.1664,
+      "step": 13733
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.49780701754386e-06,
+      "loss": 0.9037,
+      "step": 13734
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.4972587719298248e-06,
+      "loss": 1.2139,
+      "step": 13735
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.49671052631579e-06,
+      "loss": 0.9936,
+      "step": 13736
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.496162280701755e-06,
+      "loss": 0.2104,
+      "step": 13737
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.4956140350877196e-06,
+      "loss": 0.4591,
+      "step": 13738
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.4950657894736845e-06,
+      "loss": 0.3646,
+      "step": 13739
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.4945175438596493e-06,
+      "loss": 0.3362,
+      "step": 13740
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.493969298245614e-06,
+      "loss": 0.3742,
+      "step": 13741
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.493421052631579e-06,
+      "loss": 0.4093,
+      "step": 13742
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.4928728070175438e-06,
+      "loss": 0.8648,
+      "step": 13743
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.492324561403509e-06,
+      "loss": 0.4996,
+      "step": 13744
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.491776315789474e-06,
+      "loss": 0.3357,
+      "step": 13745
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.4912280701754387e-06,
+      "loss": 0.5427,
+      "step": 13746
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.4906798245614035e-06,
+      "loss": 0.36,
+      "step": 13747
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.4901315789473687e-06,
+      "loss": 0.8525,
+      "step": 13748
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.4895833333333336e-06,
+      "loss": 1.5818,
+      "step": 13749
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.4890350877192984e-06,
+      "loss": 0.6501,
+      "step": 13750
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.4884868421052632e-06,
+      "loss": 0.3715,
+      "step": 13751
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.4879385964912285e-06,
+      "loss": 0.725,
+      "step": 13752
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.4873903508771933e-06,
+      "loss": 0.4255,
+      "step": 13753
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.486842105263158e-06,
+      "loss": 0.337,
+      "step": 13754
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.486293859649123e-06,
+      "loss": 0.2797,
+      "step": 13755
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.485745614035088e-06,
+      "loss": 1.2785,
+      "step": 13756
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.485197368421053e-06,
+      "loss": 0.4196,
+      "step": 13757
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.484649122807018e-06,
+      "loss": 0.1419,
+      "step": 13758
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.4841008771929827e-06,
+      "loss": 0.8287,
+      "step": 13759
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.4835526315789475e-06,
+      "loss": 1.1146,
+      "step": 13760
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.4830043859649123e-06,
+      "loss": 0.5157,
+      "step": 13761
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.482456140350877e-06,
+      "loss": 0.7144,
+      "step": 13762
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.4819078947368424e-06,
+      "loss": 0.4339,
+      "step": 13763
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.481359649122807e-06,
+      "loss": 0.4572,
+      "step": 13764
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.480811403508772e-06,
+      "loss": 0.4382,
+      "step": 13765
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.480263157894737e-06,
+      "loss": 0.5857,
+      "step": 13766
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.4797149122807017e-06,
+      "loss": 0.3875,
+      "step": 13767
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.479166666666667e-06,
+      "loss": 0.2204,
+      "step": 13768
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.4786184210526318e-06,
+      "loss": 0.2026,
+      "step": 13769
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.4780701754385966e-06,
+      "loss": 0.1348,
+      "step": 13770
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.477521929824562e-06,
+      "loss": 0.5386,
+      "step": 13771
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.4769736842105267e-06,
+      "loss": 0.2055,
+      "step": 13772
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.4764254385964915e-06,
+      "loss": 0.2614,
+      "step": 13773
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.4758771929824563e-06,
+      "loss": 0.3838,
+      "step": 13774
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.4753289473684215e-06,
+      "loss": 0.3422,
+      "step": 13775
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.4747807017543864e-06,
+      "loss": 1.5168,
+      "step": 13776
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.474232456140351e-06,
+      "loss": 0.2034,
+      "step": 13777
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.473684210526316e-06,
+      "loss": 1.3258,
+      "step": 13778
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.473135964912281e-06,
+      "loss": 0.4392,
+      "step": 13779
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.4725877192982457e-06,
+      "loss": 1.494,
+      "step": 13780
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.472039473684211e-06,
+      "loss": 0.5857,
+      "step": 13781
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.4714912280701757e-06,
+      "loss": 0.1762,
+      "step": 13782
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.4709429824561406e-06,
+      "loss": 0.132,
+      "step": 13783
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.4703947368421054e-06,
+      "loss": 0.6464,
+      "step": 13784
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.4698464912280702e-06,
+      "loss": 0.5118,
+      "step": 13785
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.469298245614035e-06,
+      "loss": 1.0407,
+      "step": 13786
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.4687500000000003e-06,
+      "loss": 0.7855,
+      "step": 13787
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.468201754385965e-06,
+      "loss": 1.0619,
+      "step": 13788
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.46765350877193e-06,
+      "loss": 0.6037,
+      "step": 13789
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.4671052631578948e-06,
+      "loss": 0.522,
+      "step": 13790
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.46655701754386e-06,
+      "loss": 0.2413,
+      "step": 13791
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.466008771929825e-06,
+      "loss": 0.59,
+      "step": 13792
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.4654605263157897e-06,
+      "loss": 0.7451,
+      "step": 13793
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.4649122807017545e-06,
+      "loss": 0.3957,
+      "step": 13794
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.4643640350877197e-06,
+      "loss": 0.1321,
+      "step": 13795
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.4638157894736846e-06,
+      "loss": 0.2257,
+      "step": 13796
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.4632675438596494e-06,
+      "loss": 0.5671,
+      "step": 13797
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.462719298245614e-06,
+      "loss": 0.6648,
+      "step": 13798
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.462171052631579e-06,
+      "loss": 0.4432,
+      "step": 13799
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.4616228070175443e-06,
+      "loss": 0.3505,
+      "step": 13800
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.461074561403509e-06,
+      "loss": 1.4058,
+      "step": 13801
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.460526315789474e-06,
+      "loss": 0.3288,
+      "step": 13802
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.4599780701754388e-06,
+      "loss": 0.4389,
+      "step": 13803
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.4594298245614036e-06,
+      "loss": 0.3727,
+      "step": 13804
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.4588815789473684e-06,
+      "loss": 0.2421,
+      "step": 13805
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.4583333333333332e-06,
+      "loss": 0.3852,
+      "step": 13806
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.4577850877192985e-06,
+      "loss": 0.3621,
+      "step": 13807
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.4572368421052633e-06,
+      "loss": 0.1204,
+      "step": 13808
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.456688596491228e-06,
+      "loss": 0.3823,
+      "step": 13809
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.456140350877193e-06,
+      "loss": 0.2326,
+      "step": 13810
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.455592105263158e-06,
+      "loss": 0.3052,
+      "step": 13811
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.455043859649123e-06,
+      "loss": 0.1861,
+      "step": 13812
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.454495614035088e-06,
+      "loss": 0.2927,
+      "step": 13813
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.453947368421053e-06,
+      "loss": 0.1515,
+      "step": 13814
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.453399122807018e-06,
+      "loss": 0.7379,
+      "step": 13815
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.4528508771929827e-06,
+      "loss": 0.1925,
+      "step": 13816
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.4523026315789476e-06,
+      "loss": 0.4886,
+      "step": 13817
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.4517543859649124e-06,
+      "loss": 0.3056,
+      "step": 13818
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.4512061403508776e-06,
+      "loss": 0.5644,
+      "step": 13819
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.4506578947368425e-06,
+      "loss": 1.0367,
+      "step": 13820
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.4501096491228073e-06,
+      "loss": 1.1608,
+      "step": 13821
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.449561403508772e-06,
+      "loss": 0.3031,
+      "step": 13822
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.449013157894737e-06,
+      "loss": 0.6024,
+      "step": 13823
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.4484649122807018e-06,
+      "loss": 0.9255,
+      "step": 13824
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.4479166666666666e-06,
+      "loss": 0.9936,
+      "step": 13825
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.447368421052632e-06,
+      "loss": 0.5203,
+      "step": 13826
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.4468201754385967e-06,
+      "loss": 0.9409,
+      "step": 13827
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.4462719298245615e-06,
+      "loss": 0.3818,
+      "step": 13828
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.4457236842105263e-06,
+      "loss": 0.7944,
+      "step": 13829
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.4451754385964916e-06,
+      "loss": 0.273,
+      "step": 13830
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.4446271929824564e-06,
+      "loss": 1.1978,
+      "step": 13831
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.4440789473684212e-06,
+      "loss": 0.2577,
+      "step": 13832
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.443530701754386e-06,
+      "loss": 0.2126,
+      "step": 13833
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.4429824561403513e-06,
+      "loss": 0.7643,
+      "step": 13834
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.442434210526316e-06,
+      "loss": 0.3617,
+      "step": 13835
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.441885964912281e-06,
+      "loss": 0.4645,
+      "step": 13836
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.4413377192982458e-06,
+      "loss": 0.2908,
+      "step": 13837
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.440789473684211e-06,
+      "loss": 0.6054,
+      "step": 13838
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.440241228070176e-06,
+      "loss": 0.2821,
+      "step": 13839
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.4396929824561407e-06,
+      "loss": 0.1026,
+      "step": 13840
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.4391447368421055e-06,
+      "loss": 0.0545,
+      "step": 13841
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.4385964912280703e-06,
+      "loss": 0.3931,
+      "step": 13842
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.438048245614035e-06,
+      "loss": 0.4852,
+      "step": 13843
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.4375e-06,
+      "loss": 0.5674,
+      "step": 13844
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.4369517543859648e-06,
+      "loss": 0.7336,
+      "step": 13845
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.43640350877193e-06,
