diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,124776 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 20791,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.205128205128205e-08,
+      "loss": 1.6584,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.41025641025641e-08,
+      "loss": 1.706,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.615384615384617e-08,
+      "loss": 1.823,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.282051282051282e-07,
+      "loss": 1.7393,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6025641025641025e-07,
+      "loss": 1.6933,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9230769230769234e-07,
+      "loss": 1.7273,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2435897435897438e-07,
+      "loss": 1.7532,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.564102564102564e-07,
+      "loss": 1.6549,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.884615384615385e-07,
+      "loss": 1.812,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.205128205128205e-07,
+      "loss": 1.6754,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.525641025641026e-07,
+      "loss": 1.6914,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.846153846153847e-07,
+      "loss": 1.7801,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.1666666666666667e-07,
+      "loss": 1.7258,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.4871794871794876e-07,
+      "loss": 1.7635,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.807692307692308e-07,
+      "loss": 1.6713,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.128205128205128e-07,
+      "loss": 1.594,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.448717948717949e-07,
+      "loss": 1.6441,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.76923076923077e-07,
+      "loss": 1.461,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.08974358974359e-07,
+      "loss": 1.6469,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.41025641025641e-07,
+      "loss": 1.6579,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.730769230769231e-07,
+      "loss": 1.8303,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.051282051282052e-07,
+      "loss": 1.6655,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.371794871794873e-07,
+      "loss": 1.7364,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.692307692307694e-07,
+      "loss": 1.5332,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.012820512820515e-07,
+      "loss": 1.6145,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.333333333333333e-07,
+      "loss": 1.6793,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.653846153846154e-07,
+      "loss": 1.7072,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.974358974358975e-07,
+      "loss": 1.6861,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.294871794871796e-07,
+      "loss": 1.6994,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.615384615384617e-07,
+      "loss": 1.7304,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.935897435897436e-07,
+      "loss": 1.8961,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0256410256410257e-06,
+      "loss": 1.6091,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0576923076923078e-06,
+      "loss": 1.6796,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0897435897435899e-06,
+      "loss": 2.2462,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.121794871794872e-06,
+      "loss": 1.6941,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.153846153846154e-06,
+      "loss": 1.701,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.185897435897436e-06,
+      "loss": 1.617,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.217948717948718e-06,
+      "loss": 1.5928,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.25e-06,
+      "loss": 1.7541,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.282051282051282e-06,
+      "loss": 1.6325,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3141025641025643e-06,
+      "loss": 1.6734,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3461538461538462e-06,
+      "loss": 1.5646,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3782051282051285e-06,
+      "loss": 1.4728,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4102564102564104e-06,
+      "loss": 1.6235,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4423076923076922e-06,
+      "loss": 1.5541,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4743589743589745e-06,
+      "loss": 1.5738,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5064102564102564e-06,
+      "loss": 1.6759,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5384615384615387e-06,
+      "loss": 1.8205,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5705128205128206e-06,
+      "loss": 1.6301,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.602564102564103e-06,
+      "loss": 1.5861,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6346153846153848e-06,
+      "loss": 1.5601,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6666666666666667e-06,
+      "loss": 1.7804,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.698717948717949e-06,
+      "loss": 1.5381,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7307692307692308e-06,
+      "loss": 1.7726,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7628205128205131e-06,
+      "loss": 1.5988,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.794871794871795e-06,
+      "loss": 1.582,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.826923076923077e-06,
+      "loss": 1.6159,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8589743589743592e-06,
+      "loss": 1.4615,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.891025641025641e-06,
+      "loss": 1.5354,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9230769230769234e-06,
+      "loss": 1.3358,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9551282051282055e-06,
+      "loss": 1.5888,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.987179487179487e-06,
+      "loss": 1.293,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0192307692307692e-06,
+      "loss": 1.5101,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0512820512820513e-06,
+      "loss": 1.5638,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0833333333333334e-06,
+      "loss": 1.4579,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1153846153846155e-06,
+      "loss": 1.4064,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1474358974358976e-06,
+      "loss": 1.6038,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1794871794871797e-06,
+      "loss": 1.6458,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.211538461538462e-06,
+      "loss": 1.5318,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.243589743589744e-06,
+      "loss": 1.5489,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.275641025641026e-06,
+      "loss": 1.6022,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.307692307692308e-06,
+      "loss": 1.5977,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3397435897435897e-06,
+      "loss": 1.3948,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.371794871794872e-06,
+      "loss": 1.5212,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.403846153846154e-06,
+      "loss": 1.7162,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.435897435897436e-06,
+      "loss": 1.5146,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.467948717948718e-06,
+      "loss": 1.5599,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5e-06,
+      "loss": 1.5377,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5320512820512823e-06,
+      "loss": 1.6287,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.564102564102564e-06,
+      "loss": 1.5159,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5961538461538465e-06,
+      "loss": 1.6546,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6282051282051286e-06,
+      "loss": 1.482,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6602564102564107e-06,
+      "loss": 1.5794,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6923076923076923e-06,
+      "loss": 1.5853,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7243589743589744e-06,
+      "loss": 1.5775,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.756410256410257e-06,
+      "loss": 1.4825,
+      "step": 86
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7884615384615386e-06,
+      "loss": 1.5698,
+      "step": 87
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8205128205128207e-06,
+      "loss": 1.3976,
+      "step": 88
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.852564102564103e-06,
+      "loss": 1.7751,
+      "step": 89
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8846153846153845e-06,
+      "loss": 1.3708,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.916666666666667e-06,
+      "loss": 1.5284,
+      "step": 91
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.948717948717949e-06,
+      "loss": 1.5011,
+      "step": 92
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.980769230769231e-06,
+      "loss": 1.3619,
+      "step": 93
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.012820512820513e-06,
+      "loss": 1.42,
+      "step": 94
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.044871794871795e-06,
+      "loss": 1.5261,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0769230769230774e-06,
+      "loss": 1.4911,
+      "step": 96
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.108974358974359e-06,
+      "loss": 1.6726,
+      "step": 97
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.141025641025641e-06,
+      "loss": 1.4643,
+      "step": 98
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1730769230769233e-06,
+      "loss": 1.559,
+      "step": 99
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.205128205128206e-06,
+      "loss": 1.385,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2371794871794875e-06,
+      "loss": 1.2516,
+      "step": 101
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2692307692307696e-06,
+      "loss": 1.3611,
+      "step": 102
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.3012820512820517e-06,
+      "loss": 1.4104,
+      "step": 103
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 1.4268,
+      "step": 104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.365384615384616e-06,
+      "loss": 1.4816,
+      "step": 105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.397435897435898e-06,
+      "loss": 1.3671,
+      "step": 106
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4294871794871796e-06,
+      "loss": 1.3378,
+      "step": 107
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4615384615384617e-06,
+      "loss": 1.4677,
+      "step": 108
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4935897435897438e-06,
+      "loss": 1.2952,
+      "step": 109
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5256410256410263e-06,
+      "loss": 1.4342,
+      "step": 110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.557692307692308e-06,
+      "loss": 1.3349,
+      "step": 111
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.58974358974359e-06,
+      "loss": 1.2501,
+      "step": 112
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.621794871794872e-06,
+      "loss": 1.3163,
+      "step": 113
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.653846153846154e-06,
+      "loss": 1.2969,
+      "step": 114
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6858974358974363e-06,
+      "loss": 1.4542,
+      "step": 115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7179487179487184e-06,
+      "loss": 1.3675,
+      "step": 116
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7500000000000005e-06,
+      "loss": 1.5467,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.782051282051282e-06,
+      "loss": 1.3634,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8141025641025643e-06,
+      "loss": 1.2769,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.846153846153847e-06,
+      "loss": 1.3401,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.878205128205129e-06,
+      "loss": 1.2055,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.910256410256411e-06,
+      "loss": 1.4746,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.942307692307692e-06,
+      "loss": 1.5381,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.974358974358974e-06,
+      "loss": 1.2854,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.006410256410257e-06,
+      "loss": 1.2458,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.0384615384615385e-06,
+      "loss": 1.4674,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.070512820512821e-06,
+      "loss": 1.5354,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.102564102564103e-06,
+      "loss": 1.4694,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.134615384615385e-06,
+      "loss": 1.4223,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.166666666666667e-06,
+      "loss": 1.4865,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.198717948717949e-06,
+      "loss": 1.2693,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.230769230769231e-06,
+      "loss": 1.3432,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.262820512820513e-06,
+      "loss": 1.3345,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.294871794871795e-06,
+      "loss": 1.2196,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.326923076923077e-06,
+      "loss": 1.3935,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.358974358974359e-06,
+      "loss": 1.2802,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.3910256410256415e-06,
+      "loss": 1.3162,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.423076923076924e-06,
+      "loss": 1.2875,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.455128205128206e-06,
+      "loss": 1.5919,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.487179487179488e-06,
+      "loss": 1.5315,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.51923076923077e-06,
+      "loss": 1.4249,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.551282051282052e-06,
+      "loss": 1.2483,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.583333333333333e-06,
+      "loss": 1.2794,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.615384615384616e-06,
+      "loss": 1.4766,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.647435897435898e-06,
+      "loss": 1.4553,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.6794871794871795e-06,
+      "loss": 1.3542,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.711538461538462e-06,
+      "loss": 1.3127,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.743589743589744e-06,
+      "loss": 1.2754,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.775641025641027e-06,
+      "loss": 1.4411,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.807692307692308e-06,
+      "loss": 1.2539,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.83974358974359e-06,
+      "loss": 1.2849,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.871794871794872e-06,
+      "loss": 1.4912,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.903846153846154e-06,
+      "loss": 1.4248,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.935897435897436e-06,
+      "loss": 1.5004,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.967948717948718e-06,
+      "loss": 1.3539,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5e-06,
+      "loss": 1.4909,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.0320512820512825e-06,
+      "loss": 1.2355,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.064102564102565e-06,
+      "loss": 1.4575,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.096153846153846e-06,
+      "loss": 1.3186,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.128205128205128e-06,
+      "loss": 1.4748,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.160256410256411e-06,
+      "loss": 1.3552,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.192307692307693e-06,
+      "loss": 1.3832,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.224358974358975e-06,
+      "loss": 1.3234,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.256410256410257e-06,
+      "loss": 1.3415,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.288461538461539e-06,
+      "loss": 1.2838,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.320512820512821e-06,
+      "loss": 1.4086,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.3525641025641026e-06,
+      "loss": 1.473,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.384615384615385e-06,
+      "loss": 1.3032,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.416666666666667e-06,
+      "loss": 1.326,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.448717948717949e-06,
+      "loss": 1.494,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.480769230769232e-06,
+      "loss": 1.4156,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.512820512820514e-06,
+      "loss": 1.4082,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.544871794871796e-06,
+      "loss": 1.4216,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.576923076923077e-06,
+      "loss": 1.311,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.608974358974359e-06,
+      "loss": 1.3674,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.641025641025641e-06,
+      "loss": 1.3354,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.6730769230769235e-06,
+      "loss": 1.3048,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.705128205128206e-06,
+      "loss": 1.3584,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.737179487179487e-06,
+      "loss": 1.3651,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.769230769230769e-06,
+      "loss": 1.1951,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.801282051282052e-06,
+      "loss": 1.5344,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.833333333333334e-06,
+      "loss": 1.3922,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.865384615384616e-06,
+      "loss": 1.3601,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.897435897435898e-06,
+      "loss": 1.3855,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.92948717948718e-06,
+      "loss": 1.2743,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.961538461538462e-06,
+      "loss": 1.3451,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.9935897435897436e-06,
+      "loss": 1.4253,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.025641025641026e-06,
+      "loss": 1.3658,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.057692307692308e-06,
+      "loss": 1.4117,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.08974358974359e-06,
+      "loss": 1.339,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.121794871794873e-06,
+      "loss": 1.3834,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.153846153846155e-06,
+      "loss": 1.4473,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.185897435897437e-06,
+      "loss": 1.4033,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.217948717948718e-06,
+      "loss": 1.3313,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.25e-06,
+      "loss": 1.2696,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.282051282051282e-06,
+      "loss": 1.3314,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.3141025641025645e-06,
+      "loss": 1.2754,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.3461538461538466e-06,
+      "loss": 1.3617,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.378205128205129e-06,
+      "loss": 1.3472,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.410256410256412e-06,
+      "loss": 1.4046,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.442307692307693e-06,
+      "loss": 1.4866,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.474358974358975e-06,
+      "loss": 1.5231,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.506410256410257e-06,
+      "loss": 1.4017,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.538461538461539e-06,
+      "loss": 1.4336,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.570512820512821e-06,
+      "loss": 1.2645,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.602564102564103e-06,
+      "loss": 1.3518,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.6346153846153846e-06,
+      "loss": 1.4492,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 1.1839,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.698717948717949e-06,
+      "loss": 1.3047,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.730769230769232e-06,
+      "loss": 1.4281,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.762820512820514e-06,
+      "loss": 1.3856,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.794871794871796e-06,
+      "loss": 1.3144,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.826923076923078e-06,
+      "loss": 1.4261,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.858974358974359e-06,
+      "loss": 1.3892,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.891025641025641e-06,
+      "loss": 1.3787,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.923076923076923e-06,
+      "loss": 1.3381,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.9551282051282055e-06,
+      "loss": 1.3548,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.9871794871794876e-06,
+      "loss": 1.2841,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.01923076923077e-06,
+      "loss": 1.1562,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.051282051282053e-06,
+      "loss": 1.3046,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.083333333333335e-06,
+      "loss": 1.2909,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.115384615384616e-06,
+      "loss": 1.345,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.147435897435898e-06,
+      "loss": 1.3375,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.17948717948718e-06,
+      "loss": 1.2367,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.211538461538462e-06,
+      "loss": 1.2614,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.243589743589744e-06,
+      "loss": 1.2449,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.2756410256410255e-06,
+      "loss": 1.1949,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.307692307692308e-06,
+      "loss": 1.1471,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.33974358974359e-06,
+      "loss": 1.3201,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.371794871794873e-06,
+      "loss": 1.2573,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.403846153846155e-06,
+      "loss": 1.3026,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.435897435897437e-06,
+      "loss": 1.2952,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.467948717948719e-06,
+      "loss": 1.4166,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 1.4248,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.532051282051282e-06,
+      "loss": 1.3533,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.564102564102564e-06,
+      "loss": 1.1891,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.5961538461538465e-06,
+      "loss": 1.3287,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.6282051282051286e-06,
+      "loss": 1.3064,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.660256410256411e-06,
+      "loss": 1.232,
+      "step": 239
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.692307692307694e-06,
+      "loss": 1.3701,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.724358974358976e-06,
+      "loss": 1.4195,
+      "step": 241
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.756410256410258e-06,
+      "loss": 1.349,
+      "step": 242
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.78846153846154e-06,
+      "loss": 1.3964,
+      "step": 243
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.820512820512822e-06,
+      "loss": 1.3895,
+      "step": 244
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.852564102564102e-06,
+      "loss": 1.1863,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.884615384615384e-06,
+      "loss": 1.3848,
+      "step": 246
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.916666666666667e-06,
+      "loss": 1.4336,
+      "step": 247
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.948717948717949e-06,
+      "loss": 1.1889,
+      "step": 248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.980769230769232e-06,
+      "loss": 1.4881,
+      "step": 249
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.012820512820515e-06,
+      "loss": 1.1872,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.044871794871797e-06,
+      "loss": 1.2107,
+      "step": 251
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.076923076923077e-06,
+      "loss": 1.1805,
+      "step": 252
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.108974358974359e-06,
+      "loss": 1.2622,
+      "step": 253
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.141025641025641e-06,
+      "loss": 1.4153,
+      "step": 254
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.173076923076923e-06,
+      "loss": 1.3399,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.205128205128205e-06,
+      "loss": 1.0389,
+      "step": 256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.237179487179487e-06,
+      "loss": 1.3714,
+      "step": 257
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.26923076923077e-06,
+      "loss": 1.4103,
+      "step": 258
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.301282051282052e-06,
+      "loss": 1.2195,
+      "step": 259
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 1.2148,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.365384615384616e-06,
+      "loss": 1.1673,
+      "step": 261
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.397435897435898e-06,
+      "loss": 1.4305,
+      "step": 262
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.42948717948718e-06,
+      "loss": 1.361,
+      "step": 263
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.461538461538462e-06,
+      "loss": 1.236,
+      "step": 264
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.493589743589744e-06,
+      "loss": 1.3639,
+      "step": 265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.525641025641026e-06,
+      "loss": 1.365,
+      "step": 266
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.557692307692308e-06,