+      "loss": 0.1621,
+      "step": 13846
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.435855263157895e-06,
+      "loss": 0.2678,
+      "step": 13847
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.4353070175438597e-06,
+      "loss": 0.4125,
+      "step": 13848
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.4347587719298245e-06,
+      "loss": 0.4154,
+      "step": 13849
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.4342105263157898e-06,
+      "loss": 0.6596,
+      "step": 13850
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.4336622807017546e-06,
+      "loss": 0.6191,
+      "step": 13851
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.4331140350877194e-06,
+      "loss": 0.0561,
+      "step": 13852
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.4325657894736846e-06,
+      "loss": 0.2463,
+      "step": 13853
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.4320175438596495e-06,
+      "loss": 0.4329,
+      "step": 13854
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.4314692982456143e-06,
+      "loss": 0.2269,
+      "step": 13855
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.430921052631579e-06,
+      "loss": 0.6458,
+      "step": 13856
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.4303728070175444e-06,
+      "loss": 0.1825,
+      "step": 13857
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.429824561403509e-06,
+      "loss": 0.303,
+      "step": 13858
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.429276315789474e-06,
+      "loss": 0.1287,
+      "step": 13859
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.428728070175439e-06,
+      "loss": 0.8064,
+      "step": 13860
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.4281798245614037e-06,
+      "loss": 0.6432,
+      "step": 13861
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.4276315789473685e-06,
+      "loss": 0.7396,
+      "step": 13862
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.4270833333333333e-06,
+      "loss": 0.2035,
+      "step": 13863
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.426535087719298e-06,
+      "loss": 0.2856,
+      "step": 13864
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.4259868421052634e-06,
+      "loss": 0.7578,
+      "step": 13865
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.4254385964912282e-06,
+      "loss": 0.2972,
+      "step": 13866
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.424890350877193e-06,
+      "loss": 0.094,
+      "step": 13867
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.424342105263158e-06,
+      "loss": 1.2729,
+      "step": 13868
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.423793859649123e-06,
+      "loss": 0.6367,
+      "step": 13869
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.423245614035088e-06,
+      "loss": 0.1359,
+      "step": 13870
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.4226973684210528e-06,
+      "loss": 0.774,
+      "step": 13871
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.4221491228070176e-06,
+      "loss": 0.64,
+      "step": 13872
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.421600877192983e-06,
+      "loss": 0.9522,
+      "step": 13873
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.4210526315789477e-06,
+      "loss": 0.5832,
+      "step": 13874
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.4205043859649125e-06,
+      "loss": 0.4733,
+      "step": 13875
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.4199561403508773e-06,
+      "loss": 0.6669,
+      "step": 13876
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.4194078947368426e-06,
+      "loss": 0.5227,
+      "step": 13877
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.4188596491228074e-06,
+      "loss": 0.6889,
+      "step": 13878
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.418311403508772e-06,
+      "loss": 0.2987,
+      "step": 13879
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.417763157894737e-06,
+      "loss": 0.3674,
+      "step": 13880
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.417214912280702e-06,
+      "loss": 1.273,
+      "step": 13881
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.4166666666666667e-06,
+      "loss": 0.2393,
+      "step": 13882
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.416118421052632e-06,
+      "loss": 0.6946,
+      "step": 13883
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.4155701754385968e-06,
+      "loss": 0.699,
+      "step": 13884
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.4150219298245616e-06,
+      "loss": 0.7777,
+      "step": 13885
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.4144736842105264e-06,
+      "loss": 0.355,
+      "step": 13886
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.4139254385964912e-06,
+      "loss": 1.1096,
+      "step": 13887
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.413377192982456e-06,
+      "loss": 0.6398,
+      "step": 13888
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.4128289473684213e-06,
+      "loss": 0.7246,
+      "step": 13889
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.412280701754386e-06,
+      "loss": 0.1337,
+      "step": 13890
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.411732456140351e-06,
+      "loss": 0.5223,
+      "step": 13891
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.4111842105263158e-06,
+      "loss": 0.4471,
+      "step": 13892
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.410635964912281e-06,
+      "loss": 0.8611,
+      "step": 13893
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.410087719298246e-06,
+      "loss": 0.5659,
+      "step": 13894
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.4095394736842107e-06,
+      "loss": 0.531,
+      "step": 13895
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.408991228070176e-06,
+      "loss": 0.4489,
+      "step": 13896
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.4084429824561407e-06,
+      "loss": 0.2389,
+      "step": 13897
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.4078947368421056e-06,
+      "loss": 1.5515,
+      "step": 13898
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.4073464912280704e-06,
+      "loss": 0.6895,
+      "step": 13899
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.4067982456140352e-06,
+      "loss": 0.1751,
+      "step": 13900
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.40625e-06,
+      "loss": 1.452,
+      "step": 13901
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.4057017543859653e-06,
+      "loss": 0.3357,
+      "step": 13902
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.40515350877193e-06,
+      "loss": 0.4314,
+      "step": 13903
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.404605263157895e-06,
+      "loss": 0.3207,
+      "step": 13904
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.4040570175438598e-06,
+      "loss": 0.6693,
+      "step": 13905
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.4035087719298246e-06,
+      "loss": 0.6194,
+      "step": 13906
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.4029605263157894e-06,
+      "loss": 0.3549,
+      "step": 13907
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.4024122807017547e-06,
+      "loss": 1.351,
+      "step": 13908
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.4018640350877195e-06,
+      "loss": 0.5772,
+      "step": 13909
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.4013157894736843e-06,
+      "loss": 0.3502,
+      "step": 13910
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.400767543859649e-06,
+      "loss": 0.4112,
+      "step": 13911
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.4002192982456144e-06,
+      "loss": 0.7469,
+      "step": 13912
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.399671052631579e-06,
+      "loss": 1.6228,
+      "step": 13913
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.399122807017544e-06,
+      "loss": 0.6319,
+      "step": 13914
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.398574561403509e-06,
+      "loss": 0.8687,
+      "step": 13915
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.398026315789474e-06,
+      "loss": 1.1869,
+      "step": 13916
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.397478070175439e-06,
+      "loss": 0.184,
+      "step": 13917
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.3969298245614038e-06,
+      "loss": 0.2194,
+      "step": 13918
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.3963815789473686e-06,
+      "loss": 0.3506,
+      "step": 13919
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.395833333333334e-06,
+      "loss": 0.6738,
+      "step": 13920
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.3952850877192987e-06,
+      "loss": 1.1595,
+      "step": 13921
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.3947368421052635e-06,
+      "loss": 0.8866,
+      "step": 13922
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.3941885964912283e-06,
+      "loss": 0.3689,
+      "step": 13923
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.393640350877193e-06,
+      "loss": 0.2523,
+      "step": 13924
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.393092105263158e-06,
+      "loss": 1.0054,
+      "step": 13925
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.3925438596491228e-06,
+      "loss": 0.2122,
+      "step": 13926
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.3919956140350876e-06,
+      "loss": 0.5606,
+      "step": 13927
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.391447368421053e-06,
+      "loss": 0.5362,
+      "step": 13928
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.3908991228070177e-06,
+      "loss": 1.1852,
+      "step": 13929
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.3903508771929825e-06,
+      "loss": 0.6382,
+      "step": 13930
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.3898026315789473e-06,
+      "loss": 0.1385,
+      "step": 13931
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.3892543859649126e-06,
+      "loss": 1.4666,
+      "step": 13932
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.3887061403508774e-06,
+      "loss": 0.1687,
+      "step": 13933
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.3881578947368422e-06,
+      "loss": 0.8869,
+      "step": 13934
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.3876096491228075e-06,
+      "loss": 0.5818,
+      "step": 13935
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.3870614035087723e-06,
+      "loss": 0.5139,
+      "step": 13936
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.386513157894737e-06,
+      "loss": 0.2689,
+      "step": 13937
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.385964912280702e-06,
+      "loss": 0.3502,
+      "step": 13938
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.385416666666667e-06,
+      "loss": 1.1555,
+      "step": 13939
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.384868421052632e-06,
+      "loss": 0.7787,
+      "step": 13940
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.384320175438597e-06,
+      "loss": 0.2185,
+      "step": 13941
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.3837719298245617e-06,
+      "loss": 0.6175,
+      "step": 13942
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.3832236842105265e-06,
+      "loss": 0.1068,
+      "step": 13943
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.3826754385964913e-06,
+      "loss": 0.7765,
+      "step": 13944
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.382127192982456e-06,
+      "loss": 0.2428,
+      "step": 13945
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.381578947368421e-06,
+      "loss": 0.6327,
+      "step": 13946
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.3810307017543862e-06,
+      "loss": 0.7458,
+      "step": 13947
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.380482456140351e-06,
+      "loss": 0.4899,
+      "step": 13948
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.379934210526316e-06,
+      "loss": 0.3119,
+      "step": 13949
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.3793859649122807e-06,
+      "loss": 0.9955,
+      "step": 13950
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.378837719298246e-06,
+      "loss": 1.0554,
+      "step": 13951
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.3782894736842108e-06,
+      "loss": 0.389,
+      "step": 13952
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.3777412280701756e-06,
+      "loss": 0.2154,
+      "step": 13953
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.3771929824561404e-06,
+      "loss": 0.0753,
+      "step": 13954
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.3766447368421057e-06,
+      "loss": 1.6862,
+      "step": 13955
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.3760964912280705e-06,
+      "loss": 0.5503,
+      "step": 13956
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.3755482456140353e-06,
+      "loss": 0.3544,
+      "step": 13957
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.375e-06,
+      "loss": 0.1092,
+      "step": 13958
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.3744517543859654e-06,
+      "loss": 0.1724,
+      "step": 13959
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.37390350877193e-06,
+      "loss": 0.2236,
+      "step": 13960
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.373355263157895e-06,
+      "loss": 0.3808,
+      "step": 13961
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.37280701754386e-06,
+      "loss": 0.6945,
+      "step": 13962
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.3722587719298247e-06,
+      "loss": 0.0468,
+      "step": 13963
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.3717105263157895e-06,
+      "loss": 0.7082,
+      "step": 13964
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.3711622807017543e-06,
+      "loss": 0.8582,
+      "step": 13965
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.3706140350877196e-06,
+      "loss": 0.3446,
+      "step": 13966
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.3700657894736844e-06,
+      "loss": 0.1886,
+      "step": 13967
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.3695175438596492e-06,
+      "loss": 0.7745,
+      "step": 13968
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.368969298245614e-06,
+      "loss": 0.7279,
+      "step": 13969
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.368421052631579e-06,
+      "loss": 0.0958,
+      "step": 13970
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.367872807017544e-06,
+      "loss": 1.1741,
+      "step": 13971
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.367324561403509e-06,
+      "loss": 0.2953,
+      "step": 13972
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.3667763157894738e-06,
+      "loss": 0.1488,
+      "step": 13973
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.3662280701754386e-06,
+      "loss": 0.8547,
+      "step": 13974
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.365679824561404e-06,
+      "loss": 0.2548,
+      "step": 13975
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.3651315789473687e-06,
+      "loss": 0.3105,
+      "step": 13976
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.3645833333333335e-06,
+      "loss": 0.5079,
+      "step": 13977
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.3640350877192987e-06,
+      "loss": 0.1815,
+      "step": 13978
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.3634868421052636e-06,
+      "loss": 0.6429,
+      "step": 13979
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.3629385964912284e-06,
+      "loss": 0.4681,
+      "step": 13980
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.3623903508771932e-06,
+      "loss": 0.4044,
+      "step": 13981
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.361842105263158e-06,
+      "loss": 1.2298,
+      "step": 13982
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.361293859649123e-06,
+      "loss": 0.8719,
+      "step": 13983
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.3607456140350877e-06,
+      "loss": 0.6637,
+      "step": 13984
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.360197368421053e-06,
+      "loss": 0.1884,
+      "step": 13985
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.3596491228070178e-06,
+      "loss": 0.2078,
+      "step": 13986
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.3591008771929826e-06,
+      "loss": 0.4093,
+      "step": 13987
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.3585526315789474e-06,
+      "loss": 0.3977,
+      "step": 13988
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.3580043859649122e-06,
+      "loss": 0.5767,
+      "step": 13989
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.3574561403508775e-06,
+      "loss": 0.5691,
+      "step": 13990
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.3569078947368423e-06,
+      "loss": 0.2709,
+      "step": 13991
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.356359649122807e-06,
+      "loss": 0.717,
+      "step": 13992
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.355811403508772e-06,
+      "loss": 0.1231,
+      "step": 13993
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.355263157894737e-06,
+      "loss": 0.7117,
+      "step": 13994
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.354714912280702e-06,
+      "loss": 0.6172,
+      "step": 13995
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.354166666666667e-06,
+      "loss": 0.1946,
+      "step": 13996
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.3536184210526317e-06,
+      "loss": 0.2945,
+      "step": 13997
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.353070175438597e-06,
+      "loss": 0.3399,
+      "step": 13998
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.3525219298245618e-06,
+      "loss": 1.0292,
+      "step": 13999
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.3519736842105266e-06,
+      "loss": 0.3339,
+      "step": 14000
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.3514254385964914e-06,
+      "loss": 0.4077,
+      "step": 14001
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.3508771929824562e-06,
+      "loss": 0.4931,
+      "step": 14002
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.3503289473684215e-06,
+      "loss": 1.0858,
+      "step": 14003
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.3497807017543863e-06,
+      "loss": 0.2933,
+      "step": 14004
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.349232456140351e-06,
+      "loss": 0.9768,
+      "step": 14005
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.348684210526316e-06,
+      "loss": 0.6855,
+      "step": 14006
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.3481359649122808e-06,
+      "loss": 0.9481,
+      "step": 14007
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.3475877192982456e-06,
+      "loss": 0.2487,
+      "step": 14008
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.3470394736842104e-06,
+      "loss": 0.5958,
+      "step": 14009
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.3464912280701757e-06,
+      "loss": 0.6308,
+      "step": 14010
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.3459429824561405e-06,
+      "loss": 0.1146,
+      "step": 14011
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.3453947368421053e-06,
+      "loss": 0.2782,
+      "step": 14012
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.34484649122807e-06,
+      "loss": 0.6653,
+      "step": 14013
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.3442982456140354e-06,
+      "loss": 0.2637,
+      "step": 14014
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.3437500000000002e-06,
+      "loss": 1.5543,
+      "step": 14015
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.343201754385965e-06,
+      "loss": 0.181,
+      "step": 14016
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.3426535087719303e-06,
+      "loss": 0.4336,
+      "step": 14017
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.342105263157895e-06,
+      "loss": 0.6728,
+      "step": 14018
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.34155701754386e-06,
+      "loss": 0.8693,
+      "step": 14019
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.3410087719298248e-06,
+      "loss": 0.1162,
+      "step": 14020
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.3404605263157896e-06,
+      "loss": 0.9309,
+      "step": 14021
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.339912280701755e-06,
+      "loss": 0.5875,
+      "step": 14022
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.3393640350877197e-06,
+      "loss": 0.2386,
+      "step": 14023
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.3388157894736845e-06,
+      "loss": 0.104,
+      "step": 14024
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.3382675438596493e-06,
+      "loss": 0.7112,
+      "step": 14025
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.337719298245614e-06,
+      "loss": 0.3762,
+      "step": 14026
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.337171052631579e-06,
+      "loss": 0.1727,
+      "step": 14027
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.3366228070175438e-06,
+      "loss": 0.316,
+      "step": 14028
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.336074561403509e-06,
+      "loss": 0.2096,
+      "step": 14029
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.335526315789474e-06,
+      "loss": 1.1571,
+      "step": 14030
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.3349780701754387e-06,
+      "loss": 0.8002,
+      "step": 14031
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.3344298245614035e-06,
+      "loss": 0.5477,
+      "step": 14032
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.3338815789473688e-06,
+      "loss": 0.7421,
+      "step": 14033
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.3333333333333336e-06,
+      "loss": 0.8669,
+      "step": 14034
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.3327850877192984e-06,
+      "loss": 0.5622,
+      "step": 14035
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.3322368421052632e-06,
+      "loss": 0.5259,
+      "step": 14036
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.3316885964912285e-06,
+      "loss": 0.8332,
+      "step": 14037
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.3311403508771933e-06,
+      "loss": 0.1975,
+      "step": 14038
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.330592105263158e-06,
+      "loss": 0.2481,
+      "step": 14039
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.330043859649123e-06,
+      "loss": 0.9757,
+      "step": 14040
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.329495614035088e-06,
+      "loss": 0.1928,
+      "step": 14041
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.328947368421053e-06,
+      "loss": 0.5527,
+      "step": 14042
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.328399122807018e-06,
+      "loss": 0.3865,
+      "step": 14043
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.3278508771929827e-06,
+      "loss": 0.1352,
+      "step": 14044
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.3273026315789475e-06,
+      "loss": 0.8223,
+      "step": 14045
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.3267543859649123e-06,
+      "loss": 0.3373,
+      "step": 14046
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.326206140350877e-06,
+      "loss": 0.1954,
+      "step": 14047
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.325657894736842e-06,
+      "loss": 0.6983,
+      "step": 14048
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.3251096491228072e-06,
+      "loss": 0.7698,
+      "step": 14049
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.324561403508772e-06,
+      "loss": 0.2105,
+      "step": 14050
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.324013157894737e-06,
+      "loss": 0.6569,
+      "step": 14051
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.3234649122807017e-06,
+      "loss": 0.3818,
+      "step": 14052
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.322916666666667e-06,
+      "loss": 0.2967,
+      "step": 14053
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.3223684210526318e-06,
+      "loss": 0.7125,
+      "step": 14054
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.3218201754385966e-06,
+      "loss": 0.5436,
+      "step": 14055
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.321271929824562e-06,
+      "loss": 0.0958,
+      "step": 14056
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.3207236842105267e-06,
+      "loss": 0.3141,
+      "step": 14057
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.3201754385964915e-06,
+      "loss": 0.7756,
+      "step": 14058
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.3196271929824563e-06,
+      "loss": 0.0968,
+      "step": 14059
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.3190789473684216e-06,
+      "loss": 0.8517,
+      "step": 14060
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.3185307017543864e-06,
+      "loss": 0.4855,
+      "step": 14061
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.3179824561403512e-06,
+      "loss": 0.3039,
+      "step": 14062
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.317434210526316e-06,
+      "loss": 1.208,
+      "step": 14063
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.316885964912281e-06,
+      "loss": 0.3208,
+      "step": 14064
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.3163377192982457e-06,
+      "loss": 0.3633,
+      "step": 14065
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.3157894736842105e-06,
+      "loss": 0.9909,
+      "step": 14066
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.3152412280701753e-06,
+      "loss": 0.2298,
+      "step": 14067
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.3146929824561406e-06,
+      "loss": 0.3013,
+      "step": 14068
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.3141447368421054e-06,
+      "loss": 0.6724,
+      "step": 14069
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.3135964912280702e-06,
+      "loss": 0.6914,
+      "step": 14070
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.313048245614035e-06,
+      "loss": 0.3567,
+      "step": 14071