+      "loss": 1.2978,
+      "step": 267
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.58974358974359e-06,
+      "loss": 1.3621,
+      "step": 268
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.621794871794873e-06,
+      "loss": 1.1381,
+      "step": 269
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.653846153846155e-06,
+      "loss": 1.3441,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.685897435897437e-06,
+      "loss": 1.2231,
+      "step": 271
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.717948717948719e-06,
+      "loss": 1.4344,
+      "step": 272
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.750000000000001e-06,
+      "loss": 1.4075,
+      "step": 273
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.782051282051283e-06,
+      "loss": 1.322,
+      "step": 274
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.814102564102565e-06,
+      "loss": 1.4265,
+      "step": 275
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.846153846153847e-06,
+      "loss": 1.1639,
+      "step": 276
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.87820512820513e-06,
+      "loss": 1.2331,
+      "step": 277
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.910256410256411e-06,
+      "loss": 1.3329,
+      "step": 278
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.942307692307693e-06,
+      "loss": 1.3291,
+      "step": 279
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.974358974358976e-06,
+      "loss": 1.1202,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.006410256410258e-06,
+      "loss": 1.3317,
+      "step": 281
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.03846153846154e-06,
+      "loss": 1.3217,
+      "step": 282
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.070512820512822e-06,
+      "loss": 1.3008,
+      "step": 283
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.102564102564104e-06,
+      "loss": 1.2121,
+      "step": 284
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.134615384615384e-06,
+      "loss": 1.2969,
+      "step": 285
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.166666666666666e-06,
+      "loss": 1.2281,
+      "step": 286
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.198717948717949e-06,
+      "loss": 1.1955,
+      "step": 287
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.230769230769232e-06,
+      "loss": 1.3449,
+      "step": 288
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.262820512820514e-06,
+      "loss": 1.0236,
+      "step": 289
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.294871794871796e-06,
+      "loss": 1.3134,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.326923076923079e-06,
+      "loss": 1.3719,
+      "step": 291
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.358974358974359e-06,
+      "loss": 1.1758,
+      "step": 292
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.391025641025641e-06,
+      "loss": 1.0794,
+      "step": 293
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.423076923076923e-06,
+      "loss": 1.3789,
+      "step": 294
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.455128205128205e-06,
+      "loss": 1.2632,
+      "step": 295
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.487179487179487e-06,
+      "loss": 1.353,
+      "step": 296
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.51923076923077e-06,
+      "loss": 1.3279,
+      "step": 297
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.551282051282053e-06,
+      "loss": 1.3552,
+      "step": 298
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.583333333333335e-06,
+      "loss": 1.2412,
+      "step": 299
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.615384615384616e-06,
+      "loss": 1.2773,
+      "step": 300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.647435897435898e-06,
+      "loss": 1.2426,
+      "step": 301
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.67948717948718e-06,
+      "loss": 1.1011,
+      "step": 302
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.711538461538462e-06,
+      "loss": 1.2779,
+      "step": 303
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.743589743589744e-06,
+      "loss": 1.1818,
+      "step": 304
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.775641025641026e-06,
+      "loss": 1.2506,
+      "step": 305
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.807692307692308e-06,
+      "loss": 1.2458,
+      "step": 306
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.83974358974359e-06,
+      "loss": 1.2848,
+      "step": 307
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.871794871794872e-06,
+      "loss": 1.2394,
+      "step": 308
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.903846153846155e-06,
+      "loss": 1.2876,
+      "step": 309
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.935897435897437e-06,
+      "loss": 1.2625,
+      "step": 310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.967948717948719e-06,
+      "loss": 1.2673,
+      "step": 311
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1e-05,
+      "loss": 1.1818,
+      "step": 312
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0032051282051283e-05,
+      "loss": 1.241,
+      "step": 313
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0064102564102565e-05,
+      "loss": 1.2307,
+      "step": 314
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0096153846153847e-05,
+      "loss": 1.4191,
+      "step": 315
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.012820512820513e-05,
+      "loss": 1.2246,
+      "step": 316
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0160256410256411e-05,
+      "loss": 1.2571,
+      "step": 317
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0192307692307692e-05,
+      "loss": 1.2937,
+      "step": 318
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0224358974358974e-05,
+      "loss": 1.1545,
+      "step": 319
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0256410256410256e-05,
+      "loss": 1.3447,
+      "step": 320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0288461538461538e-05,
+      "loss": 1.1932,
+      "step": 321
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0320512820512822e-05,
+      "loss": 1.2214,
+      "step": 322
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0352564102564104e-05,
+      "loss": 1.3756,
+      "step": 323
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0384615384615386e-05,
+      "loss": 1.3189,
+      "step": 324
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0416666666666668e-05,
+      "loss": 1.2917,
+      "step": 325
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.044871794871795e-05,
+      "loss": 1.3198,
+      "step": 326
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0480769230769232e-05,
+      "loss": 1.2604,
+      "step": 327
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0512820512820514e-05,
+      "loss": 1.2988,
+      "step": 328
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0544871794871796e-05,
+      "loss": 1.2524,
+      "step": 329
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0576923076923078e-05,
+      "loss": 1.1864,
+      "step": 330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.060897435897436e-05,
+      "loss": 1.2258,
+      "step": 331
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0641025641025643e-05,
+      "loss": 1.3164,
+      "step": 332
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0673076923076923e-05,
+      "loss": 1.4288,
+      "step": 333
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0705128205128205e-05,
+      "loss": 1.1924,
+      "step": 334
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0737179487179487e-05,
+      "loss": 1.1762,
+      "step": 335
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.076923076923077e-05,
+      "loss": 1.2947,
+      "step": 336
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0801282051282051e-05,
+      "loss": 1.1296,
+      "step": 337
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0833333333333334e-05,
+      "loss": 1.2576,
+      "step": 338
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0865384615384616e-05,
+      "loss": 1.3062,
+      "step": 339
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0897435897435898e-05,
+      "loss": 1.354,
+      "step": 340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.092948717948718e-05,
+      "loss": 1.3044,
+      "step": 341
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0961538461538464e-05,
+      "loss": 1.272,
+      "step": 342
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0993589743589746e-05,
+      "loss": 1.1945,
+      "step": 343
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1025641025641028e-05,
+      "loss": 1.4353,
+      "step": 344
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.105769230769231e-05,
+      "loss": 1.261,
+      "step": 345
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1089743589743592e-05,
+      "loss": 1.2129,
+      "step": 346
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1121794871794872e-05,
+      "loss": 1.2461,
+      "step": 347
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1153846153846154e-05,
+      "loss": 1.3354,
+      "step": 348
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1185897435897437e-05,
+      "loss": 1.1986,
+      "step": 349
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1217948717948719e-05,
+      "loss": 1.1899,
+      "step": 350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.125e-05,
+      "loss": 1.0971,
+      "step": 351
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1282051282051283e-05,
+      "loss": 1.3751,
+      "step": 352
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1314102564102565e-05,
+      "loss": 1.3792,
+      "step": 353
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1346153846153847e-05,
+      "loss": 1.2312,
+      "step": 354
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1378205128205129e-05,
+      "loss": 1.2322,
+      "step": 355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1410256410256411e-05,
+      "loss": 1.3026,
+      "step": 356
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1442307692307693e-05,
+      "loss": 1.2985,
+      "step": 357
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1474358974358974e-05,
+      "loss": 1.1445,
+      "step": 358
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1506410256410256e-05,
+      "loss": 1.3959,
+      "step": 359
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1538461538461538e-05,
+      "loss": 1.2846,
+      "step": 360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1570512820512823e-05,
+      "loss": 1.3292,
+      "step": 361
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1602564102564104e-05,
+      "loss": 1.2235,
+      "step": 362
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1634615384615386e-05,
+      "loss": 1.0384,
+      "step": 363
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1666666666666668e-05,
+      "loss": 1.1627,
+      "step": 364
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.169871794871795e-05,
+      "loss": 1.3398,
+      "step": 365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1730769230769232e-05,
+      "loss": 1.3008,
+      "step": 366
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1762820512820514e-05,
+      "loss": 1.299,
+      "step": 367
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1794871794871796e-05,
+      "loss": 1.2413,
+      "step": 368
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1826923076923078e-05,
+      "loss": 1.2571,
+      "step": 369
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.185897435897436e-05,
+      "loss": 1.2894,
+      "step": 370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1891025641025643e-05,
+      "loss": 1.2482,
+      "step": 371
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1923076923076925e-05,
+      "loss": 1.324,
+      "step": 372
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1955128205128205e-05,
+      "loss": 1.3323,
+      "step": 373
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1987179487179487e-05,
+      "loss": 1.3034,
+      "step": 374
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.201923076923077e-05,
+      "loss": 1.2164,
+      "step": 375
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2051282051282051e-05,
+      "loss": 1.1814,
+      "step": 376
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2083333333333333e-05,
+      "loss": 1.3278,
+      "step": 377
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2115384615384615e-05,
+      "loss": 1.3259,
+      "step": 378
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2147435897435898e-05,
+      "loss": 1.1674,
+      "step": 379
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.217948717948718e-05,
+      "loss": 1.2601,
+      "step": 380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2211538461538463e-05,
+      "loss": 1.3037,
+      "step": 381
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2243589743589746e-05,
+      "loss": 1.2427,
+      "step": 382
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2275641025641028e-05,
+      "loss": 0.9605,
+      "step": 383
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.230769230769231e-05,
+      "loss": 1.1583,
+      "step": 384
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2339743589743592e-05,
+      "loss": 1.1007,
+      "step": 385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2371794871794874e-05,
+      "loss": 1.3579,
+      "step": 386
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2403846153846156e-05,
+      "loss": 1.2702,
+      "step": 387
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2435897435897436e-05,
+      "loss": 1.2808,
+      "step": 388
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2467948717948719e-05,
+      "loss": 1.1027,
+      "step": 389
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.25e-05,
+      "loss": 0.9668,
+      "step": 390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2532051282051283e-05,
+      "loss": 1.2102,
+      "step": 391
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2564102564102565e-05,
+      "loss": 1.1454,
+      "step": 392
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2596153846153847e-05,
+      "loss": 1.2059,
+      "step": 393
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2628205128205129e-05,
+      "loss": 1.3521,
+      "step": 394
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2660256410256411e-05,
+      "loss": 1.2072,
+      "step": 395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2692307692307693e-05,
+      "loss": 1.3366,
+      "step": 396
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2724358974358975e-05,
+      "loss": 1.3734,
+      "step": 397
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2756410256410257e-05,
+      "loss": 1.2696,
+      "step": 398
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2788461538461538e-05,
+      "loss": 1.3536,
+      "step": 399
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2820512820512823e-05,
+      "loss": 1.3465,
+      "step": 400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2852564102564105e-05,
+      "loss": 1.1804,
+      "step": 401
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2884615384615386e-05,
+      "loss": 1.1818,
+      "step": 402
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2916666666666668e-05,
+      "loss": 1.2518,
+      "step": 403
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.294871794871795e-05,
+      "loss": 1.2175,
+      "step": 404
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2980769230769232e-05,
+      "loss": 1.1672,
+      "step": 405
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3012820512820514e-05,
+      "loss": 1.1212,
+      "step": 406
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3044871794871796e-05,
+      "loss": 1.2239,
+      "step": 407
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3076923076923078e-05,
+      "loss": 1.2378,
+      "step": 408
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.310897435897436e-05,
+      "loss": 1.1085,
+      "step": 409
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3141025641025642e-05,
+      "loss": 1.3236,
+      "step": 410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3173076923076925e-05,
+      "loss": 1.2214,
+      "step": 411
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3205128205128207e-05,
+      "loss": 1.3217,
+      "step": 412
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3237179487179487e-05,
+      "loss": 0.9927,
+      "step": 413
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3269230769230769e-05,
+      "loss": 1.2723,
+      "step": 414
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3301282051282051e-05,
+      "loss": 1.1643,
+      "step": 415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 1.2596,
+      "step": 416
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3365384615384615e-05,
+      "loss": 1.2215,
+      "step": 417
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3397435897435897e-05,
+      "loss": 1.3239,
+      "step": 418
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.342948717948718e-05,
+      "loss": 1.2651,
+      "step": 419
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3461538461538463e-05,
+      "loss": 1.2979,
+      "step": 420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3493589743589745e-05,
+      "loss": 1.1182,
+      "step": 421
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3525641025641028e-05,
+      "loss": 1.3195,
+      "step": 422
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.355769230769231e-05,
+      "loss": 1.2061,
+      "step": 423
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3589743589743592e-05,
+      "loss": 1.2517,
+      "step": 424
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3621794871794874e-05,
+      "loss": 1.1838,
+      "step": 425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3653846153846156e-05,
+      "loss": 1.3951,
+      "step": 426
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3685897435897438e-05,
+      "loss": 1.3318,
+      "step": 427
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3717948717948718e-05,
+      "loss": 1.2975,
+      "step": 428
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.375e-05,
+      "loss": 1.2267,
+      "step": 429
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3782051282051283e-05,
+      "loss": 1.2181,
+      "step": 430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3814102564102565e-05,
+      "loss": 1.3014,
+      "step": 431
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3846153846153847e-05,
+      "loss": 1.2851,
+      "step": 432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3878205128205129e-05,
+      "loss": 1.1369,
+      "step": 433
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3910256410256411e-05,
+      "loss": 1.1196,
+      "step": 434
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3942307692307693e-05,
+      "loss": 1.1982,
+      "step": 435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3974358974358975e-05,
+      "loss": 1.2746,
+      "step": 436
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4006410256410257e-05,
+      "loss": 1.0079,
+      "step": 437
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.403846153846154e-05,
+      "loss": 1.2298,
+      "step": 438
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4070512820512823e-05,
+      "loss": 1.2987,
+      "step": 439
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4102564102564105e-05,
+      "loss": 1.3762,
+      "step": 440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4134615384615387e-05,
+      "loss": 1.1943,
+      "step": 441
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.416666666666667e-05,
+      "loss": 1.3154,
+      "step": 442
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.419871794871795e-05,
+      "loss": 1.2582,
+      "step": 443
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4230769230769232e-05,
+      "loss": 1.2877,
+      "step": 444
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4262820512820514e-05,
+      "loss": 1.2526,
+      "step": 445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4294871794871796e-05,
+      "loss": 1.1194,
+      "step": 446
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4326923076923078e-05,
+      "loss": 1.3634,
+      "step": 447
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.435897435897436e-05,
+      "loss": 1.2926,
+      "step": 448
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4391025641025642e-05,
+      "loss": 1.2797,
+      "step": 449
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4423076923076924e-05,
+      "loss": 1.2701,
+      "step": 450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4455128205128207e-05,
+      "loss": 1.0918,
+      "step": 451
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4487179487179489e-05,
+      "loss": 1.2686,
+      "step": 452
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.451923076923077e-05,
+      "loss": 1.0712,
+      "step": 453
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4551282051282051e-05,
+      "loss": 1.2371,
+      "step": 454
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4583333333333333e-05,
+      "loss": 1.1772,
+      "step": 455
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4615384615384615e-05,
+      "loss": 1.2531,
+      "step": 456
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4647435897435897e-05,
+      "loss": 1.308,
+      "step": 457
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.467948717948718e-05,
+      "loss": 1.2754,
+      "step": 458
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4711538461538463e-05,
+      "loss": 1.3516,
+      "step": 459
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4743589743589745e-05,
+      "loss": 1.3412,
+      "step": 460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4775641025641027e-05,
+      "loss": 1.2342,
+      "step": 461
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.480769230769231e-05,
+      "loss": 1.1629,
+      "step": 462
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4839743589743592e-05,
+      "loss": 1.3171,
+      "step": 463
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4871794871794874e-05,
+      "loss": 1.2646,
+      "step": 464
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4903846153846156e-05,
+      "loss": 1.3645,
+      "step": 465
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4935897435897438e-05,
+      "loss": 1.2568,
+      "step": 466
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.496794871794872e-05,
+      "loss": 1.1758,
+      "step": 467
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 1.2994,
+      "step": 468
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5032051282051282e-05,
+      "loss": 1.1975,
+      "step": 469
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5064102564102565e-05,
+      "loss": 1.2636,
+      "step": 470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5096153846153847e-05,
+      "loss": 1.3556,
+      "step": 471
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5128205128205129e-05,
+      "loss": 1.3465,
+      "step": 472
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.516025641025641e-05,
+      "loss": 1.2655,
+      "step": 473
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5192307692307693e-05,
+      "loss": 1.071,
+      "step": 474
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5224358974358975e-05,
+      "loss": 1.3133,
+      "step": 475
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5256410256410257e-05,
+      "loss": 1.3131,
+      "step": 476
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.528846153846154e-05,
+      "loss": 1.2391,
+      "step": 477
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5320512820512823e-05,
+      "loss": 1.2678,
+      "step": 478
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5352564102564103e-05,
+      "loss": 1.277,
+      "step": 479
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5384615384615387e-05,
+      "loss": 1.3633,
+      "step": 480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5416666666666668e-05,
+      "loss": 1.3478,
+      "step": 481
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.544871794871795e-05,
+      "loss": 1.257,
+      "step": 482
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5480769230769232e-05,
+      "loss": 1.2977,
+      "step": 483
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5512820512820516e-05,
+      "loss": 1.3929,
+      "step": 484
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5544871794871796e-05,
+      "loss": 1.2154,
+      "step": 485
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.557692307692308e-05,
+      "loss": 1.2018,
+      "step": 486
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.560897435897436e-05,
+      "loss": 1.241,
+      "step": 487
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5641025641025644e-05,
+      "loss": 1.2088,
+      "step": 488
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5673076923076924e-05,
+      "loss": 1.1672,
+      "step": 489
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5705128205128205e-05,
+      "loss": 1.1922,
+      "step": 490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.573717948717949e-05,
+      "loss": 1.2804,
+      "step": 491
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.576923076923077e-05,
+      "loss": 1.2799,
+      "step": 492
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5801282051282053e-05,
+      "loss": 1.2495,
+      "step": 493
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5833333333333333e-05,
+      "loss": 1.2433,
+      "step": 494
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5865384615384617e-05,
+      "loss": 1.2425,
+      "step": 495
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5897435897435897e-05,
+      "loss": 1.2072,
+      "step": 496
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.592948717948718e-05,
+      "loss": 1.1935,
+      "step": 497