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.3125000000000003e-06,
+      "loss": 0.6125,
+      "step": 14072
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.311951754385965e-06,
+      "loss": 0.4067,
+      "step": 14073
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.31140350877193e-06,
+      "loss": 0.3115,
+      "step": 14074
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.3108552631578948e-06,
+      "loss": 0.4594,
+      "step": 14075
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.31030701754386e-06,
+      "loss": 1.0557,
+      "step": 14076
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.309758771929825e-06,
+      "loss": 0.488,
+      "step": 14077
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.3092105263157897e-06,
+      "loss": 0.2482,
+      "step": 14078
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.3086622807017545e-06,
+      "loss": 0.2017,
+      "step": 14079
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.3081140350877197e-06,
+      "loss": 0.0816,
+      "step": 14080
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.3075657894736846e-06,
+      "loss": 0.5805,
+      "step": 14081
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.3070175438596494e-06,
+      "loss": 0.6676,
+      "step": 14082
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.3064692982456142e-06,
+      "loss": 1.2228,
+      "step": 14083
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.305921052631579e-06,
+      "loss": 0.3665,
+      "step": 14084
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.305372807017544e-06,
+      "loss": 0.5844,
+      "step": 14085
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.3048245614035087e-06,
+      "loss": 1.1051,
+      "step": 14086
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.304276315789474e-06,
+      "loss": 0.6377,
+      "step": 14087
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.3037280701754388e-06,
+      "loss": 0.0923,
+      "step": 14088
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.3031798245614036e-06,
+      "loss": 0.6739,
+      "step": 14089
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.3026315789473684e-06,
+      "loss": 0.9695,
+      "step": 14090
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.3020833333333332e-06,
+      "loss": 0.2315,
+      "step": 14091
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.3015350877192985e-06,
+      "loss": 1.1737,
+      "step": 14092
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.3009868421052633e-06,
+      "loss": 0.1481,
+      "step": 14093
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.300438596491228e-06,
+      "loss": 0.6377,
+      "step": 14094
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.299890350877193e-06,
+      "loss": 0.5172,
+      "step": 14095
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.2993421052631582e-06,
+      "loss": 0.5454,
+      "step": 14096
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.298793859649123e-06,
+      "loss": 0.6014,
+      "step": 14097
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.298245614035088e-06,
+      "loss": 0.3331,
+      "step": 14098
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.297697368421053e-06,
+      "loss": 0.6407,
+      "step": 14099
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.297149122807018e-06,
+      "loss": 1.4348,
+      "step": 14100
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.2966008771929828e-06,
+      "loss": 0.4422,
+      "step": 14101
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.2960526315789476e-06,
+      "loss": 0.9196,
+      "step": 14102
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.2955043859649124e-06,
+      "loss": 1.1008,
+      "step": 14103
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.2949561403508772e-06,
+      "loss": 0.6577,
+      "step": 14104
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.2944078947368425e-06,
+      "loss": 0.4186,
+      "step": 14105
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.2938596491228073e-06,
+      "loss": 0.5948,
+      "step": 14106
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.293311403508772e-06,
+      "loss": 0.1039,
+      "step": 14107
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.292763157894737e-06,
+      "loss": 0.2844,
+      "step": 14108
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.2922149122807018e-06,
+      "loss": 0.183,
+      "step": 14109
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.2916666666666666e-06,
+      "loss": 0.7984,
+      "step": 14110
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.291118421052632e-06,
+      "loss": 0.4656,
+      "step": 14111
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.2905701754385967e-06,
+      "loss": 0.9182,
+      "step": 14112
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.2900219298245615e-06,
+      "loss": 0.0877,
+      "step": 14113
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.2894736842105263e-06,
+      "loss": 1.4804,
+      "step": 14114
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.2889254385964916e-06,
+      "loss": 0.0386,
+      "step": 14115
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.2883771929824564e-06,
+      "loss": 0.4441,
+      "step": 14116
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.2878289473684212e-06,
+      "loss": 0.4423,
+      "step": 14117
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.287280701754386e-06,
+      "loss": 0.385,
+      "step": 14118
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.2867324561403513e-06,
+      "loss": 1.0949,
+      "step": 14119
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.286184210526316e-06,
+      "loss": 0.377,
+      "step": 14120
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.285635964912281e-06,
+      "loss": 0.3895,
+      "step": 14121
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.2850877192982458e-06,
+      "loss": 0.8858,
+      "step": 14122
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.2845394736842106e-06,
+      "loss": 0.445,
+      "step": 14123
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.283991228070176e-06,
+      "loss": 0.1535,
+      "step": 14124
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.2834429824561407e-06,
+      "loss": 0.3572,
+      "step": 14125
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.2828947368421055e-06,
+      "loss": 0.2605,
+      "step": 14126
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.2823464912280703e-06,
+      "loss": 0.1379,
+      "step": 14127
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.281798245614035e-06,
+      "loss": 0.1203,
+      "step": 14128
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.28125e-06,
+      "loss": 0.0732,
+      "step": 14129
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.280701754385965e-06,
+      "loss": 1.032,
+      "step": 14130
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.28015350877193e-06,
+      "loss": 0.4874,
+      "step": 14131
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.279605263157895e-06,
+      "loss": 0.3196,
+      "step": 14132
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.2790570175438597e-06,
+      "loss": 0.9107,
+      "step": 14133
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.2785087719298245e-06,
+      "loss": 0.6783,
+      "step": 14134
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.2779605263157898e-06,
+      "loss": 0.6999,
+      "step": 14135
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.2774122807017546e-06,
+      "loss": 0.7497,
+      "step": 14136
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.2768640350877194e-06,
+      "loss": 0.5918,
+      "step": 14137
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.2763157894736847e-06,
+      "loss": 0.6995,
+      "step": 14138
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.2757675438596495e-06,
+      "loss": 0.2135,
+      "step": 14139
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.2752192982456143e-06,
+      "loss": 0.5797,
+      "step": 14140
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.274671052631579e-06,
+      "loss": 1.1949,
+      "step": 14141
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.2741228070175444e-06,
+      "loss": 1.1263,
+      "step": 14142
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.273574561403509e-06,
+      "loss": 0.4188,
+      "step": 14143
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.273026315789474e-06,
+      "loss": 1.087,
+      "step": 14144
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.272478070175439e-06,
+      "loss": 0.1551,
+      "step": 14145
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.2719298245614037e-06,
+      "loss": 0.0895,
+      "step": 14146
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.2713815789473685e-06,
+      "loss": 0.2641,
+      "step": 14147
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.2708333333333333e-06,
+      "loss": 0.3428,
+      "step": 14148
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.270285087719298e-06,
+      "loss": 0.3028,
+      "step": 14149
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.2697368421052634e-06,
+      "loss": 1.263,
+      "step": 14150
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.2691885964912282e-06,
+      "loss": 0.3602,
+      "step": 14151
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.268640350877193e-06,
+      "loss": 0.6428,
+      "step": 14152
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.268092105263158e-06,
+      "loss": 0.97,
+      "step": 14153
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.267543859649123e-06,
+      "loss": 0.6569,
+      "step": 14154
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.266995614035088e-06,
+      "loss": 0.8639,
+      "step": 14155
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.2664473684210528e-06,
+      "loss": 0.9635,
+      "step": 14156
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.2658991228070176e-06,
+      "loss": 0.3777,
+      "step": 14157
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.265350877192983e-06,
+      "loss": 0.1766,
+      "step": 14158
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.2648026315789477e-06,
+      "loss": 0.9975,
+      "step": 14159
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.2642543859649125e-06,
+      "loss": 0.6155,
+      "step": 14160
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.2637061403508773e-06,
+      "loss": 0.2751,
+      "step": 14161
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.2631578947368426e-06,
+      "loss": 0.2013,
+      "step": 14162
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.2626096491228074e-06,
+      "loss": 0.2752,
+      "step": 14163
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.2620614035087722e-06,
+      "loss": 0.3134,
+      "step": 14164
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.261513157894737e-06,
+      "loss": 1.1355,
+      "step": 14165
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.260964912280702e-06,
+      "loss": 0.7821,
+      "step": 14166
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.2604166666666667e-06,
+      "loss": 1.1444,
+      "step": 14167
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.2598684210526315e-06,
+      "loss": 1.2844,
+      "step": 14168
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.2593201754385963e-06,
+      "loss": 0.4581,
+      "step": 14169
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.2587719298245616e-06,
+      "loss": 0.5411,
+      "step": 14170
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.2582236842105264e-06,
+      "loss": 0.4897,
+      "step": 14171
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.2576754385964912e-06,
+      "loss": 0.733,
+      "step": 14172
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.257127192982456e-06,
+      "loss": 0.1077,
+      "step": 14173
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.2565789473684213e-06,
+      "loss": 0.2011,
+      "step": 14174
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.256030701754386e-06,
+      "loss": 0.7195,
+      "step": 14175
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.255482456140351e-06,
+      "loss": 0.1489,
+      "step": 14176
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.2549342105263158e-06,
+      "loss": 0.1562,
+      "step": 14177
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.254385964912281e-06,
+      "loss": 0.2696,
+      "step": 14178
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.253837719298246e-06,
+      "loss": 1.2717,
+      "step": 14179
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.2532894736842107e-06,
+      "loss": 0.3434,
+      "step": 14180
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.252741228070176e-06,
+      "loss": 1.2155,
+      "step": 14181
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.2521929824561408e-06,
+      "loss": 0.871,
+      "step": 14182
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.2516447368421056e-06,
+      "loss": 0.468,
+      "step": 14183
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.2510964912280704e-06,
+      "loss": 0.3874,
+      "step": 14184
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.2505482456140352e-06,
+      "loss": 0.7588,
+      "step": 14185
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.25e-06,
+      "loss": 0.626,
+      "step": 14186
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.249451754385965e-06,
+      "loss": 0.5901,
+      "step": 14187
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.24890350877193e-06,
+      "loss": 0.3707,
+      "step": 14188
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.248355263157895e-06,
+      "loss": 0.6481,
+      "step": 14189
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.2478070175438598e-06,
+      "loss": 0.2216,
+      "step": 14190
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.2472587719298246e-06,
+      "loss": 0.1613,
+      "step": 14191
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.2467105263157894e-06,
+      "loss": 0.3247,
+      "step": 14192
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.2461622807017547e-06,
+      "loss": 0.1817,
+      "step": 14193
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.2456140350877195e-06,
+      "loss": 0.7446,
+      "step": 14194
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.2450657894736843e-06,
+      "loss": 0.6022,
+      "step": 14195
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.244517543859649e-06,
+      "loss": 0.9349,
+      "step": 14196
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.2439692982456144e-06,
+      "loss": 0.6869,
+      "step": 14197
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.2434210526315792e-06,
+      "loss": 1.0216,
+      "step": 14198
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.242872807017544e-06,
+      "loss": 0.1363,
+      "step": 14199
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.242324561403509e-06,
+      "loss": 0.2107,
+      "step": 14200
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.241776315789474e-06,
+      "loss": 0.6867,
+      "step": 14201
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.241228070175439e-06,
+      "loss": 0.1831,
+      "step": 14202
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.2406798245614038e-06,
+      "loss": 0.6652,
+      "step": 14203
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.2401315789473686e-06,
+      "loss": 0.241,
+      "step": 14204
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.2395833333333334e-06,
+      "loss": 0.1872,
+      "step": 14205
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.2390350877192982e-06,
+      "loss": 0.5224,
+      "step": 14206
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.2384868421052635e-06,
+      "loss": 0.3804,
+      "step": 14207
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.2379385964912283e-06,
+      "loss": 0.7894,
+      "step": 14208
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.237390350877193e-06,
+      "loss": 0.6179,
+      "step": 14209
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.236842105263158e-06,
+      "loss": 0.5579,
+      "step": 14210
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.236293859649123e-06,
+      "loss": 1.1113,
+      "step": 14211
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.2357456140350876e-06,
+      "loss": 0.3877,
+      "step": 14212
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.235197368421053e-06,
+      "loss": 0.3209,
+      "step": 14213
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.2346491228070177e-06,
+      "loss": 0.7211,
+      "step": 14214
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.2341008771929825e-06,
+      "loss": 0.0793,
+      "step": 14215
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.2335526315789473e-06,
+      "loss": 1.1389,
+      "step": 14216
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.2330043859649126e-06,
+      "loss": 0.389,
+      "step": 14217
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.2324561403508774e-06,
+      "loss": 0.3083,
+      "step": 14218
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.2319078947368422e-06,
+      "loss": 0.359,
+      "step": 14219
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.2313596491228075e-06,
+      "loss": 0.8417,
+      "step": 14220
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.2308114035087723e-06,
+      "loss": 1.053,
+      "step": 14221
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.230263157894737e-06,
+      "loss": 1.1491,
+      "step": 14222
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.229714912280702e-06,
+      "loss": 0.5425,
+      "step": 14223
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.2291666666666668e-06,
+      "loss": 0.87,
+      "step": 14224
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.228618421052632e-06,
+      "loss": 0.4303,
+      "step": 14225
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.228070175438597e-06,
+      "loss": 0.0911,
+      "step": 14226
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.2275219298245617e-06,
+      "loss": 0.279,
+      "step": 14227
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.2269736842105265e-06,
+      "loss": 0.1615,
+      "step": 14228
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.2264254385964913e-06,
+      "loss": 0.3752,
+      "step": 14229
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.225877192982456e-06,
+      "loss": 0.3482,
+      "step": 14230
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.225328947368421e-06,
+      "loss": 0.1153,
+      "step": 14231
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.2247807017543862e-06,
+      "loss": 0.9308,
+      "step": 14232
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.224232456140351e-06,
+      "loss": 0.417,
+      "step": 14233
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.223684210526316e-06,
+      "loss": 0.0791,
+      "step": 14234
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.2231359649122807e-06,
+      "loss": 1.0138,
+      "step": 14235
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.222587719298246e-06,
+      "loss": 0.4047,
+      "step": 14236
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.2220394736842108e-06,
+      "loss": 0.5931,
+      "step": 14237
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.2214912280701756e-06,
+      "loss": 0.9568,
+      "step": 14238
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.2209429824561404e-06,
+      "loss": 0.7103,
+      "step": 14239
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.2203947368421057e-06,
+      "loss": 0.9706,
+      "step": 14240
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.2198464912280705e-06,
+      "loss": 0.4973,
+      "step": 14241
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.2192982456140353e-06,
+      "loss": 0.1607,
+      "step": 14242
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.21875e-06,
+      "loss": 0.3637,
+      "step": 14243
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.2182017543859654e-06,
+      "loss": 0.1594,
+      "step": 14244
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.2176535087719302e-06,
+      "loss": 0.4843,
+      "step": 14245
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.217105263157895e-06,
+      "loss": 0.5735,
+      "step": 14246
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.21655701754386e-06,
+      "loss": 0.6148,
+      "step": 14247
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.2160087719298247e-06,
+      "loss": 0.8225,
+      "step": 14248
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.2154605263157895e-06,
+      "loss": 0.2605,
+      "step": 14249
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.2149122807017543e-06,
+      "loss": 0.2655,
+      "step": 14250
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.214364035087719e-06,
+      "loss": 1.0211,
+      "step": 14251
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.2138157894736844e-06,
+      "loss": 0.6088,
+      "step": 14252
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.2132675438596492e-06,
+      "loss": 0.8007,
+      "step": 14253
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.212719298245614e-06,
+      "loss": 0.7507,
+      "step": 14254
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.212171052631579e-06,
+      "loss": 0.133,
+      "step": 14255
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.211622807017544e-06,
+      "loss": 0.2136,
+      "step": 14256
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.211074561403509e-06,
+      "loss": 1.1241,
+      "step": 14257
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.2105263157894738e-06,
+      "loss": 0.2764,
+      "step": 14258
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.2099780701754386e-06,
+      "loss": 0.709,
+      "step": 14259
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.209429824561404e-06,
+      "loss": 0.5253,
+      "step": 14260
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.2088815789473687e-06,
+      "loss": 0.3044,
+      "step": 14261
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.2083333333333335e-06,
+      "loss": 0.5223,
+      "step": 14262
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.2077850877192988e-06,
+      "loss": 0.5016,
+      "step": 14263
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.2072368421052636e-06,
+      "loss": 0.2529,
+      "step": 14264
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.2066885964912284e-06,
+      "loss": 0.6055,
+      "step": 14265
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.2061403508771932e-06,
+      "loss": 0.9035,
+      "step": 14266
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.205592105263158e-06,
+      "loss": 0.4909,
+      "step": 14267
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.205043859649123e-06,
+      "loss": 0.9358,
+      "step": 14268
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.2044956140350877e-06,
+      "loss": 0.372,
+      "step": 14269
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.2039473684210525e-06,
+      "loss": 0.3179,
+      "step": 14270
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.2033991228070178e-06,
+      "loss": 0.5427,
+      "step": 14271
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.2028508771929826e-06,
+      "loss": 0.2327,
+      "step": 14272
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.2023026315789474e-06,
+      "loss": 0.7951,
+      "step": 14273
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.2017543859649122e-06,
+      "loss": 0.7458,
+      "step": 14274
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.2012061403508775e-06,
+      "loss": 0.3346,
+      "step": 14275
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.2006578947368423e-06,
+      "loss": 1.0274,
+      "step": 14276
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.200109649122807e-06,
+      "loss": 0.9045,
+      "step": 14277
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.199561403508772e-06,
+      "loss": 0.4472,
+      "step": 14278
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.1990131578947372e-06,
+      "loss": 1.4884,
+      "step": 14279
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.198464912280702e-06,
+      "loss": 0.058,
+      "step": 14280
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.197916666666667e-06,
+      "loss": 0.4927,
+      "step": 14281
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.1973684210526317e-06,
+      "loss": 0.9915,
+      "step": 14282
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.196820175438597e-06,
+      "loss": 0.9114,
+      "step": 14283
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.1962719298245618e-06,
+      "loss": 0.6608,
+      "step": 14284
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.1957236842105266e-06,
+      "loss": 0.2073,
+      "step": 14285
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.1951754385964914e-06,
+      "loss": 0.5611,
+      "step": 14286
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.1946271929824562e-06,
+      "loss": 0.0917,
+      "step": 14287
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.194078947368421e-06,
+      "loss": 0.3125,
+      "step": 14288
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.193530701754386e-06,
+      "loss": 0.1495,
+      "step": 14289
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.192982456140351e-06,
+      "loss": 0.3989,
+      "step": 14290
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.192434210526316e-06,
+      "loss": 0.0704,
+      "step": 14291
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.1918859649122808e-06,
+      "loss": 0.2717,
+      "step": 14292
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.1913377192982456e-06,
+      "loss": 0.3831,
+      "step": 14293
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.1907894736842104e-06,
+      "loss": 0.5758,
+      "step": 14294
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.1902412280701757e-06,
+      "loss": 0.4392,
+      "step": 14295
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.1896929824561405e-06,
+      "loss": 0.5031,
+      "step": 14296
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.1891447368421053e-06,
+      "loss": 0.8847,
+      "step": 14297