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5961538461538465e-05,
+      "loss": 1.2351,
+      "step": 498
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5993589743589745e-05,
+      "loss": 1.0223,
+      "step": 499
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.602564102564103e-05,
+      "loss": 1.3472,
+      "step": 500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.605769230769231e-05,
+      "loss": 1.2993,
+      "step": 501
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6089743589743593e-05,
+      "loss": 1.3237,
+      "step": 502
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6121794871794874e-05,
+      "loss": 1.3379,
+      "step": 503
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6153846153846154e-05,
+      "loss": 1.2652,
+      "step": 504
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6185897435897438e-05,
+      "loss": 1.2648,
+      "step": 505
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6217948717948718e-05,
+      "loss": 1.3044,
+      "step": 506
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6250000000000002e-05,
+      "loss": 1.3896,
+      "step": 507
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6282051282051282e-05,
+      "loss": 1.2779,
+      "step": 508
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6314102564102566e-05,
+      "loss": 1.3504,
+      "step": 509
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6346153846153847e-05,
+      "loss": 1.1524,
+      "step": 510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.637820512820513e-05,
+      "loss": 1.2961,
+      "step": 511
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.641025641025641e-05,
+      "loss": 1.2047,
+      "step": 512
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6442307692307695e-05,
+      "loss": 1.2471,
+      "step": 513
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6474358974358975e-05,
+      "loss": 1.2093,
+      "step": 514
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6506410256410255e-05,
+      "loss": 1.196,
+      "step": 515
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.653846153846154e-05,
+      "loss": 1.2589,
+      "step": 516
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6570512820512823e-05,
+      "loss": 1.2546,
+      "step": 517
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6602564102564103e-05,
+      "loss": 1.3353,
+      "step": 518
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6634615384615387e-05,
+      "loss": 1.277,
+      "step": 519
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6666666666666667e-05,
+      "loss": 1.3438,
+      "step": 520
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.669871794871795e-05,
+      "loss": 1.2031,
+      "step": 521
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.673076923076923e-05,
+      "loss": 1.2086,
+      "step": 522
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6762820512820515e-05,
+      "loss": 1.1698,
+      "step": 523
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6794871794871796e-05,
+      "loss": 1.1413,
+      "step": 524
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.682692307692308e-05,
+      "loss": 1.1253,
+      "step": 525
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.685897435897436e-05,
+      "loss": 1.1394,
+      "step": 526
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6891025641025644e-05,
+      "loss": 1.3933,
+      "step": 527
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6923076923076924e-05,
+      "loss": 1.2707,
+      "step": 528
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6955128205128205e-05,
+      "loss": 1.2372,
+      "step": 529
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.698717948717949e-05,
+      "loss": 1.2319,
+      "step": 530
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.701923076923077e-05,
+      "loss": 1.3081,
+      "step": 531
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7051282051282053e-05,
+      "loss": 1.2748,
+      "step": 532
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7083333333333333e-05,
+      "loss": 1.2318,
+      "step": 533
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7115384615384617e-05,
+      "loss": 1.0839,
+      "step": 534
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7147435897435897e-05,
+      "loss": 1.1139,
+      "step": 535
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.717948717948718e-05,
+      "loss": 1.1557,
+      "step": 536
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7211538461538465e-05,
+      "loss": 1.2063,
+      "step": 537
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7243589743589745e-05,
+      "loss": 1.2088,
+      "step": 538
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.727564102564103e-05,
+      "loss": 1.2577,
+      "step": 539
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.730769230769231e-05,
+      "loss": 1.1256,
+      "step": 540
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7339743589743593e-05,
+      "loss": 1.299,
+      "step": 541
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7371794871794873e-05,
+      "loss": 1.3747,
+      "step": 542
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7403846153846157e-05,
+      "loss": 1.3573,
+      "step": 543
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7435897435897438e-05,
+      "loss": 1.1525,
+      "step": 544
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7467948717948718e-05,
+      "loss": 1.1919,
+      "step": 545
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7500000000000002e-05,
+      "loss": 1.1634,
+      "step": 546
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7532051282051282e-05,
+      "loss": 1.3564,
+      "step": 547
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7564102564102566e-05,
+      "loss": 1.3202,
+      "step": 548
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7596153846153846e-05,
+      "loss": 1.1916,
+      "step": 549
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.762820512820513e-05,
+      "loss": 1.343,
+      "step": 550
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.766025641025641e-05,
+      "loss": 1.2938,
+      "step": 551
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7692307692307694e-05,
+      "loss": 1.0608,
+      "step": 552
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7724358974358975e-05,
+      "loss": 1.1655,
+      "step": 553
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.775641025641026e-05,
+      "loss": 1.2242,
+      "step": 554
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.778846153846154e-05,
+      "loss": 1.4397,
+      "step": 555
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7820512820512823e-05,
+      "loss": 1.2841,
+      "step": 556
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7852564102564107e-05,
+      "loss": 1.2839,
+      "step": 557
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7884615384615387e-05,
+      "loss": 1.1583,
+      "step": 558
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7916666666666667e-05,
+      "loss": 1.3296,
+      "step": 559
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.794871794871795e-05,
+      "loss": 1.1572,
+      "step": 560
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.798076923076923e-05,
+      "loss": 1.2161,
+      "step": 561
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8012820512820515e-05,
+      "loss": 1.2533,
+      "step": 562
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8044871794871796e-05,
+      "loss": 1.1311,
+      "step": 563
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.807692307692308e-05,
+      "loss": 1.2228,
+      "step": 564
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.810897435897436e-05,
+      "loss": 1.1746,
+      "step": 565
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8141025641025644e-05,
+      "loss": 1.3532,
+      "step": 566
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8173076923076924e-05,
+      "loss": 1.3908,
+      "step": 567
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8205128205128208e-05,
+      "loss": 1.2681,
+      "step": 568
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8237179487179488e-05,
+      "loss": 1.2067,
+      "step": 569
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.826923076923077e-05,
+      "loss": 1.2994,
+      "step": 570
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8301282051282052e-05,
+      "loss": 1.2351,
+      "step": 571
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8333333333333333e-05,
+      "loss": 1.4352,
+      "step": 572
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8365384615384617e-05,
+      "loss": 1.3703,
+      "step": 573
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8397435897435897e-05,
+      "loss": 1.2003,
+      "step": 574
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.842948717948718e-05,
+      "loss": 1.1446,
+      "step": 575
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8461538461538465e-05,
+      "loss": 1.2639,
+      "step": 576
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8493589743589745e-05,
+      "loss": 1.276,
+      "step": 577
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.852564102564103e-05,
+      "loss": 1.3213,
+      "step": 578
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.855769230769231e-05,
+      "loss": 1.2576,
+      "step": 579
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8589743589743593e-05,
+      "loss": 1.3625,
+      "step": 580
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8621794871794873e-05,
+      "loss": 1.2288,
+      "step": 581
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8653846153846157e-05,
+      "loss": 1.332,
+      "step": 582
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8685897435897438e-05,
+      "loss": 1.3088,
+      "step": 583
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8717948717948718e-05,
+      "loss": 1.317,
+      "step": 584
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8750000000000002e-05,
+      "loss": 1.1452,
+      "step": 585
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8782051282051282e-05,
+      "loss": 1.1868,
+      "step": 586
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8814102564102566e-05,
+      "loss": 1.3063,
+      "step": 587
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8846153846153846e-05,
+      "loss": 1.2478,
+      "step": 588
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.887820512820513e-05,
+      "loss": 1.2739,
+      "step": 589
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.891025641025641e-05,
+      "loss": 1.0377,
+      "step": 590
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8942307692307694e-05,
+      "loss": 1.2701,
+      "step": 591
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8974358974358975e-05,
+      "loss": 1.2256,
+      "step": 592
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.900641025641026e-05,
+      "loss": 1.1509,
+      "step": 593
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.903846153846154e-05,
+      "loss": 1.3291,
+      "step": 594
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9070512820512823e-05,
+      "loss": 1.181,
+      "step": 595
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9102564102564106e-05,
+      "loss": 1.3482,
+      "step": 596
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9134615384615387e-05,
+      "loss": 1.1896,
+      "step": 597
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.916666666666667e-05,
+      "loss": 1.2062,
+      "step": 598
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.919871794871795e-05,
+      "loss": 1.2855,
+      "step": 599
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.923076923076923e-05,
+      "loss": 1.16,
+      "step": 600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9262820512820515e-05,
+      "loss": 1.2095,
+      "step": 601
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9294871794871796e-05,
+      "loss": 1.1864,
+      "step": 602
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.932692307692308e-05,
+      "loss": 1.3061,
+      "step": 603
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.935897435897436e-05,
+      "loss": 1.1976,
+      "step": 604
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9391025641025644e-05,
+      "loss": 1.2462,
+      "step": 605
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9423076923076924e-05,
+      "loss": 1.1893,
+      "step": 606
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9455128205128208e-05,
+      "loss": 1.2936,
+      "step": 607
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9487179487179488e-05,
+      "loss": 1.1363,
+      "step": 608
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9519230769230772e-05,
+      "loss": 1.2328,
+      "step": 609
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9551282051282052e-05,
+      "loss": 1.1231,
+      "step": 610
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9583333333333333e-05,
+      "loss": 1.1794,
+      "step": 611
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9615384615384617e-05,
+      "loss": 1.2345,
+      "step": 612
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9647435897435897e-05,
+      "loss": 1.2046,
+      "step": 613
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.967948717948718e-05,
+      "loss": 1.2378,
+      "step": 614
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9711538461538465e-05,
+      "loss": 1.2917,
+      "step": 615
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9743589743589745e-05,
+      "loss": 1.2803,
+      "step": 616
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.977564102564103e-05,
+      "loss": 1.3038,
+      "step": 617
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980769230769231e-05,
+      "loss": 1.2947,
+      "step": 618
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9839743589743593e-05,
+      "loss": 1.3283,
+      "step": 619
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9871794871794873e-05,
+      "loss": 1.1069,
+      "step": 620
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9903846153846157e-05,
+      "loss": 1.2436,
+      "step": 621
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9935897435897437e-05,
+      "loss": 1.2223,
+      "step": 622
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.996794871794872e-05,
+      "loss": 1.4128,
+      "step": 623
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2e-05,
+      "loss": 1.1411,
+      "step": 624
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999999878664707e-05,
+      "loss": 1.2986,
+      "step": 625
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999951465882e-05,
+      "loss": 1.3501,
+      "step": 626
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999890798236e-05,
+      "loss": 1.1473,
+      "step": 627
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999998058635333e-05,
+      "loss": 1.3262,
+      "step": 628
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999696661776e-05,
+      "loss": 1.1186,
+      "step": 629
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999563192967e-05,
+      "loss": 1.0817,
+      "step": 630
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999994054571096e-05,
+      "loss": 1.0715,
+      "step": 631
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999992234542078e-05,
+      "loss": 1.2833,
+      "step": 632
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999990171842654e-05,
+      "loss": 1.2457,
+      "step": 633
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999987866472878e-05,
+      "loss": 1.2947,
+      "step": 634
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999985318432804e-05,
+      "loss": 1.3566,
+      "step": 635
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999982527722498e-05,
+      "loss": 1.2403,
+      "step": 636
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999979494342022e-05,
+      "loss": 1.2124,
+      "step": 637
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999976218291455e-05,
+      "loss": 1.2933,
+      "step": 638
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999972699570876e-05,
+      "loss": 1.1772,
+      "step": 639
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999968938180364e-05,
+      "loss": 1.3607,
+      "step": 640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999964934120016e-05,
+      "loss": 1.3246,
+      "step": 641
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999996068738993e-05,
+      "loss": 1.377,
+      "step": 642
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999956197990205e-05,
+      "loss": 1.1585,
+      "step": 643
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999951465920953e-05,
+      "loss": 1.1391,
+      "step": 644
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999946491182284e-05,
+      "loss": 1.2883,
+      "step": 645
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999941273774327e-05,
+      "loss": 1.2339,
+      "step": 646
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999935813697204e-05,
+      "loss": 1.2161,
+      "step": 647
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999930110951043e-05,
+      "loss": 1.2542,
+      "step": 648
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999992416553599e-05,
+      "loss": 1.259,
+      "step": 649
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999917977452187e-05,
+      "loss": 1.2274,
+      "step": 650
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999911546699785e-05,
+      "loss": 1.1867,
+      "step": 651
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999904873278933e-05,
+      "loss": 1.0949,
+      "step": 652
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999897957189802e-05,
+      "loss": 1.2379,
+      "step": 653
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999890798432556e-05,
+      "loss": 1.2583,
+      "step": 654
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999883397007366e-05,
+      "loss": 1.291,
+      "step": 655
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999987575291442e-05,
+      "loss": 1.2824,
+      "step": 656
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999867866153894e-05,
+      "loss": 1.2318,
+      "step": 657
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999859736725984e-05,
+      "loss": 1.0811,
+      "step": 658
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999851364630886e-05,
+      "loss": 1.2604,
+      "step": 659
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999842749868808e-05,
+      "loss": 1.3026,
+      "step": 660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999833892439952e-05,
+      "loss": 1.3638,
+      "step": 661
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999824792344536e-05,
+      "loss": 1.1684,
+      "step": 662
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999981544958278e-05,
+      "loss": 1.1301,
+      "step": 663
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999805864154913e-05,
+      "loss": 1.3734,
+      "step": 664
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999796036061164e-05,
+      "loss": 1.2218,
+      "step": 665
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999785965301776e-05,
+      "loss": 1.307,
+      "step": 666
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999977565187699e-05,
+      "loss": 1.1156,
+      "step": 667
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999765095787055e-05,
+      "loss": 1.2382,
+      "step": 668
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999975429703223e-05,
+      "loss": 1.283,
+      "step": 669
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999974325561278e-05,
+      "loss": 1.2129,
+      "step": 670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999731971528965e-05,
+      "loss": 1.1246,
+      "step": 671
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999972044478107e-05,
+      "loss": 1.2581,
+      "step": 672
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999970867536936e-05,
+      "loss": 1.3525,
+      "step": 673
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999696663294133e-05,
+      "loss": 1.2455,
+      "step": 674
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999684408555673e-05,
+      "loss": 1.2458,
+      "step": 675
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999671911154285e-05,
+      "loss": 1.2249,
+      "step": 676
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999659171090263e-05,
+      "loss": 1.2822,
+      "step": 677
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999646188363925e-05,
+      "loss": 1.1638,
+      "step": 678
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999632962975578e-05,
+      "loss": 1.314,
+      "step": 679
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999961949492555e-05,
+      "loss": 1.1874,
+      "step": 680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999960578421416e-05,
+      "loss": 1.1661,
+      "step": 681
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999959183084175e-05,
+      "loss": 1.385,
+      "step": 682
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999957763480865e-05,
+      "loss": 1.3277,
+      "step": 683
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999956319611521e-05,
+      "loss": 1.2707,
+      "step": 684
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999548514761785e-05,
+      "loss": 1.3873,
+      "step": 685
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999953359074872e-05,
+      "loss": 1.2605,
+      "step": 686
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999951842407638e-05,
+      "loss": 1.2476,
+      "step": 687
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999503014745138e-05,
+      "loss": 1.1489,
+      "step": 688
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999487362755366e-05,
+      "loss": 1.1578,
+      "step": 689
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999471468107444e-05,
+      "loss": 1.3635,
+      "step": 690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999455330801752e-05,
+      "loss": 1.2727,
+      "step": 691
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999943895083869e-05,
+      "loss": 1.1463,
+      "step": 692
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999942232821865e-05,
+      "loss": 1.1905,
+      "step": 693
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999940546294204e-05,
+      "loss": 1.2359,
+      "step": 694
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999388355009266e-05,
+      "loss": 1.1262,
+      "step": 695
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999371004420744e-05,
+      "loss": 1.1489,
+      "step": 696
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999353411176893e-05,
+      "loss": 1.4049,
+      "step": 697
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999335575278143e-05,
+      "loss": 1.3056,
+      "step": 698
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999317496724925e-05,
+      "loss": 1.1688,
+      "step": 699
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999929917551768e-05,
+      "loss": 1.2602,
+      "step": 700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999928061165685e-05,
+      "loss": 1.1983,
+      "step": 701
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999261805142885e-05,
+      "loss": 1.3124,
+      "step": 702
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999242755976246e-05,
+      "loss": 1.294,
+      "step": 703
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999922346415739e-05,
+      "loss": 1.306,
+      "step": 704
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999203929686786e-05,
+      "loss": 1.2559,
+      "step": 705
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999184152564907e-05,
+      "loss": 1.1984,
+      "step": 706
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999916413279224e-05,
+      "loss": 1.1387,
+      "step": 707
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999143870369265e-05,
+      "loss": 1.1556,
+      "step": 708
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999123365296473e-05,
+      "loss": 1.2127,
+      "step": 709
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999102617574366e-05,
+      "loss": 1.2985,
+      "step": 710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999908162720344e-05,
+      "loss": 1.3099,
+      "step": 711
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999060394184214e-05,
+      "loss": 1.1193,
+      "step": 712
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99990389185172e-05,
+      "loss": 1.334,
+      "step": 713
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999901720020291e-05,
+      "loss": 1.1827,
+      "step": 714
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998995239241883e-05,
+      "loss": 1.3041,
+      "step": 715
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998973035634648e-05,
+      "loss": 1.0518,
+      "step": 716
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998950589381743e-05,
+      "loss": 1.2184,
+      "step": 717
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998927900483714e-05,
+      "loss": 1.2713,
+      "step": 718
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998904968941107e-05,
+      "loss": 1.3417,
+      "step": 719
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998881794754484e-05,
+      "loss": 1.2617,
+      "step": 720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998858377924408e-05,
+      "loss": 1.0888,
+      "step": 721
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998834718451444e-05,
+      "loss": 1.2785,
+      "step": 722
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999881081633616e-05,
+      "loss": 1.1735,
+      "step": 723
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999878667157915e-05,
+      "loss": 1.2768,
+      "step": 724
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998762284180996e-05,
+      "loss": 1.3057,
+      "step": 725
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999873765414228e-05,
+      "loss": 1.3346,
+      "step": 726
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999871278146361e-05,
+      "loss": 1.18,
+      "step": 727