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.18859649122807e-06,
+      "loss": 0.2087,
+      "step": 14298
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.1880482456140354e-06,
+      "loss": 0.1272,
+      "step": 14299
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.1875000000000002e-06,
+      "loss": 0.1194,
+      "step": 14300
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.186951754385965e-06,
+      "loss": 0.8395,
+      "step": 14301
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.1864035087719303e-06,
+      "loss": 0.6871,
+      "step": 14302
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.185855263157895e-06,
+      "loss": 0.4506,
+      "step": 14303
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.18530701754386e-06,
+      "loss": 0.5589,
+      "step": 14304
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.1847587719298248e-06,
+      "loss": 1.1085,
+      "step": 14305
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.1842105263157896e-06,
+      "loss": 0.4657,
+      "step": 14306
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.1836622807017544e-06,
+      "loss": 0.4825,
+      "step": 14307
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.1831140350877193e-06,
+      "loss": 0.9251,
+      "step": 14308
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.1825657894736845e-06,
+      "loss": 0.5003,
+      "step": 14309
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.1820175438596493e-06,
+      "loss": 0.2052,
+      "step": 14310
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.181469298245614e-06,
+      "loss": 0.5197,
+      "step": 14311
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.180921052631579e-06,
+      "loss": 0.5722,
+      "step": 14312
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.180372807017544e-06,
+      "loss": 0.5921,
+      "step": 14313
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.179824561403509e-06,
+      "loss": 0.2683,
+      "step": 14314
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.179276315789474e-06,
+      "loss": 1.2342,
+      "step": 14315
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.1787280701754387e-06,
+      "loss": 0.2356,
+      "step": 14316
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.1781798245614035e-06,
+      "loss": 1.2799,
+      "step": 14317
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.1776315789473688e-06,
+      "loss": 0.7768,
+      "step": 14318
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.1770833333333336e-06,
+      "loss": 0.2124,
+      "step": 14319
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.1765350877192984e-06,
+      "loss": 0.3667,
+      "step": 14320
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.1759868421052632e-06,
+      "loss": 0.1523,
+      "step": 14321
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.1754385964912285e-06,
+      "loss": 0.7273,
+      "step": 14322
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.1748903508771933e-06,
+      "loss": 1.4903,
+      "step": 14323
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.174342105263158e-06,
+      "loss": 0.4345,
+      "step": 14324
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.173793859649123e-06,
+      "loss": 0.2523,
+      "step": 14325
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.173245614035088e-06,
+      "loss": 0.5128,
+      "step": 14326
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.172697368421053e-06,
+      "loss": 0.8141,
+      "step": 14327
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.172149122807018e-06,
+      "loss": 1.2699,
+      "step": 14328
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.1716008771929827e-06,
+      "loss": 0.5406,
+      "step": 14329
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.1710526315789475e-06,
+      "loss": 0.1231,
+      "step": 14330
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.1705043859649123e-06,
+      "loss": 0.4037,
+      "step": 14331
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.169956140350877e-06,
+      "loss": 0.1665,
+      "step": 14332
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.169407894736842e-06,
+      "loss": 0.3379,
+      "step": 14333
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.1688596491228072e-06,
+      "loss": 0.1116,
+      "step": 14334
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.168311403508772e-06,
+      "loss": 0.8213,
+      "step": 14335
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.167763157894737e-06,
+      "loss": 0.8169,
+      "step": 14336
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.1672149122807017e-06,
+      "loss": 0.8897,
+      "step": 14337
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.166666666666667e-06,
+      "loss": 0.6384,
+      "step": 14338
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.1661184210526318e-06,
+      "loss": 0.2186,
+      "step": 14339
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.1655701754385966e-06,
+      "loss": 0.9099,
+      "step": 14340
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.165021929824562e-06,
+      "loss": 0.9549,
+      "step": 14341
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.1644736842105267e-06,
+      "loss": 0.1085,
+      "step": 14342
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.1639254385964915e-06,
+      "loss": 0.2959,
+      "step": 14343
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.1633771929824563e-06,
+      "loss": 0.5697,
+      "step": 14344
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.162828947368421e-06,
+      "loss": 0.4808,
+      "step": 14345
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.1622807017543864e-06,
+      "loss": 0.6171,
+      "step": 14346
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.1617324561403512e-06,
+      "loss": 0.086,
+      "step": 14347
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.161184210526316e-06,
+      "loss": 0.3068,
+      "step": 14348
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.160635964912281e-06,
+      "loss": 0.2073,
+      "step": 14349
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.1600877192982457e-06,
+      "loss": 0.7089,
+      "step": 14350
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.1595394736842105e-06,
+      "loss": 0.827,
+      "step": 14351
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.1589912280701753e-06,
+      "loss": 0.5221,
+      "step": 14352
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.1584429824561406e-06,
+      "loss": 0.7487,
+      "step": 14353
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.1578947368421054e-06,
+      "loss": 0.5493,
+      "step": 14354
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.1573464912280702e-06,
+      "loss": 0.4578,
+      "step": 14355
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.156798245614035e-06,
+      "loss": 0.4976,
+      "step": 14356
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.1562500000000003e-06,
+      "loss": 0.3083,
+      "step": 14357
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.155701754385965e-06,
+      "loss": 0.1521,
+      "step": 14358
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.15515350877193e-06,
+      "loss": 0.2153,
+      "step": 14359
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.154605263157895e-06,
+      "loss": 1.3003,
+      "step": 14360
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.15405701754386e-06,
+      "loss": 0.6771,
+      "step": 14361
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.153508771929825e-06,
+      "loss": 1.1832,
+      "step": 14362
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.1529605263157897e-06,
+      "loss": 1.0524,
+      "step": 14363
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.1524122807017545e-06,
+      "loss": 1.0608,
+      "step": 14364
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.1518640350877198e-06,
+      "loss": 0.1881,
+      "step": 14365
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.1513157894736846e-06,
+      "loss": 0.998,
+      "step": 14366
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.1507675438596494e-06,
+      "loss": 1.0654,
+      "step": 14367
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.1502192982456142e-06,
+      "loss": 0.4197,
+      "step": 14368
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.149671052631579e-06,
+      "loss": 0.2828,
+      "step": 14369
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.149122807017544e-06,
+      "loss": 0.4436,
+      "step": 14370
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.1485745614035087e-06,
+      "loss": 0.6327,
+      "step": 14371
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.1480263157894735e-06,
+      "loss": 1.1667,
+      "step": 14372
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.1474780701754388e-06,
+      "loss": 0.5733,
+      "step": 14373
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.1469298245614036e-06,
+      "loss": 0.276,
+      "step": 14374
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.1463815789473684e-06,
+      "loss": 0.5658,
+      "step": 14375
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.1458333333333333e-06,
+      "loss": 0.2941,
+      "step": 14376
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.1452850877192985e-06,
+      "loss": 0.4475,
+      "step": 14377
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.1447368421052633e-06,
+      "loss": 0.1589,
+      "step": 14378
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.144188596491228e-06,
+      "loss": 0.2477,
+      "step": 14379
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.143640350877193e-06,
+      "loss": 0.7243,
+      "step": 14380
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.1430921052631582e-06,
+      "loss": 1.6568,
+      "step": 14381
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.142543859649123e-06,
+      "loss": 0.9216,
+      "step": 14382
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.141995614035088e-06,
+      "loss": 1.0501,
+      "step": 14383
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.141447368421053e-06,
+      "loss": 0.4955,
+      "step": 14384
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.140899122807018e-06,
+      "loss": 0.265,
+      "step": 14385
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.1403508771929828e-06,
+      "loss": 0.6362,
+      "step": 14386
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.1398026315789476e-06,
+      "loss": 2.1293,
+      "step": 14387
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.1392543859649124e-06,
+      "loss": 1.1407,
+      "step": 14388
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.1387061403508772e-06,
+      "loss": 1.8298,
+      "step": 14389
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.138157894736842e-06,
+      "loss": 0.3263,
+      "step": 14390
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.137609649122807e-06,
+      "loss": 0.2253,
+      "step": 14391
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.137061403508772e-06,
+      "loss": 1.1451,
+      "step": 14392
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.136513157894737e-06,
+      "loss": 1.2069,
+      "step": 14393
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.135964912280702e-06,
+      "loss": 0.757,
+      "step": 14394
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.1354166666666666e-06,
+      "loss": 0.2253,
+      "step": 14395
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.134868421052632e-06,
+      "loss": 1.2172,
+      "step": 14396
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.1343201754385967e-06,
+      "loss": 1.072,
+      "step": 14397
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.1337719298245615e-06,
+      "loss": 0.2558,
+      "step": 14398
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.1332236842105263e-06,
+      "loss": 0.504,
+      "step": 14399
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.1326754385964916e-06,
+      "loss": 0.331,
+      "step": 14400
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.1321271929824564e-06,
+      "loss": 0.231,
+      "step": 14401
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.1315789473684212e-06,
+      "loss": 0.7011,
+      "step": 14402
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.131030701754386e-06,
+      "loss": 0.5162,
+      "step": 14403
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.1304824561403513e-06,
+      "loss": 0.1981,
+      "step": 14404
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.129934210526316e-06,
+      "loss": 0.7032,
+      "step": 14405
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.129385964912281e-06,
+      "loss": 1.0073,
+      "step": 14406
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.1288377192982458e-06,
+      "loss": 0.305,
+      "step": 14407
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.1282894736842106e-06,
+      "loss": 0.3121,
+      "step": 14408
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.1277412280701754e-06,
+      "loss": 0.1804,
+      "step": 14409
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.1271929824561407e-06,
+      "loss": 0.7979,
+      "step": 14410
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.1266447368421055e-06,
+      "loss": 0.3491,
+      "step": 14411
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.1260964912280703e-06,
+      "loss": 0.8969,
+      "step": 14412
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.125548245614035e-06,
+      "loss": 0.6301,
+      "step": 14413
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.125e-06,
+      "loss": 0.9811,
+      "step": 14414
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.124451754385965e-06,
+      "loss": 0.4087,
+      "step": 14415
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.12390350877193e-06,
+      "loss": 0.8626,
+      "step": 14416
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.123355263157895e-06,
+      "loss": 0.4753,
+      "step": 14417
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.1228070175438597e-06,
+      "loss": 0.7767,
+      "step": 14418
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.1222587719298245e-06,
+      "loss": 0.5087,
+      "step": 14419
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.1217105263157898e-06,
+      "loss": 0.2524,
+      "step": 14420
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.1211622807017546e-06,
+      "loss": 0.7185,
+      "step": 14421
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.1206140350877194e-06,
+      "loss": 0.9133,
+      "step": 14422
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.1200657894736847e-06,
+      "loss": 0.5427,
+      "step": 14423
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.1195175438596495e-06,
+      "loss": 0.616,
+      "step": 14424
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.1189692982456143e-06,
+      "loss": 0.7084,
+      "step": 14425
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.118421052631579e-06,
+      "loss": 0.4812,
+      "step": 14426
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.117872807017544e-06,
+      "loss": 0.1047,
+      "step": 14427
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.117324561403509e-06,
+      "loss": 0.6549,
+      "step": 14428
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.116776315789474e-06,
+      "loss": 0.8444,
+      "step": 14429
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.116228070175439e-06,
+      "loss": 0.4913,
+      "step": 14430
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.1156798245614037e-06,
+      "loss": 0.2524,
+      "step": 14431
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.1151315789473685e-06,
+      "loss": 0.6122,
+      "step": 14432
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.1145833333333333e-06,
+      "loss": 0.1871,
+      "step": 14433
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.114035087719298e-06,
+      "loss": 0.085,
+      "step": 14434
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.1134868421052634e-06,
+      "loss": 0.1274,
+      "step": 14435
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.1129385964912282e-06,
+      "loss": 0.6606,
+      "step": 14436
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.112390350877193e-06,
+      "loss": 0.5238,
+      "step": 14437
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.111842105263158e-06,
+      "loss": 0.5664,
+      "step": 14438
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.111293859649123e-06,
+      "loss": 0.0556,
+      "step": 14439
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.110745614035088e-06,
+      "loss": 0.1912,
+      "step": 14440
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.110197368421053e-06,
+      "loss": 0.6598,
+      "step": 14441
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.1096491228070176e-06,
+      "loss": 0.6265,
+      "step": 14442
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.109100877192983e-06,
+      "loss": 1.0472,
+      "step": 14443
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.1085526315789477e-06,
+      "loss": 0.5067,
+      "step": 14444
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.1080043859649125e-06,
+      "loss": 0.6793,
+      "step": 14445
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.1074561403508773e-06,
+      "loss": 0.9556,
+      "step": 14446
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.1069078947368426e-06,
+      "loss": 0.6897,
+      "step": 14447
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.1063596491228074e-06,
+      "loss": 1.042,
+      "step": 14448
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.1058114035087722e-06,
+      "loss": 0.7389,
+      "step": 14449
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.105263157894737e-06,
+      "loss": 0.9527,
+      "step": 14450
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.104714912280702e-06,
+      "loss": 0.3807,
+      "step": 14451
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.1041666666666667e-06,
+      "loss": 1.164,
+      "step": 14452
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.1036184210526315e-06,
+      "loss": 0.4505,
+      "step": 14453
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.1030701754385964e-06,
+      "loss": 0.4505,
+      "step": 14454
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.1025219298245616e-06,
+      "loss": 0.8948,
+      "step": 14455
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.1019736842105264e-06,
+      "loss": 0.1631,
+      "step": 14456
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.1014254385964913e-06,
+      "loss": 0.5006,
+      "step": 14457
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.100877192982456e-06,
+      "loss": 1.3301,
+      "step": 14458
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.1003289473684213e-06,
+      "loss": 0.4571,
+      "step": 14459
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.099780701754386e-06,
+      "loss": 0.667,
+      "step": 14460
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.099232456140351e-06,
+      "loss": 0.5494,
+      "step": 14461
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.098684210526316e-06,
+      "loss": 0.1882,
+      "step": 14462
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.098135964912281e-06,
+      "loss": 0.327,
+      "step": 14463
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.097587719298246e-06,
+      "loss": 0.526,
+      "step": 14464
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.0970394736842107e-06,
+      "loss": 0.6045,
+      "step": 14465
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.096491228070176e-06,
+      "loss": 0.8167,
+      "step": 14466
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.0959429824561408e-06,
+      "loss": 0.8853,
+      "step": 14467
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.0953947368421056e-06,
+      "loss": 0.4247,
+      "step": 14468
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.0948464912280704e-06,
+      "loss": 0.1784,
+      "step": 14469
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.0942982456140352e-06,
+      "loss": 0.9769,
+      "step": 14470
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.09375e-06,
+      "loss": 0.4006,
+      "step": 14471
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.093201754385965e-06,
+      "loss": 0.4405,
+      "step": 14472
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.0926535087719297e-06,
+      "loss": 0.0626,
+      "step": 14473
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.092105263157895e-06,
+      "loss": 0.6911,
+      "step": 14474
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.09155701754386e-06,
+      "loss": 0.7616,
+      "step": 14475
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.0910087719298246e-06,
+      "loss": 0.1688,
+      "step": 14476
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.0904605263157894e-06,
+      "loss": 0.4836,
+      "step": 14477
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.0899122807017547e-06,
+      "loss": 0.841,
+      "step": 14478
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.0893640350877195e-06,
+      "loss": 0.6292,
+      "step": 14479
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.0888157894736843e-06,
+      "loss": 0.5443,
+      "step": 14480
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.088267543859649e-06,
+      "loss": 1.4008,
+      "step": 14481
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.0877192982456144e-06,
+      "loss": 0.661,
+      "step": 14482
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.0871710526315792e-06,
+      "loss": 0.2866,
+      "step": 14483
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.086622807017544e-06,
+      "loss": 0.5403,
+      "step": 14484
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.086074561403509e-06,
+      "loss": 0.3724,
+      "step": 14485
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.085526315789474e-06,
+      "loss": 0.3334,
+      "step": 14486
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.084978070175439e-06,
+      "loss": 0.1382,
+      "step": 14487
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.0844298245614038e-06,
+      "loss": 0.1827,
+      "step": 14488
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.0838815789473686e-06,
+      "loss": 0.8992,
+      "step": 14489
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.0833333333333334e-06,
+      "loss": 0.0559,
+      "step": 14490
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.0827850877192983e-06,
+      "loss": 0.3346,
+      "step": 14491
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.082236842105263e-06,
+      "loss": 0.487,
+      "step": 14492
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.0816885964912283e-06,
+      "loss": 1.0325,
+      "step": 14493
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.081140350877193e-06,
+      "loss": 1.499,
+      "step": 14494
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.080592105263158e-06,
+      "loss": 0.7512,
+      "step": 14495
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.080043859649123e-06,
+      "loss": 0.4407,
+      "step": 14496
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0794956140350876e-06,
+      "loss": 0.2751,
+      "step": 14497
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.078947368421053e-06,
+      "loss": 1.1387,
+      "step": 14498
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0783991228070177e-06,
+      "loss": 0.3531,
+      "step": 14499
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0778508771929825e-06,
+      "loss": 0.4054,
+      "step": 14500
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0773026315789474e-06,
+      "loss": 1.325,
+      "step": 14501
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0767543859649126e-06,
+      "loss": 0.3444,
+      "step": 14502
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0762061403508774e-06,
+      "loss": 0.3818,
+      "step": 14503
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0756578947368422e-06,
+      "loss": 1.035,
+      "step": 14504
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0751096491228075e-06,
+      "loss": 0.2626,
+      "step": 14505
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0745614035087723e-06,
+      "loss": 0.3935,
+      "step": 14506
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.074013157894737e-06,
+      "loss": 0.0832,
+      "step": 14507
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.073464912280702e-06,
+      "loss": 0.4611,
+      "step": 14508
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.072916666666667e-06,
+      "loss": 0.9547,
+      "step": 14509
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0723684210526316e-06,
+      "loss": 0.1878,
+      "step": 14510
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0718201754385964e-06,
+      "loss": 0.3688,
+      "step": 14511
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0712719298245617e-06,
+      "loss": 0.5761,
+      "step": 14512
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0707236842105265e-06,
+      "loss": 0.307,
+      "step": 14513
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.0701754385964913e-06,
+      "loss": 0.6157,
+      "step": 14514
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.069627192982456e-06,
+      "loss": 1.3562,
+      "step": 14515
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.069078947368421e-06,
+      "loss": 0.7717,
+      "step": 14516
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.0685307017543862e-06,
+      "loss": 0.7949,
+      "step": 14517
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.067982456140351e-06,
+      "loss": 0.2605,
+      "step": 14518
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.067434210526316e-06,
+      "loss": 0.6059,
+      "step": 14519
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.0668859649122807e-06,
+      "loss": 0.5229,
+      "step": 14520
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.066337719298246e-06,
+      "loss": 0.6155,
+      "step": 14521
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.0657894736842108e-06,
+      "loss": 0.492,
+      "step": 14522
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.0652412280701756e-06,
+      "loss": 1.0982,