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998687666145585e-05,
+      "loss": 1.2947,
+      "step": 728
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998662308188813e-05,
+      "loss": 1.2694,
+      "step": 729
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998636707593913e-05,
+      "loss": 1.1362,
+      "step": 730
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998610864361506e-05,
+      "loss": 1.2618,
+      "step": 731
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999858477849222e-05,
+      "loss": 1.251,
+      "step": 732
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998558449986684e-05,
+      "loss": 1.2083,
+      "step": 733
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998531878845536e-05,
+      "loss": 1.3571,
+      "step": 734
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999850506506943e-05,
+      "loss": 1.2645,
+      "step": 735
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999847800865901e-05,
+      "loss": 1.2654,
+      "step": 736
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998450709614928e-05,
+      "loss": 1.2852,
+      "step": 737
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998423167937852e-05,
+      "loss": 1.2494,
+      "step": 738
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998395383628457e-05,
+      "loss": 1.3072,
+      "step": 739
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998367356687405e-05,
+      "loss": 1.2336,
+      "step": 740
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998339087115378e-05,
+      "loss": 1.1061,
+      "step": 741
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998310574913074e-05,
+      "loss": 1.2132,
+      "step": 742
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999828182008117e-05,
+      "loss": 1.2281,
+      "step": 743
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998252822620373e-05,
+      "loss": 1.2778,
+      "step": 744
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998223582531386e-05,
+      "loss": 1.3342,
+      "step": 745
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998194099814913e-05,
+      "loss": 1.117,
+      "step": 746
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998164374471673e-05,
+      "loss": 1.3337,
+      "step": 747
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998134406502384e-05,
+      "loss": 1.298,
+      "step": 748
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999810419590778e-05,
+      "loss": 1.2684,
+      "step": 749
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998073742688592e-05,
+      "loss": 1.3133,
+      "step": 750
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998043046845555e-05,
+      "loss": 1.1408,
+      "step": 751
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998012108379417e-05,
+      "loss": 1.1187,
+      "step": 752
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997980927290928e-05,
+      "loss": 1.1013,
+      "step": 753
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997949503580844e-05,
+      "loss": 1.233,
+      "step": 754
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999791783724993e-05,
+      "loss": 1.2874,
+      "step": 755
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999788592829895e-05,
+      "loss": 1.1837,
+      "step": 756
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999785377672869e-05,
+      "loss": 1.3788,
+      "step": 757
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997821382539914e-05,
+      "loss": 1.265,
+      "step": 758
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997788745733415e-05,
+      "loss": 1.2434,
+      "step": 759
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997755866309988e-05,
+      "loss": 1.16,
+      "step": 760
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999772274427043e-05,
+      "loss": 1.0885,
+      "step": 761
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999768937961554e-05,
+      "loss": 1.1185,
+      "step": 762
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997655772346132e-05,
+      "loss": 1.1649,
+      "step": 763
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999762192246302e-05,
+      "loss": 1.2236,
+      "step": 764
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997587829967027e-05,
+      "loss": 1.1557,
+      "step": 765
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999755349485898e-05,
+      "loss": 1.2449,
+      "step": 766
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999751891713971e-05,
+      "loss": 1.1788,
+      "step": 767
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997484096810054e-05,
+      "loss": 1.319,
+      "step": 768
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999744903387087e-05,
+      "loss": 1.2918,
+      "step": 769
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997413728322992e-05,
+      "loss": 1.3302,
+      "step": 770
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997378180167285e-05,
+      "loss": 1.1213,
+      "step": 771
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999734238940461e-05,
+      "loss": 1.3056,
+      "step": 772
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997306356035838e-05,
+      "loss": 1.302,
+      "step": 773
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997270080061843e-05,
+      "loss": 1.2117,
+      "step": 774
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997233561483503e-05,
+      "loss": 1.0029,
+      "step": 775
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997196800301704e-05,
+      "loss": 1.1652,
+      "step": 776
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997159796517342e-05,
+      "loss": 1.3511,
+      "step": 777
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999712255013131e-05,
+      "loss": 1.2406,
+      "step": 778
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997085061144514e-05,
+      "loss": 1.298,
+      "step": 779
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997047329557867e-05,
+      "loss": 1.1713,
+      "step": 780
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997009355372276e-05,
+      "loss": 1.27,
+      "step": 781
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996971138588674e-05,
+      "loss": 1.3915,
+      "step": 782
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996932679207977e-05,
+      "loss": 1.2591,
+      "step": 783
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999689397723113e-05,
+      "loss": 1.2356,
+      "step": 784
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996855032659065e-05,
+      "loss": 1.1715,
+      "step": 785
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996815845492722e-05,
+      "loss": 1.3336,
+      "step": 786
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996776415733063e-05,
+      "loss": 1.2034,
+      "step": 787
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996736743381037e-05,
+      "loss": 1.2405,
+      "step": 788
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996696828437613e-05,
+      "loss": 1.336,
+      "step": 789
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996656670903753e-05,
+      "loss": 1.1536,
+      "step": 790
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999661627078044e-05,
+      "loss": 1.1751,
+      "step": 791
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996575628068645e-05,
+      "loss": 1.2148,
+      "step": 792
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996534742769355e-05,
+      "loss": 1.2368,
+      "step": 793
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999649361488357e-05,
+      "loss": 1.2365,
+      "step": 794
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996452244412285e-05,
+      "loss": 1.2354,
+      "step": 795
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996410631356496e-05,
+      "loss": 1.2867,
+      "step": 796
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996368775717228e-05,
+      "loss": 1.1538,
+      "step": 797
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996326677495482e-05,
+      "loss": 1.3332,
+      "step": 798
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996284336692286e-05,
+      "loss": 1.1594,
+      "step": 799
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996241753308673e-05,
+      "loss": 1.2407,
+      "step": 800
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996198927345665e-05,
+      "loss": 1.2637,
+      "step": 801
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996155858804307e-05,
+      "loss": 1.279,
+      "step": 802
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996112547685645e-05,
+      "loss": 1.2878,
+      "step": 803
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996068993990727e-05,
+      "loss": 1.218,
+      "step": 804
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996025197720615e-05,
+      "loss": 1.3537,
+      "step": 805
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999598115887637e-05,
+      "loss": 1.2573,
+      "step": 806
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995936877459053e-05,
+      "loss": 1.2912,
+      "step": 807
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999589235346975e-05,
+      "loss": 1.0626,
+      "step": 808
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995847586909537e-05,
+      "loss": 1.289,
+      "step": 809
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995802577779498e-05,
+      "loss": 1.294,
+      "step": 810
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995757326080727e-05,
+      "loss": 1.0917,
+      "step": 811
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999571183181432e-05,
+      "loss": 1.2915,
+      "step": 812
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995666094981385e-05,
+      "loss": 1.1936,
+      "step": 813
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995620115583033e-05,
+      "loss": 1.3362,
+      "step": 814
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999557389362037e-05,
+      "loss": 1.3747,
+      "step": 815
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995527429094534e-05,
+      "loss": 1.2194,
+      "step": 816
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995480722006636e-05,
+      "loss": 1.1555,
+      "step": 817
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999543377235782e-05,
+      "loss": 1.2619,
+      "step": 818
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999538658014922e-05,
+      "loss": 1.2614,
+      "step": 819
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995339145381982e-05,
+      "loss": 1.245,
+      "step": 820
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999529146805726e-05,
+      "loss": 1.2831,
+      "step": 821
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999524354817621e-05,
+      "loss": 1.311,
+      "step": 822
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999519538573999e-05,
+      "loss": 1.2444,
+      "step": 823
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995146980749777e-05,
+      "loss": 1.3217,
+      "step": 824
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999509833320674e-05,
+      "loss": 1.3011,
+      "step": 825
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999504944311206e-05,
+      "loss": 1.069,
+      "step": 826
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995000310466927e-05,
+      "loss": 1.2717,
+      "step": 827
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999495093527253e-05,
+      "loss": 1.1305,
+      "step": 828
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994901317530067e-05,
+      "loss": 1.2821,
+      "step": 829
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999485145724074e-05,
+      "loss": 1.2644,
+      "step": 830
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994801354405768e-05,
+      "loss": 1.2,
+      "step": 831
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994751009026355e-05,
+      "loss": 1.2813,
+      "step": 832
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994700421103734e-05,
+      "loss": 1.3734,
+      "step": 833
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994649590639124e-05,
+      "loss": 1.0997,
+      "step": 834
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999459851763376e-05,
+      "loss": 1.2062,
+      "step": 835
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994547202088886e-05,
+      "loss": 1.2562,
+      "step": 836
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994495644005746e-05,
+      "loss": 1.1882,
+      "step": 837
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994443843385583e-05,
+      "loss": 1.3298,
+      "step": 838
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994391800229666e-05,
+      "loss": 1.0893,
+      "step": 839
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999433951453925e-05,
+      "loss": 1.0441,
+      "step": 840
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999428698631561e-05,
+      "loss": 1.1184,
+      "step": 841
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994234215560014e-05,
+      "loss": 1.3511,
+      "step": 842
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994181202273745e-05,
+      "loss": 1.1674,
+      "step": 843
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999412794645809e-05,
+      "loss": 1.3425,
+      "step": 844
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994074448114342e-05,
+      "loss": 1.241,
+      "step": 845
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99940207072438e-05,
+      "loss": 1.209,
+      "step": 846
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993966723847766e-05,
+      "loss": 1.2059,
+      "step": 847
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999391249792755e-05,
+      "loss": 1.2116,
+      "step": 848
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999385802948447e-05,
+      "loss": 1.4,
+      "step": 849
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993803318519845e-05,
+      "loss": 1.3135,
+      "step": 850
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993748365035004e-05,
+      "loss": 1.1667,
+      "step": 851
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999369316903128e-05,
+      "loss": 1.2607,
+      "step": 852
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999363773051002e-05,
+      "loss": 1.2779,
+      "step": 853
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993582049472554e-05,
+      "loss": 1.1996,
+      "step": 854
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993526125920245e-05,
+      "loss": 1.1729,
+      "step": 855
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993469959854445e-05,
+      "loss": 1.2715,
+      "step": 856
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993413551276523e-05,
+      "loss": 1.2908,
+      "step": 857
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993356900187843e-05,
+      "loss": 1.2154,
+      "step": 858
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993300006589775e-05,
+      "loss": 1.0274,
+      "step": 859
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999324287048371e-05,
+      "loss": 1.1801,
+      "step": 860
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999318549187103e-05,
+      "loss": 1.2086,
+      "step": 861
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993127870753123e-05,
+      "loss": 1.0794,
+      "step": 862
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993070007131395e-05,
+      "loss": 1.2756,
+      "step": 863
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993011901007245e-05,
+      "loss": 1.3556,
+      "step": 864
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992953552382085e-05,
+      "loss": 1.129,
+      "step": 865
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992894961257332e-05,
+      "loss": 1.2093,
+      "step": 866
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992836127634402e-05,
+      "loss": 1.2394,
+      "step": 867
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999277705151473e-05,
+      "loss": 1.3471,
+      "step": 868
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992717732899746e-05,
+      "loss": 1.164,
+      "step": 869
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992658171790893e-05,
+      "loss": 1.2158,
+      "step": 870
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999259836818961e-05,
+      "loss": 1.2518,
+      "step": 871
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992538322097352e-05,
+      "loss": 1.1442,
+      "step": 872
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992478033515578e-05,
+      "loss": 1.1456,
+      "step": 873
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992417502445746e-05,
+      "loss": 1.3147,
+      "step": 874
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992356728889332e-05,
+      "loss": 1.2068,
+      "step": 875
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992295712847802e-05,
+      "loss": 1.263,
+      "step": 876
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992234454322643e-05,
+      "loss": 1.1478,
+      "step": 877
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992172953315343e-05,
+      "loss": 1.2127,
+      "step": 878
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992111209827386e-05,
+      "loss": 1.3162,
+      "step": 879
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999204922386028e-05,
+      "loss": 1.2211,
+      "step": 880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999198699541552e-05,
+      "loss": 1.1908,
+      "step": 881
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999192452449463e-05,
+      "loss": 1.1333,
+      "step": 882
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999186181109911e-05,
+      "loss": 1.396,
+      "step": 883
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999179885523049e-05,
+      "loss": 1.1761,
+      "step": 884
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991735656890293e-05,
+      "loss": 1.4088,
+      "step": 885
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991672216080064e-05,
+      "loss": 1.2891,
+      "step": 886
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999160853280133e-05,
+      "loss": 1.218,
+      "step": 887
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991544607055642e-05,
+      "loss": 1.1503,
+      "step": 888
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999148043884455e-05,
+      "loss": 1.2971,
+      "step": 889
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991416028169612e-05,
+      "loss": 1.1763,
+      "step": 890
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999135137503239e-05,
+      "loss": 1.1336,
+      "step": 891
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991286479434456e-05,
+      "loss": 1.231,
+      "step": 892
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999122134137738e-05,
+      "loss": 1.2357,
+      "step": 893
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991155960862743e-05,
+      "loss": 1.186,
+      "step": 894
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999109033789214e-05,
+      "loss": 1.4233,
+      "step": 895
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991024472467156e-05,
+      "loss": 1.1226,
+      "step": 896
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990958364589388e-05,
+      "loss": 1.3527,
+      "step": 897
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999089201426044e-05,
+      "loss": 1.2493,
+      "step": 898
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999082542148193e-05,
+      "loss": 1.2601,
+      "step": 899
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990758586255467e-05,
+      "loss": 1.2308,
+      "step": 900
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990691508582678e-05,
+      "loss": 1.2949,
+      "step": 901
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990624188465183e-05,
+      "loss": 1.2801,
+      "step": 902
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990556625904623e-05,
+      "loss": 1.3565,
+      "step": 903
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990488820902633e-05,
+      "loss": 1.284,
+      "step": 904
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990420773460864e-05,
+      "loss": 1.1609,
+      "step": 905
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999035248358096e-05,
+      "loss": 1.3245,
+      "step": 906
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999028395126458e-05,
+      "loss": 1.2477,
+      "step": 907
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990215176513394e-05,
+      "loss": 1.2339,
+      "step": 908
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990146159329065e-05,
+      "loss": 1.1649,
+      "step": 909
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990076899713268e-05,
+      "loss": 1.2207,
+      "step": 910
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999000739766768e-05,
+      "loss": 1.262,
+      "step": 911
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989937653193995e-05,
+      "loss": 1.3906,
+      "step": 912
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989867666293904e-05,
+      "loss": 1.2355,
+      "step": 913
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989797436969103e-05,
+      "loss": 1.2883,
+      "step": 914
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989726965221298e-05,
+      "loss": 1.2961,
+      "step": 915
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99896562510522e-05,
+      "loss": 1.1586,
+      "step": 916
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989585294463518e-05,
+      "loss": 1.2783,
+      "step": 917
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998951409545698e-05,
+      "loss": 1.1843,
+      "step": 918
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989442654034315e-05,
+      "loss": 1.2637,
+      "step": 919
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989370970197257e-05,
+      "loss": 1.2778,
+      "step": 920
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998929904394754e-05,
+      "loss": 1.3436,
+      "step": 921
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989226875286913e-05,
+      "loss": 1.2798,
+      "step": 922
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989154464217128e-05,
+      "loss": 1.1775,
+      "step": 923
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989081810739943e-05,
+      "loss": 1.2188,
+      "step": 924
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989008914857115e-05,
+      "loss": 1.2963,
+      "step": 925
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998893577657042e-05,
+      "loss": 1.1023,
+      "step": 926
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998886239588163e-05,
+      "loss": 1.1814,
+      "step": 927
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988788772792523e-05,
+      "loss": 1.1401,
+      "step": 928
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998871490730489e-05,
+      "loss": 1.2351,
+      "step": 929
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988640799420524e-05,
+      "loss": 1.1938,
+      "step": 930
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988566449141223e-05,
+      "loss": 1.3766,
+      "step": 931
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988491856468785e-05,
+      "loss": 1.1886,
+      "step": 932
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988417021405027e-05,
+      "loss": 1.0902,
+      "step": 933
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998834194395176e-05,
+      "loss": 1.2032,
+      "step": 934
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988266624110813e-05,
+      "loss": 1.2086,
+      "step": 935
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988191061884005e-05,
+      "loss": 1.3511,
+      "step": 936
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988115257273176e-05,
+      "loss": 1.1567,
+      "step": 937
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988039210280167e-05,
+      "loss": 1.1911,
+      "step": 938
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987962920906816e-05,
+      "loss": 1.0997,
+      "step": 939
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998788638915498e-05,
+      "loss": 1.1776,
+      "step": 940
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987809615026513e-05,
+      "loss": 1.249,
+      "step": 941
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998773259852328e-05,
+      "loss": 1.2391,
+      "step": 942
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987655339647153e-05,
+      "loss": 1.3302,
+      "step": 943
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99875778384e-05,
+      "loss": 1.2822,
+      "step": 944
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998750009478371e-05,
+      "loss": 1.2102,
+      "step": 945
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987422108800157e-05,
+      "loss": 1.3991,
+      "step": 946
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998734388045125e-05,
+      "loss": 1.3216,
+      "step": 947
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998726540973887e-05,
+      "loss": 1.2474,
+      "step": 948
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987186696664937e-05,
+      "loss": 1.2626,
+      "step": 949
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987107741231352e-05,
+      "loss": 1.0399,
+      "step": 950
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987028543440035e-05,
+      "loss": 1.2634,
+      "step": 951
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986949103292904e-05,
+      "loss": 1.1896,
+      "step": 952
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998686942079189e-05,
+      "loss": 1.1303,
+      "step": 953
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986789495938922e-05,
+      "loss": 1.3772,
+      "step": 954
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986709328735944e-05,
+      "loss": 1.3507,
+      "step": 955
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986628919184904e-05,
+      "loss": 1.2493,
+      "step": 956
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986548267287745e-05,
+      "loss": 1.2646,
+      "step": 957
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998646737304643e-05,
+      "loss": 1.2537,
+      "step": 958