+      "step": 14523
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.0646929824561404e-06,
+      "loss": 0.2036,
+      "step": 14524
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.0641447368421057e-06,
+      "loss": 0.5909,
+      "step": 14525
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.0635964912280705e-06,
+      "loss": 0.7253,
+      "step": 14526
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.0630482456140353e-06,
+      "loss": 0.485,
+      "step": 14527
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.0625e-06,
+      "loss": 0.236,
+      "step": 14528
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.061951754385965e-06,
+      "loss": 0.3728,
+      "step": 14529
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.06140350877193e-06,
+      "loss": 0.5312,
+      "step": 14530
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.060855263157895e-06,
+      "loss": 0.9256,
+      "step": 14531
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.06030701754386e-06,
+      "loss": 0.3653,
+      "step": 14532
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.0597587719298247e-06,
+      "loss": 0.5519,
+      "step": 14533
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.0592105263157895e-06,
+      "loss": 0.3615,
+      "step": 14534
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.0586622807017544e-06,
+      "loss": 0.4891,
+      "step": 14535
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.058114035087719e-06,
+      "loss": 1.003,
+      "step": 14536
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.0575657894736844e-06,
+      "loss": 0.0536,
+      "step": 14537
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.0570175438596493e-06,
+      "loss": 0.1062,
+      "step": 14538
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.056469298245614e-06,
+      "loss": 0.3644,
+      "step": 14539
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.055921052631579e-06,
+      "loss": 0.2996,
+      "step": 14540
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.055372807017544e-06,
+      "loss": 0.5502,
+      "step": 14541
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.054824561403509e-06,
+      "loss": 0.1214,
+      "step": 14542
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.054276315789474e-06,
+      "loss": 0.0661,
+      "step": 14543
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.0537280701754386e-06,
+      "loss": 0.2446,
+      "step": 14544
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.053179824561404e-06,
+      "loss": 0.126,
+      "step": 14545
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.0526315789473687e-06,
+      "loss": 0.4141,
+      "step": 14546
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.0520833333333335e-06,
+      "loss": 0.4599,
+      "step": 14547
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.0515350877192983e-06,
+      "loss": 0.9293,
+      "step": 14548
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.0509868421052636e-06,
+      "loss": 1.2566,
+      "step": 14549
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.0504385964912284e-06,
+      "loss": 0.8568,
+      "step": 14550
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.0498903508771932e-06,
+      "loss": 0.5582,
+      "step": 14551
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.049342105263158e-06,
+      "loss": 0.4311,
+      "step": 14552
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.048793859649123e-06,
+      "loss": 0.862,
+      "step": 14553
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.0482456140350877e-06,
+      "loss": 0.1366,
+      "step": 14554
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.0476973684210525e-06,
+      "loss": 0.178,
+      "step": 14555
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.047149122807018e-06,
+      "loss": 1.3309,
+      "step": 14556
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.0466008771929826e-06,
+      "loss": 0.4294,
+      "step": 14557
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.0460526315789474e-06,
+      "loss": 1.0764,
+      "step": 14558
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.0455043859649123e-06,
+      "loss": 0.3318,
+      "step": 14559
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.0449561403508775e-06,
+      "loss": 0.5164,
+      "step": 14560
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.0444078947368423e-06,
+      "loss": 0.3609,
+      "step": 14561
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.043859649122807e-06,
+      "loss": 0.1584,
+      "step": 14562
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.043311403508772e-06,
+      "loss": 1.1428,
+      "step": 14563
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.0427631578947372e-06,
+      "loss": 1.0607,
+      "step": 14564
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.042214912280702e-06,
+      "loss": 0.6905,
+      "step": 14565
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.041666666666667e-06,
+      "loss": 0.5887,
+      "step": 14566
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.0411184210526317e-06,
+      "loss": 0.5584,
+      "step": 14567
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.040570175438597e-06,
+      "loss": 0.8509,
+      "step": 14568
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.0400219298245618e-06,
+      "loss": 0.3302,
+      "step": 14569
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.0394736842105266e-06,
+      "loss": 0.2227,
+      "step": 14570
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.0389254385964914e-06,
+      "loss": 0.2617,
+      "step": 14571
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.0383771929824563e-06,
+      "loss": 0.5046,
+      "step": 14572
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.037828947368421e-06,
+      "loss": 0.3461,
+      "step": 14573
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.037280701754386e-06,
+      "loss": 0.7096,
+      "step": 14574
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.0367324561403507e-06,
+      "loss": 0.6509,
+      "step": 14575
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.036184210526316e-06,
+      "loss": 0.3748,
+      "step": 14576
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.035635964912281e-06,
+      "loss": 0.8532,
+      "step": 14577
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.0350877192982456e-06,
+      "loss": 0.6016,
+      "step": 14578
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.0345394736842105e-06,
+      "loss": 0.6264,
+      "step": 14579
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.0339912280701757e-06,
+      "loss": 0.5035,
+      "step": 14580
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.0334429824561405e-06,
+      "loss": 0.3374,
+      "step": 14581
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.0328947368421053e-06,
+      "loss": 0.2626,
+      "step": 14582
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.03234649122807e-06,
+      "loss": 0.1737,
+      "step": 14583
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.0317982456140354e-06,
+      "loss": 0.3979,
+      "step": 14584
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.0312500000000002e-06,
+      "loss": 0.6182,
+      "step": 14585
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.030701754385965e-06,
+      "loss": 0.6054,
+      "step": 14586
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.0301535087719303e-06,
+      "loss": 0.514,
+      "step": 14587
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.029605263157895e-06,
+      "loss": 0.5705,
+      "step": 14588
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.02905701754386e-06,
+      "loss": 0.5645,
+      "step": 14589
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.028508771929825e-06,
+      "loss": 0.3554,
+      "step": 14590
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.0279605263157896e-06,
+      "loss": 0.6839,
+      "step": 14591
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.0274122807017544e-06,
+      "loss": 0.4973,
+      "step": 14592
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.0268640350877193e-06,
+      "loss": 0.4221,
+      "step": 14593
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.026315789473684e-06,
+      "loss": 0.4333,
+      "step": 14594
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.0257675438596493e-06,
+      "loss": 0.9245,
+      "step": 14595
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.025219298245614e-06,
+      "loss": 0.1779,
+      "step": 14596
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.024671052631579e-06,
+      "loss": 0.6172,
+      "step": 14597
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.024122807017544e-06,
+      "loss": 0.2594,
+      "step": 14598
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.023574561403509e-06,
+      "loss": 0.2959,
+      "step": 14599
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.023026315789474e-06,
+      "loss": 1.1072,
+      "step": 14600
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.0224780701754387e-06,
+      "loss": 0.2065,
+      "step": 14601
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.0219298245614035e-06,
+      "loss": 0.5823,
+      "step": 14602
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.0213815789473688e-06,
+      "loss": 1.4617,
+      "step": 14603
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.0208333333333336e-06,
+      "loss": 0.1719,
+      "step": 14604
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.0202850877192984e-06,
+      "loss": 0.5466,
+      "step": 14605
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.0197368421052633e-06,
+      "loss": 0.4906,
+      "step": 14606
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.0191885964912285e-06,
+      "loss": 0.5342,
+      "step": 14607
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.0186403508771933e-06,
+      "loss": 0.9153,
+      "step": 14608
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.018092105263158e-06,
+      "loss": 0.0942,
+      "step": 14609
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.017543859649123e-06,
+      "loss": 0.0813,
+      "step": 14610
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.016995614035088e-06,
+      "loss": 0.4782,
+      "step": 14611
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.0164473684210526e-06,
+      "loss": 0.4563,
+      "step": 14612
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.0158991228070175e-06,
+      "loss": 0.2106,
+      "step": 14613
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.0153508771929827e-06,
+      "loss": 0.7232,
+      "step": 14614
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.0148026315789475e-06,
+      "loss": 1.1052,
+      "step": 14615
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.0142543859649124e-06,
+      "loss": 0.5046,
+      "step": 14616
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.013706140350877e-06,
+      "loss": 0.5098,
+      "step": 14617
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.013157894736842e-06,
+      "loss": 0.377,
+      "step": 14618
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.0126096491228072e-06,
+      "loss": 0.9313,
+      "step": 14619
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.012061403508772e-06,
+      "loss": 0.2213,
+      "step": 14620
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.011513157894737e-06,
+      "loss": 1.3985,
+      "step": 14621
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.0109649122807017e-06,
+      "loss": 0.2488,
+      "step": 14622
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.010416666666667e-06,
+      "loss": 0.3869,
+      "step": 14623
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.009868421052632e-06,
+      "loss": 0.2351,
+      "step": 14624
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.0093201754385966e-06,
+      "loss": 0.1658,
+      "step": 14625
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.008771929824562e-06,
+      "loss": 0.3808,
+      "step": 14626
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.0082236842105267e-06,
+      "loss": 0.3606,
+      "step": 14627
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.0076754385964915e-06,
+      "loss": 0.9378,
+      "step": 14628
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.0071271929824563e-06,
+      "loss": 0.5683,
+      "step": 14629
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.006578947368421e-06,
+      "loss": 0.3796,
+      "step": 14630
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.006030701754386e-06,
+      "loss": 0.5702,
+      "step": 14631
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.0054824561403512e-06,
+      "loss": 0.258,
+      "step": 14632
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.7851137994668854,
+      "eval_loss": 0.6621229648590088,
+      "eval_roc_auc": 0.9461420793448543,
+      "eval_runtime": 183.612,
+      "eval_samples_per_second": 26.561,
+      "eval_steps_per_second": 3.322,
+      "step": 14632
+    }
+  ],
+  "max_steps": 18290,
+  "num_train_epochs": 10,
+  "total_flos": 2.65606641927936e+18,
+  "trial_name": null,
+  "trial_params": null
+}