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986386236462926e-05,
+      "loss": 1.3714,
+      "step": 959
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998630485753919e-05,
+      "loss": 1.2176,
+      "step": 960
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998622323627721e-05,
+      "loss": 1.1262,
+      "step": 961
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998614137267896e-05,
+      "loss": 1.131,
+      "step": 962
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998605926674642e-05,
+      "loss": 1.2023,
+      "step": 963
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985976918481595e-05,
+      "loss": 1.3105,
+      "step": 964
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985894327886476e-05,
+      "loss": 1.2491,
+      "step": 965
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985811494963074e-05,
+      "loss": 1.1576,
+      "step": 966
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985728419713388e-05,
+      "loss": 1.3004,
+      "step": 967
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998564510213944e-05,
+      "loss": 1.2429,
+      "step": 968
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985561542243256e-05,
+      "loss": 1.3425,
+      "step": 969
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998547774002686e-05,
+      "loss": 1.2309,
+      "step": 970
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985393695492283e-05,
+      "loss": 1.3126,
+      "step": 971
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998530940864157e-05,
+      "loss": 1.2733,
+      "step": 972
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998522487947676e-05,
+      "loss": 1.1492,
+      "step": 973
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985140107999908e-05,
+      "loss": 1.2833,
+      "step": 974
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985055094213072e-05,
+      "loss": 1.3055,
+      "step": 975
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984969838118316e-05,
+      "loss": 1.1628,
+      "step": 976
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984884339717704e-05,
+      "loss": 1.3542,
+      "step": 977
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984798599013315e-05,
+      "loss": 0.9628,
+      "step": 978
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984712616007226e-05,
+      "loss": 1.2179,
+      "step": 979
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998462639070153e-05,
+      "loss": 1.1616,
+      "step": 980
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998453992309831e-05,
+      "loss": 1.2166,
+      "step": 981
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984453213199673e-05,
+      "loss": 1.297,
+      "step": 982
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984366261007716e-05,
+      "loss": 1.0402,
+      "step": 983
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998427906652456e-05,
+      "loss": 1.1697,
+      "step": 984
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984191629752306e-05,
+      "loss": 1.3859,
+      "step": 985
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998410395069309e-05,
+      "loss": 1.133,
+      "step": 986
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998401602934903e-05,
+      "loss": 1.0854,
+      "step": 987
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983927865722262e-05,
+      "loss": 1.2883,
+      "step": 988
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983839459814925e-05,
+      "loss": 1.1957,
+      "step": 989
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998375081162917e-05,
+      "loss": 1.0899,
+      "step": 990
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998366192116714e-05,
+      "loss": 1.1705,
+      "step": 991
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983572788431e-05,
+      "loss": 1.2168,
+      "step": 992
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983483413422907e-05,
+      "loss": 1.207,
+      "step": 993
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998339379614503e-05,
+      "loss": 1.2724,
+      "step": 994
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983303936599553e-05,
+      "loss": 1.1754,
+      "step": 995
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983213834788643e-05,
+      "loss": 1.2103,
+      "step": 996
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983123490714492e-05,
+      "loss": 1.1977,
+      "step": 997
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983032904379296e-05,
+      "loss": 1.2669,
+      "step": 998
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982942075785247e-05,
+      "loss": 1.3245,
+      "step": 999
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982851004934557e-05,
+      "loss": 1.2506,
+      "step": 1000
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998275969182943e-05,
+      "loss": 1.268,
+      "step": 1001
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998266813647208e-05,
+      "loss": 1.1517,
+      "step": 1002
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982576338864738e-05,
+      "loss": 1.2147,
+      "step": 1003
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982484299009624e-05,
+      "loss": 1.3175,
+      "step": 1004
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982392016908975e-05,
+      "loss": 1.1898,
+      "step": 1005
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982299492565028e-05,
+      "loss": 1.2517,
+      "step": 1006
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982206725980026e-05,
+      "loss": 1.1476,
+      "step": 1007
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982113717156225e-05,
+      "loss": 1.1334,
+      "step": 1008
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982020466095882e-05,
+      "loss": 1.1792,
+      "step": 1009
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981926972801257e-05,
+      "loss": 1.1987,
+      "step": 1010
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998183323727462e-05,
+      "loss": 1.2462,
+      "step": 1011
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981739259518246e-05,
+      "loss": 1.1164,
+      "step": 1012
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981645039534417e-05,
+      "loss": 1.1552,
+      "step": 1013
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981550577325417e-05,
+      "loss": 1.1909,
+      "step": 1014
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998145587289354e-05,
+      "loss": 1.3222,
+      "step": 1015
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998136092624108e-05,
+      "loss": 1.163,
+      "step": 1016
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998126573737035e-05,
+      "loss": 1.1998,
+      "step": 1017
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981170306283647e-05,
+      "loss": 1.3237,
+      "step": 1018
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99810746329833e-05,
+      "loss": 1.0621,
+      "step": 1019
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998097871747162e-05,
+      "loss": 1.1676,
+      "step": 1020
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980882559750947e-05,
+      "loss": 1.3889,
+      "step": 1021
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99807861598236e-05,
+      "loss": 1.1837,
+      "step": 1022
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980689517691928e-05,
+      "loss": 1.2093,
+      "step": 1023
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980592633358276e-05,
+      "loss": 1.3354,
+      "step": 1024
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998049550682499e-05,
+      "loss": 1.2364,
+      "step": 1025
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998039813809443e-05,
+      "loss": 1.131,
+      "step": 1026
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998030052716896e-05,
+      "loss": 1.2661,
+      "step": 1027
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980202674050945e-05,
+      "loss": 1.2164,
+      "step": 1028
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980104578742762e-05,
+      "loss": 1.3367,
+      "step": 1029
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998000624124679e-05,
+      "loss": 1.3424,
+      "step": 1030
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979907661565418e-05,
+      "loss": 1.269,
+      "step": 1031
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979808839701037e-05,
+      "loss": 1.2531,
+      "step": 1032
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979709775656048e-05,
+      "loss": 1.1118,
+      "step": 1033
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979610469432847e-05,
+      "loss": 1.297,
+      "step": 1034
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997951092103385e-05,
+      "loss": 1.256,
+      "step": 1035
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979411130461475e-05,
+      "loss": 1.3325,
+      "step": 1036
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997931109771814e-05,
+      "loss": 1.3052,
+      "step": 1037
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997921082280627e-05,
+      "loss": 1.2008,
+      "step": 1038
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99791103057283e-05,
+      "loss": 1.306,
+      "step": 1039
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979009546486675e-05,
+      "loss": 1.1101,
+      "step": 1040
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978908545083833e-05,
+      "loss": 1.4173,
+      "step": 1041
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978807301522226e-05,
+      "loss": 1.3261,
+      "step": 1042
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978705815804312e-05,
+      "loss": 1.3688,
+      "step": 1043
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978604087932557e-05,
+      "loss": 1.2663,
+      "step": 1044
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997850211790943e-05,
+      "loss": 1.179,
+      "step": 1045
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978399905737395e-05,
+      "loss": 1.2626,
+      "step": 1046
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978297451418945e-05,
+      "loss": 1.1171,
+      "step": 1047
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978194754956558e-05,
+      "loss": 1.3336,
+      "step": 1048
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997809181635273e-05,
+      "loss": 1.1139,
+      "step": 1049
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977988635609957e-05,
+      "loss": 1.4329,
+      "step": 1050
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977885212730745e-05,
+      "loss": 1.1361,
+      "step": 1051
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977781547717604e-05,
+      "loss": 1.2852,
+      "step": 1052
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977677640573048e-05,
+      "loss": 1.3523,
+      "step": 1053
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977573491299597e-05,
+      "loss": 1.139,
+      "step": 1054
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997746909989978e-05,
+      "loss": 1.1717,
+      "step": 1055
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977364466376135e-05,
+      "loss": 1.0198,
+      "step": 1056
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977259590731193e-05,
+      "loss": 1.3434,
+      "step": 1057
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977154472967504e-05,
+      "loss": 1.1846,
+      "step": 1058
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977049113087615e-05,
+      "loss": 1.238,
+      "step": 1059
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997694351109409e-05,
+      "loss": 1.2377,
+      "step": 1060
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976837666989484e-05,
+      "loss": 1.0752,
+      "step": 1061
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976731580776373e-05,
+      "loss": 1.0998,
+      "step": 1062
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976625252457322e-05,
+      "loss": 1.1403,
+      "step": 1063
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976518682034917e-05,
+      "loss": 1.2848,
+      "step": 1064
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976411869511746e-05,
+      "loss": 1.1334,
+      "step": 1065
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976304814890396e-05,
+      "loss": 1.2218,
+      "step": 1066
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997619751817347e-05,
+      "loss": 1.103,
+      "step": 1067
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976089979363566e-05,
+      "loss": 1.1543,
+      "step": 1068
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99759821984633e-05,
+      "loss": 1.1688,
+      "step": 1069
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975874175475284e-05,
+      "loss": 1.2032,
+      "step": 1070
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997576591040214e-05,
+      "loss": 1.1436,
+      "step": 1071
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975657403246492e-05,
+      "loss": 1.2209,
+      "step": 1072
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975548654010978e-05,
+      "loss": 1.2002,
+      "step": 1073
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975439662698234e-05,
+      "loss": 1.2382,
+      "step": 1074
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975330429310908e-05,
+      "loss": 1.1636,
+      "step": 1075
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975220953851648e-05,
+      "loss": 1.2928,
+      "step": 1076
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975111236323112e-05,
+      "loss": 1.2209,
+      "step": 1077
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997500127672796e-05,
+      "loss": 1.1427,
+      "step": 1078
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974891075068864e-05,
+      "loss": 1.2704,
+      "step": 1079
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974780631348495e-05,
+      "loss": 1.0925,
+      "step": 1080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997466994556954e-05,
+      "loss": 1.2529,
+      "step": 1081
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974559017734676e-05,
+      "loss": 1.1912,
+      "step": 1082
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99744478478466e-05,
+      "loss": 1.2611,
+      "step": 1083
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974336435908005e-05,
+      "loss": 1.268,
+      "step": 1084
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99742247819216e-05,
+      "loss": 1.1333,
+      "step": 1085
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974112885890094e-05,
+      "loss": 1.0851,
+      "step": 1086
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974000747816203e-05,
+      "loss": 1.2656,
+      "step": 1087
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973888367702643e-05,
+      "loss": 1.2211,
+      "step": 1088
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973775745552146e-05,
+      "loss": 1.0267,
+      "step": 1089
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973662881367442e-05,
+      "loss": 1.3079,
+      "step": 1090
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973549775151273e-05,
+      "loss": 1.1812,
+      "step": 1091
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997343642690638e-05,
+      "loss": 1.2295,
+      "step": 1092
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973322836635517e-05,
+      "loss": 1.1402,
+      "step": 1093
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973209004341442e-05,
+      "loss": 1.3507,
+      "step": 1094
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997309493002691e-05,
+      "loss": 1.34,
+      "step": 1095
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972980613694698e-05,
+      "loss": 1.1633,
+      "step": 1096
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972866055347572e-05,
+      "loss": 1.2704,
+      "step": 1097
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972751254988317e-05,
+      "loss": 1.2576,
+      "step": 1098
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997263621261972e-05,
+      "loss": 1.1648,
+      "step": 1099
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997252092824457e-05,
+      "loss": 1.2776,
+      "step": 1100
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972405401865663e-05,
+      "loss": 0.9589,
+      "step": 1101
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997228963348581e-05,
+      "loss": 1.24,
+      "step": 1102
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997217362310781e-05,
+      "loss": 1.1634,
+      "step": 1103
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972057370734483e-05,
+      "loss": 1.1486,
+      "step": 1104
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971940876368653e-05,
+      "loss": 1.3036,
+      "step": 1105
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971824140013143e-05,
+      "loss": 1.129,
+      "step": 1106
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971707161670786e-05,
+      "loss": 1.2761,
+      "step": 1107
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997158994134442e-05,
+      "loss": 1.3766,
+      "step": 1108
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99714724790369e-05,
+      "loss": 1.1519,
+      "step": 1109
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971354774751063e-05,
+      "loss": 1.155,
+      "step": 1110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971236828489768e-05,
+      "loss": 1.3924,
+      "step": 1111
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971118640255883e-05,
+      "loss": 1.1159,
+      "step": 1112
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997100021005227e-05,
+      "loss": 1.0808,
+      "step": 1113
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997088153788181e-05,
+      "loss": 1.3389,
+      "step": 1114
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970762623747373e-05,
+      "loss": 1.191,
+      "step": 1115
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970643467651853e-05,
+      "loss": 1.2612,
+      "step": 1116
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970524069598136e-05,
+      "loss": 1.1557,
+      "step": 1117
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970404429589126e-05,
+      "loss": 1.2793,
+      "step": 1118
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997028454762772e-05,
+      "loss": 1.2276,
+      "step": 1119
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997016442371683e-05,
+      "loss": 1.247,
+      "step": 1120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997004405785937e-05,
+      "loss": 1.2822,
+      "step": 1121
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969923450058264e-05,
+      "loss": 1.0169,
+      "step": 1122
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969802600316433e-05,
+      "loss": 1.274,
+      "step": 1123
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969681508636814e-05,
+      "loss": 1.2249,
+      "step": 1124
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969560175022343e-05,
+      "loss": 1.2475,
+      "step": 1125
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996943859947597e-05,
+      "loss": 1.0302,
+      "step": 1126
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996931678200064e-05,
+      "loss": 1.1662,
+      "step": 1127
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996919472259931e-05,
+      "loss": 1.2672,
+      "step": 1128
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996907242127494e-05,
+      "loss": 1.1046,
+      "step": 1129
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968949878030503e-05,
+      "loss": 1.1448,
+      "step": 1130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996882709286897e-05,
+      "loss": 1.3466,
+      "step": 1131
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996870406579332e-05,
+      "loss": 1.2283,
+      "step": 1132
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968580796806542e-05,
+      "loss": 1.3796,
+      "step": 1133
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968457285911624e-05,
+      "loss": 1.3129,
+      "step": 1134
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996833353311156e-05,
+      "loss": 1.3196,
+      "step": 1135
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996820953840936e-05,
+      "loss": 1.2377,
+      "step": 1136
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968085301808026e-05,
+      "loss": 1.2302,
+      "step": 1137
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996796082331058e-05,
+      "loss": 1.3188,
+      "step": 1138
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967836102920043e-05,
+      "loss": 1.1691,
+      "step": 1139
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996771114063943e-05,
+      "loss": 1.2856,
+      "step": 1140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996758593647179e-05,
+      "loss": 1.2892,
+      "step": 1141
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996746049042015e-05,
+      "loss": 1.3419,
+      "step": 1142
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967334802487553e-05,
+      "loss": 1.1602,
+      "step": 1143
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996720887267706e-05,
+      "loss": 1.1587,
+      "step": 1144
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9967082700991712e-05,
+      "loss": 1.2549,
+      "step": 1145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966956287434586e-05,
+      "loss": 1.2163,
+      "step": 1146
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996682963200874e-05,
+      "loss": 1.2345,
+      "step": 1147
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966702734717248e-05,
+      "loss": 1.2305,
+      "step": 1148
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966575595563195e-05,
+      "loss": 1.2027,
+      "step": 1149
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996644821454966e-05,
+      "loss": 1.2792,
+      "step": 1150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996632059167974e-05,
+      "loss": 1.1574,
+      "step": 1151
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996619272695653e-05,
+      "loss": 1.1609,
+      "step": 1152
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996606462038313e-05,
+      "loss": 1.2553,
+      "step": 1153
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965936271962652e-05,
+      "loss": 1.2397,
+      "step": 1154
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965807681698208e-05,
+      "loss": 1.2286,
+      "step": 1155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996567884959292e-05,
+      "loss": 1.1057,
+      "step": 1156
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965549775649914e-05,
+      "loss": 1.2155,
+      "step": 1157
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965420459872325e-05,
+      "loss": 1.1432,
+      "step": 1158
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965290902263286e-05,
+      "loss": 1.1826,
+      "step": 1159
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965161102825944e-05,
+      "loss": 1.3797,
+      "step": 1160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996503106156345e-05,
+      "loss": 1.39,
+      "step": 1161
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964900778478958e-05,
+      "loss": 1.3645,
+      "step": 1162
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996477025357563e-05,
+      "loss": 1.306,
+      "step": 1163
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964639486856634e-05,
+      "loss": 1.1909,
+      "step": 1164
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996450847832514e-05,
+      "loss": 1.1721,
+      "step": 1165
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996437722798433e-05,
+      "loss": 1.2819,
+      "step": 1166
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996424573583739e-05,
+      "loss": 1.236,
+      "step": 1167
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996411400188751e-05,
+      "loss": 1.3016,
+      "step": 1168
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963982026137886e-05,
+      "loss": 1.2812,
+      "step": 1169
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963849808591723e-05,
+      "loss": 1.2618,
+      "step": 1170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963717349252226e-05,
+      "loss": 1.2886,
+      "step": 1171
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996358464812261e-05,
+      "loss": 1.2599,
+      "step": 1172
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99634517052061e-05,
+      "loss": 1.204,
+      "step": 1173
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963318520505915e-05,
+      "loss": 1.1608,
+      "step": 1174
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963185094025293e-05,
+      "loss": 1.2497,
+      "step": 1175
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996305142576747e-05,
+      "loss": 1.1431,
+      "step": 1176
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962917515735686e-05,
+      "loss": 1.1112,
+      "step": 1177
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962783363933193e-05,
+      "loss": 1.2135,
+      "step": 1178
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996264897036325e-05,
+      "loss": 1.1269,
+      "step": 1179
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962514335029116e-05,
+      "loss": 1.1477,
+      "step": 1180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962379457934058e-05,
+      "loss": 1.3544,
+      "step": 1181
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962244339081347e-05,
+      "loss": 1.2752,
+      "step": 1182
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962108978474265e-05,
+      "loss": 1.2329,
+      "step": 1183
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961973376116096e-05,
+      "loss": 1.2283,
+      "step": 1184
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996183753201013e-05,
+      "loss": 1.0855,
+      "step": 1185
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996170144615966e-05,
+      "loss": 1.2385,
+      "step": 1186
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961565118567997e-05,
+      "loss": 1.3697,
+      "step": 1187
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961428549238445e-05,
+      "loss": 1.1743,
+      "step": 1188
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961291738174316e-05,
+      "loss": 1.2727,
+      "step": 1189
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996115468537893e-05,
+      "loss": 1.2826,
+      "step": 1190
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961017390855616e-05,
+      "loss": 1.3785,
+      "step": 1191
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960879854607702e-05,
+      "loss": 1.2824,
+      "step": 1192
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996074207663853e-05,
+      "loss": 1.3485,
+      "step": 1193
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960604056951444e-05,
+      "loss": 1.2071,
+      "step": 1194
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960465795549787e-05,
+      "loss": 1.2335,
+      "step": 1195
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996032729243692e-05,
+      "loss": 1.2443,
+      "step": 1196
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99601885476162e-05,
+      "loss": 1.0954,
+      "step": 1197
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960049561090995e-05,
+      "loss": 1.1573,
+      "step": 1198
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995991033286468e-05,
+      "loss": 1.2146,
+      "step": 1199
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959770862940632e-05,
+      "loss": 1.3447,
+      "step": 1200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959631151322235e-05,
+      "loss": 1.2871,
+      "step": 1201
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995949119801288e-05,
+      "loss": 1.0627,
+      "step": 1202
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995935100301597e-05,
+      "loss": 1.3185,
+      "step": 1203
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959210566334893e-05,
+      "loss": 1.2826,
+      "step": 1204
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959069887973067e-05,
+      "loss": 1.0502,
+      "step": 1205
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958928967933903e-05,
+      "loss": 0.9676,
+      "step": 1206
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995878780622082e-05,
+      "loss": 1.1575,
+      "step": 1207
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958646402837247e-05,
+      "loss": 1.1738,
+      "step": 1208
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995850475778661e-05,
+      "loss": 1.1055,
+      "step": 1209
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958362871072353e-05,
+      "loss": 1.1716,
+      "step": 1210
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958220742697915e-05,
+      "loss": 1.1922,
+      "step": 1211
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995807837266674e-05,
+      "loss": 1.2321,
+      "step": 1212
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957935760982297e-05,
+      "loss": 1.3087,
+      "step": 1213
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957792907648033e-05,
+      "loss": 1.2639,
+      "step": 1214
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957649812667422e-05,
+      "loss": 1.132,
+      "step": 1215
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957506476043934e-05,
+      "loss": 1.2745,
+      "step": 1216
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995736289778105e-05,
+      "loss": 1.1925,
+      "step": 1217
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957219077882245e-05,
+      "loss": 1.1463,
+      "step": 1218
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957075016351023e-05,
+      "loss": 1.2627,
+      "step": 1219
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995693071319087e-05,
+      "loss": 1.3496,
+      "step": 1220
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995678616840529e-05,
+      "loss": 1.3386,
+      "step": 1221
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956641381997795e-05,
+      "loss": 1.2478,
+      "step": 1222
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956496353971894e-05,
+      "loss": 1.3026,
+      "step": 1223
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995635108433111e-05,
+      "loss": 1.2367,
+      "step": 1224
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995620557307896e-05,
+      "loss": 1.2754,
+      "step": 1225
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956059820218982e-05,
+      "loss": 1.1829,
+      "step": 1226
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955913825754713e-05,
+      "loss": 1.2645,
+      "step": 1227
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955767589689697e-05,
+      "loss": 1.2839,
+      "step": 1228
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955621112027476e-05,
+      "loss": 1.1689,
+      "step": 1229
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995547439277161e-05,
+      "loss": 1.1256,
+      "step": 1230
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955327431925663e-05,
+      "loss": 1.3079,
+      "step": 1231
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955180229493193e-05,
+      "loss": 1.1696,
+      "step": 1232
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955032785477778e-05,
+      "loss": 1.085,
+      "step": 1233
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954885099882992e-05,
+      "loss": 1.3601,
+      "step": 1234
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954737172712422e-05,
+      "loss": 1.2841,
+      "step": 1235
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954589003969657e-05,
+      "loss": 1.1302,
+      "step": 1236
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954440593658294e-05,
+      "loss": 1.2575,
+      "step": 1237
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995429194178193e-05,
+      "loss": 1.3168,
+      "step": 1238
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954143048344175e-05,
+      "loss": 1.398,
+      "step": 1239
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995399391334864e-05,
+      "loss": 1.117,
+      "step": 1240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995384453679895e-05,
+      "loss": 1.3484,
+      "step": 1241
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953694918698726e-05,
+      "loss": 1.3178,
+      "step": 1242
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99535450590516e-05,
+      "loss": 1.2742,
+      "step": 1243
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953394957861206e-05,
+      "loss": 1.3006,
+      "step": 1244
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953244615131187e-05,
+      "loss": 1.2901,
+      "step": 1245
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953094030865197e-05,
+      "loss": 1.1588,
+      "step": 1246
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995294320506688e-05,
+      "loss": 1.1041,
+      "step": 1247
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952792137739908e-05,
+      "loss": 1.2087,
+      "step": 1248
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995264082888794e-05,
+      "loss": 1.3398,
+      "step": 1249
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952489278514644e-05,
+      "loss": 1.3929,
+      "step": 1250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952337486623704e-05,
+      "loss": 1.1961,
+      "step": 1251
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952185453218803e-05,
+      "loss": 1.0494,
+      "step": 1252
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952033178303632e-05,
+      "loss": 1.1308,
+      "step": 1253
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995188066188188e-05,
+      "loss": 1.0657,
+      "step": 1254
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951727903957252e-05,
+      "loss": 1.2636,
+      "step": 1255
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951574904533456e-05,
+      "loss": 1.2071,
+      "step": 1256
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951421663614204e-05,
+      "loss": 1.2073,
+      "step": 1257
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951268181203213e-05,
+      "loss": 1.2823,
+      "step": 1258
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995111445730421e-05,
+      "loss": 1.2393,
+      "step": 1259
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950960491920923e-05,
+      "loss": 1.2487,
+      "step": 1260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950806285057092e-05,
+      "loss": 1.1947,
+      "step": 1261
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950651836716453e-05,
+      "loss": 1.0367,
+      "step": 1262
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950497146902757e-05,
+      "loss": 1.2617,
+      "step": 1263
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950342215619764e-05,
+      "loss": 1.2343,
+      "step": 1264
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950187042871226e-05,
+      "loss": 1.2814,
+      "step": 1265
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995003162866091e-05,
+      "loss": 1.1466,
+      "step": 1266
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994987597299259e-05,
+      "loss": 1.173,
+      "step": 1267
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994972007587004e-05,
+      "loss": 1.1522,
+      "step": 1268
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949563937297045e-05,
+      "loss": 1.0375,
+      "step": 1269
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949407557277394e-05,
+      "loss": 1.1829,
+      "step": 1270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949250935814884e-05,
+      "loss": 1.1732,
+      "step": 1271
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994909407291331e-05,
+      "loss": 1.243,
+      "step": 1272
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948936968576483e-05,
+      "loss": 1.1718,
+      "step": 1273
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948779622808215e-05,
+      "loss": 1.1322,
+      "step": 1274
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948622035612326e-05,
+      "loss": 1.0975,
+      "step": 1275
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948464206992635e-05,
+      "loss": 1.1532,
+      "step": 1276
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948306136952976e-05,
+      "loss": 1.3005,
+      "step": 1277
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948147825497184e-05,
+      "loss": 1.0659,
+      "step": 1278
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99479892726291e-05,
+      "loss": 1.1063,
+      "step": 1279
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947830478352578e-05,
+      "loss": 1.1851,
+      "step": 1280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994767144267146e-05,
+      "loss": 1.1558,
+      "step": 1281
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994751216558961e-05,
+      "loss": 1.222,
+      "step": 1282
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947352647110895e-05,
+      "loss": 1.2143,
+      "step": 1283
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947192887239188e-05,
+      "loss": 1.2682,
+      "step": 1284
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947032885978365e-05,
+      "loss": 1.3809,
+      "step": 1285
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99468726433323e-05,
+      "loss": 1.0527,
+      "step": 1286
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946712159304894e-05,
+      "loss": 1.2901,
+      "step": 1287
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946551433900033e-05,
+      "loss": 1.2868,
+      "step": 1288
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946390467121627e-05,
+      "loss": 1.2427,
+      "step": 1289
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994622925897357e-05,
+      "loss": 1.3041,
+      "step": 1290
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994606780945978e-05,
+      "loss": 1.2063,
+      "step": 1291
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945906118584173e-05,
+      "loss": 1.2091,
+      "step": 1292
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994574418635068e-05,
+      "loss": 1.2198,
+      "step": 1293
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994558201276322e-05,
+      "loss": 1.1687,
+      "step": 1294
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945419597825736e-05,
+      "loss": 1.1849,
+      "step": 1295
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945256941542163e-05,
+      "loss": 1.278,
+      "step": 1296
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945094043916456e-05,
+      "loss": 1.2511,
+      "step": 1297
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944930904952564e-05,
+      "loss": 1.2785,
+      "step": 1298
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944767524654446e-05,
+      "loss": 1.2932,
+      "step": 1299
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944603903026064e-05,
+      "loss": 1.2441,
+      "step": 1300
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944440040071392e-05,
+      "loss": 1.305,
+      "step": 1301
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944275935794407e-05,
+      "loss": 1.1689,
+      "step": 1302
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944111590199088e-05,
+      "loss": 1.2251,
+      "step": 1303
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994394700328943e-05,
+      "loss": 1.1146,
+      "step": 1304
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994378217506942e-05,
+      "loss": 1.2152,
+      "step": 1305
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994361710554306e-05,
+      "loss": 1.2244,
+      "step": 1306
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943451794714354e-05,
+      "loss": 1.2198,
+      "step": 1307
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994328624258732e-05,
+      "loss": 1.4133,
+      "step": 1308
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943120449165963e-05,
+      "loss": 1.2817,
+      "step": 1309
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942954414454322e-05,
+      "loss": 1.1482,
+      "step": 1310
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942788138456418e-05,
+      "loss": 1.2434,
+      "step": 1311
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942621621176283e-05,
+      "loss": 1.3106,
+      "step": 1312
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942454862617962e-05,
+      "loss": 1.2204,
+      "step": 1313
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942287862785502e-05,
+      "loss": 1.239,
+      "step": 1314
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942120621682957e-05,
+      "loss": 1.2278,
+      "step": 1315
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994195313931438e-05,
+      "loss": 1.2348,
+      "step": 1316
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994178541568384e-05,
+      "loss": 1.2139,
+      "step": 1317
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941617450795406e-05,
+      "loss": 1.1293,
+      "step": 1318
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941449244653154e-05,
+      "loss": 1.0861,
+      "step": 1319
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941280797261163e-05,
+      "loss": 1.2474,
+      "step": 1320
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941112108623523e-05,
+      "loss": 1.2639,
+      "step": 1321
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940943178744333e-05,
+      "loss": 1.1696,
+      "step": 1322
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940774007627684e-05,
+      "loss": 1.218,
+      "step": 1323
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940604595277687e-05,
+      "loss": 1.2865,
+      "step": 1324
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940434941698447e-05,
+      "loss": 1.0606,
+      "step": 1325
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940265046894086e-05,
+      "loss": 1.0773,
+      "step": 1326
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940094910868727e-05,
+      "loss": 1.232,
+      "step": 1327
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939924533626492e-05,
+      "loss": 1.1547,
+      "step": 1328
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993975391517153e-05,
+      "loss": 1.1672,
+      "step": 1329
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939583055507964e-05,
+      "loss": 1.0783,
+      "step": 1330
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939411954639953e-05,
+      "loss": 1.1382,
+      "step": 1331
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939240612571642e-05,
+      "loss": 1.3196,
+      "step": 1332
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939069029307193e-05,
+      "loss": 1.3884,
+      "step": 1333
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993889720485077e-05,
+      "loss": 1.1807,
+      "step": 1334
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993872513920654e-05,
+      "loss": 1.2268,
+      "step": 1335
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993855283237868e-05,
+      "loss": 1.1647,
+      "step": 1336
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993838028437137e-05,
+      "loss": 1.3209,
+      "step": 1337
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938207495188797e-05,
+      "loss": 1.0788,
+      "step": 1338
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938034464835155e-05,
+      "loss": 1.1162,
+      "step": 1339
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937861193314648e-05,
+      "loss": 1.2678,
+      "step": 1340
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937687680631473e-05,
+      "loss": 1.3395,
+      "step": 1341
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937513926789843e-05,
+      "loss": 1.279,
+      "step": 1342
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937339931793975e-05,
+      "loss": 1.2522,
+      "step": 1343
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937165695648092e-05,
+      "loss": 1.2956,
+      "step": 1344
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993699121835642e-05,
+      "loss": 1.0166,
+      "step": 1345
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936816499923198e-05,
+      "loss": 1.1812,
+      "step": 1346
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936641540352665e-05,
+      "loss": 1.1623,
+      "step": 1347
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993646633964906e-05,
+      "loss": 1.2415,
+      "step": 1348
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993629089781664e-05,
+      "loss": 1.2402,
+      "step": 1349
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936115214859663e-05,
+      "loss": 1.3216,
+      "step": 1350
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9935939290782386e-05,
+      "loss": 1.0251,
+      "step": 1351
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935763125589086e-05,
+      "loss": 1.2363,
+      "step": 1352
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935586719284036e-05,
+      "loss": 1.2165,
+      "step": 1353
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935410071871514e-05,
+      "loss": 1.107,
+      "step": 1354
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993523318335581e-05,
+      "loss": 1.3687,
+      "step": 1355
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935056053741216e-05,
+      "loss": 1.1018,
+      "step": 1356
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934878683032028e-05,
+      "loss": 1.2965,
+      "step": 1357
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993470107123255e-05,
+      "loss": 1.1664,
+      "step": 1358
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934523218347096e-05,
+      "loss": 1.2753,
+      "step": 1359
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993434512437998e-05,
+      "loss": 1.3247,
+      "step": 1360
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934166789335526e-05,
+      "loss": 1.3067,
+      "step": 1361
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933988213218054e-05,
+      "loss": 1.1939,
+      "step": 1362
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933809396031908e-05,
+      "loss": 1.2985,
+      "step": 1363
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993363033778142e-05,
+      "loss": 1.1871,
+      "step": 1364
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933451038470936e-05,
+      "loss": 1.1556,
+      "step": 1365
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933271498104808e-05,
+      "loss": 1.3168,
+      "step": 1366
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933091716687397e-05,
+      "loss": 1.2999,
+      "step": 1367
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932911694223064e-05,
+      "loss": 1.2026,
+      "step": 1368
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993273143071617e-05,
+      "loss": 1.1705,
+      "step": 1369
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932550926171096e-05,
+      "loss": 1.1635,
+      "step": 1370
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932370180592226e-05,
+      "loss": 1.2966,
+      "step": 1371
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932189193983937e-05,
+      "loss": 1.212,
+      "step": 1372
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932007966350627e-05,
+      "loss": 1.0871,
+      "step": 1373
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931826497696694e-05,
+      "loss": 1.1762,
+      "step": 1374
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993164478802654e-05,
+      "loss": 1.2923,
+      "step": 1375
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931462837344578e-05,
+      "loss": 1.2277,
+      "step": 1376
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931280645655216e-05,
+      "loss": 1.1974,
+      "step": 1377
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993109821296288e-05,
+      "loss": 1.1312,
+      "step": 1378
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930915539271996e-05,
+      "loss": 1.2176,
+      "step": 1379
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930732624587e-05,
+      "loss": 1.3546,
+      "step": 1380
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930549468912326e-05,
+      "loss": 1.267,
+      "step": 1381
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930366072252424e-05,
+      "loss": 1.219,
+      "step": 1382
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930182434611736e-05,
+      "loss": 1.0165,
+      "step": 1383
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929998555994732e-05,
+      "loss": 1.2051,
+      "step": 1384
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992981443640586e-05,
+      "loss": 1.1443,
+      "step": 1385
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929630075849597e-05,
+      "loss": 1.3595,
+      "step": 1386
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929445474330413e-05,
+      "loss": 1.1942,
+      "step": 1387
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929260631852792e-05,
+      "loss": 1.3035,
+      "step": 1388
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992907554842121e-05,
+      "loss": 1.2347,
+      "step": 1389
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928890224040168e-05,
+      "loss": 1.1708,
+      "step": 1390
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992870465871416e-05,
+      "loss": 1.2822,
+      "step": 1391
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992851885244769e-05,
+      "loss": 1.228,
+      "step": 1392
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928332805245266e-05,
+      "loss": 1.1916,
+      "step": 1393
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928146517111404e-05,
+      "loss": 1.0364,
+      "step": 1394
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927959988050622e-05,
+      "loss": 1.047,
+      "step": 1395
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992777321806745e-05,
+      "loss": 1.1207,
+      "step": 1396
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927586207166417e-05,
+      "loss": 1.3755,
+      "step": 1397
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927398955352062e-05,
+      "loss": 1.1063,
+      "step": 1398
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992721146262893e-05,
+      "loss": 1.2314,
+      "step": 1399
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992702372900157e-05,
+      "loss": 1.2806,
+      "step": 1400
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926835754474543e-05,
+      "loss": 1.2412,
+      "step": 1401
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926647539052403e-05,
+      "loss": 1.0768,
+      "step": 1402
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926459082739717e-05,
+      "loss": 1.2475,
+      "step": 1403
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926270385541067e-05,
+      "loss": 1.2896,
+      "step": 1404
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926081447461025e-05,
+      "loss": 1.3261,
+      "step": 1405
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925892268504176e-05,
+      "loss": 1.2531,
+      "step": 1406
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925702848675117e-05,
+      "loss": 1.1269,
+      "step": 1407
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925513187978437e-05,
+      "loss": 1.151,
+      "step": 1408
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992532328641874e-05,
+      "loss": 1.2836,
+      "step": 1409
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925133144000643e-05,
+      "loss": 1.2588,
+      "step": 1410
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924942760728748e-05,
+      "loss": 1.2032,
+      "step": 1411
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992475213660768e-05,
+      "loss": 1.1095,
+      "step": 1412
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924561271642066e-05,
+      "loss": 1.3561,
+      "step": 1413
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992437016583654e-05,
+      "loss": 1.323,
+      "step": 1414
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924178819195732e-05,
+      "loss": 1.1382,
+      "step": 1415
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992398723172429e-05,
+      "loss": 1.1661,
+      "step": 1416
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923795403426865e-05,
+      "loss": 1.3431,
+      "step": 1417
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923603334308114e-05,
+      "loss": 1.076,
+      "step": 1418
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992341102437269e-05,
+      "loss": 1.2505,
+      "step": 1419
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923218473625264e-05,
+      "loss": 1.114,
+      "step": 1420
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992302568207051e-05,
+      "loss": 1.278,
+      "step": 1421
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922832649713108e-05,
+      "loss": 1.0404,
+      "step": 1422
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922639376557734e-05,
+      "loss": 1.2714,
+      "step": 1423
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922445862609088e-05,
+      "loss": 1.1653,
+      "step": 1424
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992225210787186e-05,
+      "loss": 1.1777,
+      "step": 1425
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922058112350754e-05,
+      "loss": 1.0792,
+      "step": 1426
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992186387605048e-05,
+      "loss": 1.0176,
+      "step": 1427
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921669398975745e-05,
+      "loss": 1.2038,
+      "step": 1428
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921474681131273e-05,
+      "loss": 1.2806,
+      "step": 1429
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992127972252179e-05,
+      "loss": 1.2927,
+      "step": 1430
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992108452315202e-05,
+      "loss": 1.0807,
+      "step": 1431
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920889083026716e-05,
+      "loss": 1.1187,
+      "step": 1432
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920693402150604e-05,
+      "loss": 1.2138,
+      "step": 1433
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992049748052844e-05,
+      "loss": 1.2749,
+      "step": 1434
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920301318164978e-05,
+      "loss": 1.2309,
+      "step": 1435
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920104915064974e-05,
+      "loss": 1.2342,
+      "step": 1436
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919908271233198e-05,
+      "loss": 1.2884,
+      "step": 1437
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919711386674425e-05,
+      "loss": 1.1106,
+      "step": 1438
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991951426139343e-05,
+      "loss": 1.3037,
+      "step": 1439
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919316895394993e-05,
+      "loss": 1.211,
+      "step": 1440
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919119288683908e-05,
+      "loss": 1.188,
+      "step": 1441
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918921441264966e-05,
+      "loss": 1.2106,
+      "step": 1442
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918723353142973e-05,
+      "loss": 1.2333,
+      "step": 1443
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918525024322738e-05,
+      "loss": 1.159,
+      "step": 1444
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918326454809066e-05,
+      "loss": 1.2541,
+      "step": 1445
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991812764460678e-05,
+      "loss": 1.2803,
+      "step": 1446
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917928593720705e-05,
+      "loss": 1.0533,
+      "step": 1447
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991772930215567e-05,
+      "loss": 1.2282,
+      "step": 1448
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917529769916513e-05,
+      "loss": 1.2763,
+      "step": 1449
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917329997008075e-05,
+      "loss": 1.2462,
+      "step": 1450
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99171299834352e-05,
+      "loss": 1.0816,
+      "step": 1451
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991692972920275e-05,
+      "loss": 1.195,
+      "step": 1452
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991672923431558e-05,
+      "loss": 1.2355,
+      "step": 1453
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916528498778554e-05,
+      "loss": 1.3125,
+      "step": 1454
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916327522596545e-05,
+      "loss": 1.2731,
+      "step": 1455
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916126305774427e-05,
+      "loss": 1.2261,
+      "step": 1456
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991592484831709e-05,
+      "loss": 1.2022,
+      "step": 1457
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915723150229417e-05,
+      "loss": 1.2555,
+      "step": 1458
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915521211516307e-05,
+      "loss": 1.0927,
+      "step": 1459
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915319032182655e-05,
+      "loss": 1.3106,
+      "step": 1460
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915116612233367e-05,
+      "loss": 1.1672,
+      "step": 1461
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991491395167336e-05,
+      "loss": 1.1799,
+      "step": 1462
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914711050507556e-05,
+      "loss": 1.1829,
+      "step": 1463
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991450790874087e-05,
+      "loss": 1.1287,
+      "step": 1464
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991430452637823e-05,
+      "loss": 1.2091,
+      "step": 1465
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991410090342458e-05,
+      "loss": 1.2158,
+      "step": 1466
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913897039884855e-05,
+      "loss": 1.0089,
+      "step": 1467
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913692935764006e-05,
+      "loss": 1.1175,
+      "step": 1468
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913488591066986e-05,
+      "loss": 1.2156,
+      "step": 1469
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991328400579875e-05,
+      "loss": 1.2486,
+      "step": 1470
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913079179964266e-05,
+      "loss": 1.2164,
+      "step": 1471
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912874113568503e-05,
+      "loss": 1.3371,
+      "step": 1472
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912668806616437e-05,
+      "loss": 1.186,
+      "step": 1473
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912463259113055e-05,
+      "loss": 1.2181,
+      "step": 1474
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912257471063338e-05,
+      "loss": 1.2534,
+      "step": 1475
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912051442472283e-05,
+      "loss": 1.183,
+      "step": 1476
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911845173344894e-05,
+      "loss": 1.1068,
+      "step": 1477
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991163866368617e-05,
+      "loss": 1.2747,
+      "step": 1478
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991143191350112e-05,
+      "loss": 1.1885,
+      "step": 1479
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991122492279477e-05,
+      "loss": 1.2612,
+      "step": 1480
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991101769157214e-05,
+      "loss": 1.1729,
+      "step": 1481
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910810219838257e-05,
+      "loss": 1.4029,
+      "step": 1482
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991060250759816e-05,
+      "loss": 1.1349,
+      "step": 1483
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991039455485688e-05,
+      "loss": 1.2365,
+      "step": 1484
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910186361619473e-05,
+      "loss": 1.1374,
+      "step": 1485
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909977927890988e-05,
+      "loss": 1.0521,
+      "step": 1486
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990976925367648e-05,
+      "loss": 1.294,
+      "step": 1487
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909560338981014e-05,
+      "loss": 1.2392,
+      "step": 1488
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990935118380967e-05,
+      "loss": 1.3442,
+      "step": 1489
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909141788167506e-05,
+      "loss": 1.1825,
+      "step": 1490
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908932152059618e-05,
+      "loss": 1.196,
+      "step": 1491
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908722275491084e-05,
+      "loss": 1.2665,
+      "step": 1492
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908512158467007e-05,
+      "loss": 1.1571,
+      "step": 1493
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908301800992475e-05,
+      "loss": 1.1374,
+      "step": 1494
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99080912030726e-05,
+      "loss": 1.136,
+      "step": 1495
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990788036471249e-05,
+      "loss": 1.3739,
+      "step": 1496
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990766928591726e-05,
+      "loss": 1.0597,
+      "step": 1497
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907457966692036e-05,
+      "loss": 1.2848,
+      "step": 1498
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990724640704194e-05,
+      "loss": 1.1383,
+      "step": 1499
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990703460697211e-05,
+      "loss": 1.3076,
+      "step": 1500
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990682256648769e-05,
+      "loss": 1.3344,
+      "step": 1501
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990661028559382e-05,
+      "loss": 1.2203,
+      "step": 1502
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990639776429565e-05,
+      "loss": 1.1949,
+      "step": 1503
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906185002598343e-05,
+      "loss": 1.0855,
+      "step": 1504
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905972000507057e-05,
+      "loss": 1.1892,
+      "step": 1505
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905758758026966e-05,
+      "loss": 1.1436,
+      "step": 1506
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905545275163235e-05,
+      "loss": 1.2022,
+      "step": 1507
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905331551921056e-05,
+      "loss": 1.2679,
+      "step": 1508
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905117588305612e-05,
+      "loss": 1.2692,
+      "step": 1509
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904903384322095e-05,
+      "loss": 1.0904,
+      "step": 1510
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904688939975697e-05,
+      "loss": 1.3013,
+      "step": 1511
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990447425527163e-05,
+      "loss": 1.2284,
+      "step": 1512
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99042593302151e-05,
+      "loss": 1.1082,
+      "step": 1513
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904044164811325e-05,
+      "loss": 1.1854,
+      "step": 1514
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903828759065524e-05,
+      "loss": 1.2563,
+      "step": 1515
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903613112982925e-05,
+      "loss": 1.3354,
+      "step": 1516
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990339722656876e-05,
+      "loss": 1.2577,
+      "step": 1517
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990318109982827e-05,
+      "loss": 1.2997,
+      "step": 1518
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902964732766702e-05,
+      "loss": 1.18,
+      "step": 1519
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99027481253893e-05,
+      "loss": 1.2377,
+      "step": 1520
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902531277701323e-05,
+      "loss": 1.2898,
+      "step": 1521
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902314189708037e-05,
+      "loss": 1.1776,
+      "step": 1522
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902096861414706e-05,
+      "loss": 1.1859,
+      "step": 1523
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901879292826604e-05,
+      "loss": 1.0938,
+      "step": 1524
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901661483949015e-05,
+      "loss": 1.1763,
+      "step": 1525
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990144343478722e-05,
+      "loss": 1.2965,
+      "step": 1526
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990122514534651e-05,
+      "loss": 1.2198,
+      "step": 1527
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901006615632187e-05,
+      "loss": 1.3245,
+      "step": 1528
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900787845649548e-05,
+      "loss": 1.3404,
+      "step": 1529
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990056883540391e-05,
+      "loss": 1.2712,
+      "step": 1530
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990034958490058e-05,
+      "loss": 1.2181,
+      "step": 1531
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990013009414488e-05,
+      "loss": 1.1637,
+      "step": 1532
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989991036314214e-05,
+      "loss": 0.95,
+      "step": 1533
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899690391897694e-05,
+      "loss": 1.2688,
+      "step": 1534
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989947018041687e-05,
+      "loss": 1.2269,
+      "step": 1535
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899249728705018e-05,
+      "loss": 1.2608,
+      "step": 1536
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989902903676749e-05,
+      "loss": 1.2371,
+      "step": 1537
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898808104609638e-05,
+      "loss": 1.3357,
+      "step": 1538
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898586932236826e-05,
+      "loss": 1.1653,
+      "step": 1539
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989836551965442e-05,
+      "loss": 1.2771,
+      "step": 1540
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898143866867792e-05,
+      "loss": 1.3155,
+      "step": 1541
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989792197388232e-05,
+      "loss": 1.2605,
+      "step": 1542
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897699840703393e-05,
+      "loss": 1.1614,
+      "step": 1543
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.98974774673364e-05,
+      "loss": 1.3289,
+      "step": 1544
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897254853786735e-05,
+      "loss": 1.3035,
+      "step": 1545
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.98970320000598e-05,
+      "loss": 1.2095,
+      "step": 1546
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896808906161005e-05,
+      "loss": 1.3592,
+      "step": 1547
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896585572095764e-05,
+      "loss": 1.2597,
+      "step": 1548
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896361997869496e-05,
+      "loss": 1.2318,
+      "step": 1549
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896138183487626e-05,
+      "loss": 1.3372,
+      "step": 1550
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895914128955588e-05,
+      "loss": 1.2742,
+      "step": 1551
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895689834278813e-05,
+      "loss": 1.1867,
+      "step": 1552
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989546529946275e-05,
+      "loss": 1.2649,
+      "step": 1553
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895240524512845e-05,
+      "loss": 1.2354,
+      "step": 1554
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895015509434555e-05,
+      "loss": 1.1658,
+      "step": 1555
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894790254233338e-05,
+      "loss": 1.1695,
+      "step": 1556
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894564758914662e-05,
+      "loss": 1.2504,
+      "step": 1557
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894339023484e-05,
+      "loss": 1.2027,
+      "step": 1558
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989411304794682e-05,
+      "loss": 1.2842,
+      "step": 1559
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989388683230862e-05,
+      "loss": 1.2107,
+      "step": 1560
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893660376574883e-05,
+      "loss": 1.1256,
+      "step": 1561
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893433680751105e-05,
+      "loss": 1.1814,
+      "step": 1562
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893206744842787e-05,
+      "loss": 1.2675,
+      "step": 1563
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892979568855435e-05,
+      "loss": 1.1783,
+      "step": 1564
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892752152794565e-05,
+      "loss": 1.1652,
+      "step": 1565
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892524496665692e-05,
+      "loss": 1.2373,
+      "step": 1566
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989229660047434e-05,
+      "loss": 1.2151,
+      "step": 1567
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892068464226044e-05,
+      "loss": 1.0836,
+      "step": 1568
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989184008792634e-05,
+      "loss": 1.2738,
+      "step": 1569
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891611471580767e-05,
+      "loss": 1.1579,
+      "step": 1570
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989138261519487e-05,
+      "loss": 1.1891,
+      "step": 1571
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989115351877421e-05,
+      "loss": 1.3115,
+      "step": 1572
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890924182324345e-05,
+      "loss": 1.292,
+      "step": 1573
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989069460585083e-05,
+      "loss": 1.2118,
+      "step": 1574
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890464789359253e-05,
+      "loss": 1.2522,
+      "step": 1575
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989023473285518e-05,
+      "loss": 1.2686,
+      "step": 1576
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890004436344197e-05,
+      "loss": 1.0832,
+      "step": 1577
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988977389983189e-05,
+      "loss": 1.3374,
+      "step": 1578
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889543123323854e-05,
+      "loss": 1.1627,
+      "step": 1579
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889312106825694e-05,
+      "loss": 1.2428,
+      "step": 1580
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988908085034301e-05,
+      "loss": 1.2229,
+      "step": 1581
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988884935388142e-05,
+      "loss": 1.3273,
+      "step": 1582
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988861761744653e-05,
+      "loss": 1.1028,
+      "step": 1583
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988838564104398e-05,
+      "loss": 1.2802,
+      "step": 1584
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888153424679387e-05,
+      "loss": 1.172,
+      "step": 1585
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887920968358394e-05,
+      "loss": 1.1951,
+      "step": 1586
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887688272086637e-05,
+      "loss": 1.1848,
+      "step": 1587
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887455335869762e-05,
+      "loss": 1.0306,
+      "step": 1588
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887222159713427e-05,
+      "loss": 1.2194,
+      "step": 1589
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886988743623284e-05,
+      "loss": 1.169,
+      "step": 1590
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886755087605004e-05,
+      "loss": 1.0504,
+      "step": 1591
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886521191664255e-05,
+      "loss": 1.2027,
+      "step": 1592
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988628705580671e-05,
+      "loss": 1.2665,
+      "step": 1593
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886052680038048e-05,
+      "loss": 1.3467,
+      "step": 1594
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885818064363968e-05,
+      "loss": 1.2991,
+      "step": 1595
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885583208790154e-05,
+      "loss": 1.1263,
+      "step": 1596
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988534811332231e-05,
+      "loss": 1.2963,
+      "step": 1597
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885112777966135e-05,
+      "loss": 1.1377,
+      "step": 1598
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884877202727345e-05,
+      "loss": 1.2506,
+      "step": 1599
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988464138761166e-05,
+      "loss": 1.2749,
+      "step": 1600
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884405332624793e-05,
+      "loss": 1.2695,
+      "step": 1601
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988416903777248e-05,
+      "loss": 1.1747,
+      "step": 1602
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883932503060452e-05,
+      "loss": 1.2673,
+      "step": 1603
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988369572849445e-05,
+      "loss": 1.1353,
+      "step": 1604
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883458714080222e-05,
+      "loss": 1.15,
+      "step": 1605
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883221459823515e-05,
+      "loss": 1.1052,
+      "step": 1606
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882983965730086e-05,
+      "loss": 1.1832,
+      "step": 1607
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882746231805705e-05,
+      "loss": 1.3124,
+      "step": 1608
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882508258056136e-05,
+      "loss": 1.166,
+      "step": 1609
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882270044487155e-05,
+      "loss": 1.196,
+      "step": 1610
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882031591104543e-05,
+      "loss": 1.1879,
+      "step": 1611
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881792897914086e-05,
+      "loss": 1.2686,
+      "step": 1612
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881553964921574e-05,
+      "loss": 1.2981,
+      "step": 1613
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881314792132812e-05,
+      "loss": 1.1324,
+      "step": 1614
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881075379553597e-05,
+      "loss": 1.2432,
+      "step": 1615
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880835727189742e-05,
+      "loss": 1.2153,
+      "step": 1616
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880595835047062e-05,
+      "loss": 1.2931,
+      "step": 1617
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988035570313138e-05,
+      "loss": 1.3875,
+      "step": 1618
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880115331448526e-05,
+      "loss": 1.3227,
+      "step": 1619
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879874720004326e-05,
+      "loss": 1.2833,
+      "step": 1620
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879633868804624e-05,
+      "loss": 1.1764,
+      "step": 1621
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879392777855258e-05,
+      "loss": 1.2147,
+      "step": 1622
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879151447162086e-05,
+      "loss": 1.2515,
+      "step": 1623
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987890987673096e-05,
+      "loss": 1.3057,
+      "step": 1624
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987866806656775e-05,
+      "loss": 1.2632,
+      "step": 1625
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878426016678313e-05,
+      "loss": 1.1165,
+      "step": 1626
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987818372706853e-05,
+      "loss": 1.2562,
+      "step": 1627
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877941197744277e-05,
+      "loss": 1.1912,
+      "step": 1628
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877698428711444e-05,
+      "loss": 1.2604,
+      "step": 1629
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877455419975917e-05,
+      "loss": 1.0693,
+      "step": 1630
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877212171543595e-05,
+      "loss": 1.1734,
+      "step": 1631
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876968683420384e-05,
+      "loss": 1.1465,
+      "step": 1632
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876724955612188e-05,
+      "loss": 1.15,
+      "step": 1633
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876480988124923e-05,
+      "loss": 1.1678,
+      "step": 1634
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876236780964513e-05,
+      "loss": 1.1954,
+      "step": 1635
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875992334136878e-05,
+      "loss": 1.1741,
+      "step": 1636
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875747647647956e-05,
+      "loss": 1.2723,
+      "step": 1637
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875502721503683e-05,
+      "loss": 1.091,
+      "step": 1638
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987525755571e-05,
+      "loss": 1.2227,
+      "step": 1639
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987501215027286e-05,
+      "loss": 1.2618,
+      "step": 1640
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9874766505198214e-05,
+      "loss": 1.3179,
+      "step": 1641
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9874520620492026e-05,
+      "loss": 1.1194,
+      "step": 1642
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987427449616026e-05,
+      "loss": 1.2671,
+      "step": 1643
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9874028132208897e-05,
+      "loss": 0.982,
+      "step": 1644
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873781528643905e-05,
+      "loss": 1.2362,
+      "step": 1645
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873534685471277e-05,
+      "loss": 1.1581,
+      "step": 1646
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873287602696996e-05,
+      "loss": 1.2399,
+      "step": 1647
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873040280327062e-05,
+      "loss": 1.2516,
+      "step": 1648
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872792718367476e-05,
+      "loss": 1.013,
+      "step": 1649
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872544916824244e-05,
+      "loss": 1.1628,
+      "step": 1650
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987229687570338e-05,
+      "loss": 1.2305,
+      "step": 1651
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987204859501091e-05,
+      "loss": 1.2515,
+      "step": 1652
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871800074752848e-05,
+      "loss": 1.3653,
+      "step": 1653
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871551314935233e-05,
+      "loss": 1.104,
+      "step": 1654
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98713023155641e-05,
+      "loss": 1.0482,
+      "step": 1655
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987105307664549e-05,
+      "loss": 1.3093,
+      "step": 1656
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870803598185446e-05,
+      "loss": 1.2885,
+      "step": 1657
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870553880190032e-05,
+      "loss": 1.2178,
+      "step": 1658
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870303922665305e-05,
+      "loss": 1.2466,
+      "step": 1659
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870053725617326e-05,
+      "loss": 1.3232,
+      "step": 1660
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986980328905217e-05,
+      "loss": 1.2206,
+      "step": 1661
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869552612975917e-05,
+      "loss": 0.9897,
+      "step": 1662
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869301697394646e-05,
+      "loss": 1.2179,
+      "step": 1663
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869050542314446e-05,
+      "loss": 1.181,
+      "step": 1664
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9868799147741417e-05,
+      "loss": 1.2234,
+      "step": 1665
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986854751368165e-05,
+      "loss": 1.2891,
+      "step": 1666
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986829564014126e-05,
+      "loss": 1.2442,
+      "step": 1667
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9868043527126358e-05,
+      "loss": 1.0578,
+      "step": 1668
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867791174643057e-05,
+      "loss": 1.2294,
+      "step": 1669
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986753858269749e-05,
+      "loss": 1.1264,
+      "step": 1670
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986728575129578e-05,
+      "loss": 1.1827,
+      "step": 1671
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986703268044406e-05,
+      "loss": 1.1649,
+      "step": 1672
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866779370148475e-05,
+      "loss": 1.1388,
+      "step": 1673
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866525820415173e-05,
+      "loss": 1.2832,
+      "step": 1674
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98662720312503e-05,
+      "loss": 1.2433,
+      "step": 1675
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866018002660027e-05,
+      "loss": 1.1563,
+      "step": 1676
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865763734650514e-05,
+      "loss": 1.2449,
+      "step": 1677
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865509227227924e-05,
+      "loss": 1.3671,
+      "step": 1678
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986525448039844e-05,
+      "loss": 1.3715,
+      "step": 1679
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9864999494168245e-05,
+      "loss": 1.1815,
+      "step": 1680
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9864744268543522e-05,
+      "loss": 1.2023,
+      "step": 1681
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9864488803530467e-05,
+      "loss": 1.2284,
+      "step": 1682
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9864233099135278e-05,
+      "loss": 1.3258,
+      "step": 1683
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863977155364164e-05,
+      "loss": 1.2123,
+      "step": 1684
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986372097222333e-05,
+      "loss": 1.3644,
+      "step": 1685
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863464549719e-05,
+      "loss": 1.134,
+      "step": 1686
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986320788785739e-05,
+      "loss": 1.1198,
+      "step": 1687
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862950986644732e-05,
+      "loss": 1.1581,
+      "step": 1688
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862693846087258e-05,
+      "loss": 1.3675,
+      "step": 1689
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986243646619121e-05,
+      "loss": 1.181,
+      "step": 1690
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862178846962837e-05,
+      "loss": 1.2126,
+      "step": 1691
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861920988408382e-05,
+      "loss": 1.1949,
+      "step": 1692
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986166289053411e-05,
+      "loss": 1.2965,
+      "step": 1693
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861404553346282e-05,
+      "loss": 1.1585,
+      "step": 1694
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861145976851167e-05,
+      "loss": 1.2869,
+      "step": 1695
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986088716105504e-05,
+      "loss": 1.2601,
+      "step": 1696
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986062810596418e-05,
+      "loss": 1.2804,
+      "step": 1697
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860368811584875e-05,
+      "loss": 1.168,
+      "step": 1698
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860109277923417e-05,
+      "loss": 1.196,
+      "step": 1699
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859849504986105e-05,
+      "loss": 1.2003,
+      "step": 1700
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859589492779244e-05,
+      "loss": 1.2612,
+      "step": 1701
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985932924130914e-05,
+      "loss": 1.3124,
+      "step": 1702
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859068750582112e-05,
+      "loss": 1.1859,
+      "step": 1703
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985880802060448e-05,
+      "loss": 1.1997,
+      "step": 1704
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858547051382565e-05,
+      "loss": 1.0901,
+      "step": 1705
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858285842922715e-05,
+      "loss": 1.138,
+      "step": 1706
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858024395231256e-05,
+      "loss": 1.1474,
+      "step": 1707
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857762708314535e-05,
+      "loss": 1.2398,
+      "step": 1708
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857500782178905e-05,
+      "loss": 1.3717,
+      "step": 1709
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857238616830718e-05,
+      "loss": 1.256,
+      "step": 1710
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856976212276344e-05,
+      "loss": 1.1847,
+      "step": 1711
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856713568522143e-05,
+      "loss": 1.004,
+      "step": 1712
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985645068557449e-05,
+      "loss": 1.2501,
+      "step": 1713
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985618756343977e-05,
+      "loss": 1.1148,
+      "step": 1714
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855924202124358e-05,
+      "loss": 1.2643,
+      "step": 1715
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855660601634656e-05,
+      "loss": 1.2342,
+      "step": 1716
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855396761977052e-05,
+      "loss": 1.1747,
+      "step": 1717
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985513268315795e-05,
+      "loss": 1.1905,
+      "step": 1718
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985486836518377e-05,
+      "loss": 1.1336,
+      "step": 1719
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854603808060907e-05,
+      "loss": 1.2266,
+      "step": 1720
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854339011795795e-05,
+      "loss": 1.182,
+      "step": 1721
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854073976394858e-05,
+      "loss": 1.0497,
+      "step": 1722
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853808701864522e-05,
+      "loss": 1.324,
+      "step": 1723
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985354318821123e-05,
+      "loss": 1.3321,
+      "step": 1724
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853277435441422e-05,
+      "loss": 1.3032,
+      "step": 1725
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985301144356155e-05,
+      "loss": 1.3112,
+      "step": 1726
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852745212578063e-05,
+      "loss": 1.2029,
+      "step": 1727
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852478742497426e-05,
+      "loss": 1.3641,
+      "step": 1728
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985221203332611e-05,
+      "loss": 1.2457,
+      "step": 1729
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985194508507058e-05,
+      "loss": 1.1046,
+      "step": 1730
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9851677897737314e-05,
+      "loss": 1.2157,
+      "step": 1731
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98514104713328e-05,
+      "loss": 1.2768,
+      "step": 1732
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9851142805863523e-05,
+      "loss": 1.205,
+      "step": 1733
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850874901335984e-05,
+      "loss": 1.1879,
+      "step": 1734
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850606757756683e-05,
+      "loss": 1.2406,
+      "step": 1735
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850338375132125e-05,
+      "loss": 1.2385,
+      "step": 1736
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985006975346882e-05,
+      "loss": 1.1465,
+      "step": 1737
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849800892773293e-05,
+      "loss": 1.2532,
+      "step": 1738
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849531793052064e-05,
+      "loss": 1.2353,
+      "step": 1739
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849262454311663e-05,
+      "loss": 1.3306,
+      "step": 1740
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848992876558633e-05,
+      "loss": 1.1564,
+      "step": 1741
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848723059799508e-05,
+      "loss": 1.2329,
+      "step": 1742
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848453004040838e-05,
+      "loss": 1.1927,
+      "step": 1743
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984818270928918e-05,
+      "loss": 1.2122,
+      "step": 1744
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847912175551085e-05,
+      "loss": 1.3066,
+      "step": 1745
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847641402833127e-05,
+      "loss": 1.1923,
+      "step": 1746
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847370391141872e-05,
+      "loss": 1.3465,
+      "step": 1747
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847099140483896e-05,
+      "loss": 1.1659,
+      "step": 1748
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846827650865787e-05,
+      "loss": 1.2248,
+      "step": 1749
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846555922294123e-05,
+      "loss": 1.2832,
+      "step": 1750
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984628395477551e-05,
+      "loss": 1.3811,
+      "step": 1751
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984601174831654e-05,
+      "loss": 1.1193,
+      "step": 1752
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845739302923822e-05,
+      "loss": 1.1855,
+      "step": 1753
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984546661860397e-05,
+      "loss": 1.0899,
+      "step": 1754
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845193695363592e-05,
+      "loss": 1.1582,
+      "step": 1755
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984492053320932e-05,
+      "loss": 1.3366,
+      "step": 1756
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844647132147773e-05,
+      "loss": 1.3697,
+      "step": 1757
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98443734921856e-05,
+      "loss": 1.1838,
+      "step": 1758
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984409961332943e-05,
+      "loss": 1.175,
+      "step": 1759
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843825495585912e-05,
+      "loss": 1.1218,
+      "step": 1760
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98435511389617e-05,
+      "loss": 1.2601,
+      "step": 1761
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843276543463452e-05,
+      "loss": 1.2547,
+      "step": 1762
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843001709097832e-05,
+      "loss": 1.3804,
+      "step": 1763
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842726635871503e-05,
+      "loss": 1.2134,
+      "step": 1764
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842451323791146e-05,
+      "loss": 1.2876,
+      "step": 1765
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842175772863442e-05,
+      "loss": 1.1707,
+      "step": 1766
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984189998309508e-05,
+      "loss": 1.1266,
+      "step": 1767
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9841623954492744e-05,
+      "loss": 1.2684,
+      "step": 1768
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.984134768706314e-05,
+      "loss": 1.0313,
+      "step": 1769
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9841071180812972e-05,
+      "loss": 1.3077,
+      "step": 1770
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840794435748946e-05,
+      "loss": 1.2421,
+      "step": 1771
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.984051745187778e-05,
+      "loss": 1.3594,
+      "step": 1772
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840240229206194e-05,
+      "loss": 1.2438,
+      "step": 1773
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839962767740918e-05,
+      "loss": 1.33,
+      "step": 1774
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839685067488683e-05,
+      "loss": 1.1987,
+      "step": 1775
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983940712845623e-05,
+      "loss": 1.0079,
+      "step": 1776
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839128950650302e-05,
+      "loss": 1.2513,
+      "step": 1777
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838850534077652e-05,
+      "loss": 1.2896,
+      "step": 1778
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983857187874503e-05,
+      "loss": 1.1317,
+      "step": 1779
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838292984659207e-05,
+      "loss": 1.2553,
+      "step": 1780
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838013851826948e-05,
+      "loss": 1.0026,
+      "step": 1781
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983773448025502e-05,
+      "loss": 1.1217,
+      "step": 1782
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983745486995021e-05,
+      "loss": 1.1569,
+      "step": 1783
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837175020919304e-05,
+      "loss": 1.2556,
+      "step": 1784
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983689493316909e-05,
+      "loss": 1.3112,
+      "step": 1785
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983661460670636e-05,
+      "loss": 1.1988,
+      "step": 1786
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836334041537928e-05,
+      "loss": 1.2588,
+      "step": 1787
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836053237670594e-05,
+      "loss": 1.0768,
+      "step": 1788
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835772195111174e-05,
+      "loss": 1.1653,
+      "step": 1789
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835490913866492e-05,
+      "loss": 1.0849,
+      "step": 1790
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835209393943366e-05,
+      "loss": 1.1029,
+      "step": 1791
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834927635348635e-05,
+      "loss": 1.2332,
+      "step": 1792
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983464563808913e-05,
+      "loss": 1.3446,
+      "step": 1793
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834363402171704e-05,
+      "loss": 1.1353,
+      "step": 1794
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834080927603198e-05,
+      "loss": 1.1411,
+      "step": 1795
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833798214390467e-05,
+      "loss": 1.253,
+      "step": 1796
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833515262540376e-05,
+      "loss": 1.3454,
+      "step": 1797
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833232072059786e-05,
+      "loss": 1.251,
+      "step": 1798
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832948642955574e-05,
+      "loss": 1.2811,
+      "step": 1799
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983266497523462e-05,
+      "loss": 1.1023,
+      "step": 1800
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832381068903797e-05,
+      "loss": 1.2434,
+      "step": 1801
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832096923970002e-05,
+      "loss": 1.1331,
+      "step": 1802
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983181254044013e-05,
+      "loss": 1.2834,
+      "step": 1803
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983152791832109e-05,
+      "loss": 1.1814,
+      "step": 1804
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9831243057619774e-05,
+      "loss": 1.1303,
+      "step": 1805
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98309579583431e-05,
+      "loss": 1.2768,
+      "step": 1806
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983067262049799e-05,
+      "loss": 1.314,
+      "step": 1807
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983038704409137e-05,
+      "loss": 1.1085,
+      "step": 1808
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830101229130162e-05,
+      "loss": 1.295,
+      "step": 1809
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9829815175621306e-05,
+      "loss": 1.0677,
+      "step": 1810
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9829528883571747e-05,
+      "loss": 1.1126,
+      "step": 1811
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982924235298843e-05,
+      "loss": 1.2433,
+      "step": 1812
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828955583878306e-05,
+      "loss": 1.4026,
+      "step": 1813
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828668576248336e-05,
+      "loss": 1.073,
+      "step": 1814
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828381330105487e-05,
+      "loss": 1.2981,
+      "step": 1815
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828093845456726e-05,
+      "loss": 1.1695,
+      "step": 1816
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982780612230903e-05,
+      "loss": 1.2934,
+      "step": 1817
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827518160669382e-05,
+      "loss": 1.3327,
+      "step": 1818
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827229960544773e-05,
+      "loss": 1.1952,
+      "step": 1819
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826941521942187e-05,
+      "loss": 1.1726,
+      "step": 1820
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826652844868636e-05,
+      "loss": 1.2736,
+      "step": 1821
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826363929331118e-05,
+      "loss": 1.1258,
+      "step": 1822
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826074775336645e-05,
+      "loss": 1.1393,
+      "step": 1823
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825785382892236e-05,
+      "loss": 1.0825,
+      "step": 1824
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825495752004912e-05,
+      "loss": 1.134,
+      "step": 1825
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98252058826817e-05,
+      "loss": 1.2842,
+      "step": 1826
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982491577492964e-05,
+      "loss": 1.1826,
+      "step": 1827
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982462542875576e-05,
+      "loss": 1.215,
+      "step": 1828
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9824334844167122e-05,
+      "loss": 1.216,
+      "step": 1829
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982404402117077e-05,
+      "loss": 1.1504,
+      "step": 1830
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823752959773758e-05,
+      "loss": 0.9752,
+      "step": 1831
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823461659983154e-05,
+      "loss": 1.2695,
+      "step": 1832
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823170121806023e-05,
+      "loss": 1.227,
+      "step": 1833
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822878345249443e-05,
+      "loss": 1.2705,
+      "step": 1834
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822586330320492e-05,
+      "loss": 1.197,
+      "step": 1835
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982229407702626e-05,
+      "loss": 1.2759,
+      "step": 1836
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822001585373835e-05,
+      "loss": 1.2143,
+      "step": 1837
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821708855370322e-05,
+      "loss": 1.4197,
+      "step": 1838
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821415887022813e-05,
+      "loss": 1.2465,
+      "step": 1839
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821122680338428e-05,
+      "loss": 1.1533,
+      "step": 1840
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982082923532428e-05,
+      "loss": 1.1961,
+      "step": 1841
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820535551987486e-05,
+      "loss": 1.1955,
+      "step": 1842
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820241630335176e-05,
+      "loss": 1.1124,
+      "step": 1843
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9819947470374484e-05,
+      "loss": 1.2007,
+      "step": 1844
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981965307211254e-05,
+      "loss": 1.249,
+      "step": 1845
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98193584355565e-05,
+      "loss": 1.1176,
+      "step": 1846
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9819063560713507e-05,
+      "loss": 1.183,
+      "step": 1847
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981876844759072e-05,
+      "loss": 1.302,
+      "step": 1848
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9818473096195295e-05,
+      "loss": 1.1018,
+      "step": 1849
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9818177506534406e-05,
+      "loss": 1.2842,
+      "step": 1850
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981788167861522e-05,
+      "loss": 1.2912,
+      "step": 1851
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817585612444923e-05,
+      "loss": 1.2346,
+      "step": 1852
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817289308030695e-05,
+      "loss": 1.301,
+      "step": 1853
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9816992765379725e-05,
+      "loss": 1.191,
+      "step": 1854
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981669598449921e-05,
+      "loss": 0.9876,
+      "step": 1855
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981639896539636e-05,
+      "loss": 1.0499,
+      "step": 1856
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981610170807837e-05,
+      "loss": 1.2254,
+      "step": 1857
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981580421255246e-05,
+      "loss": 1.1503,
+      "step": 1858
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981550647882585e-05,
+      "loss": 1.1683,
+      "step": 1859
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9815208506905764e-05,
+      "loss": 1.2219,
+      "step": 1860
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814910296799436e-05,
+      "loss": 1.0759,
+      "step": 1861
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814611848514095e-05,
+      "loss": 1.1692,
+      "step": 1862
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814313162056993e-05,
+      "loss": 1.2189,
+      "step": 1863
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981401423743537e-05,
+      "loss": 1.3463,
+      "step": 1864
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813715074656482e-05,
+      "loss": 1.1657,
+      "step": 1865
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813415673727594e-05,
+      "loss": 1.0866,
+      "step": 1866
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813116034655966e-05,
+      "loss": 1.2247,
+      "step": 1867
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812816157448874e-05,
+      "loss": 1.0849,
+      "step": 1868
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812516042113588e-05,
+      "loss": 1.0852,
+      "step": 1869
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812215688657397e-05,
+      "loss": 1.2083,
+      "step": 1870
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811915097087587e-05,
+      "loss": 1.133,
+      "step": 1871
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811614267411453e-05,
+      "loss": 1.3001,
+      "step": 1872
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98113131996363e-05,
+      "loss": 1.2867,
+      "step": 1873
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811011893769424e-05,
+      "loss": 1.1203,
+      "step": 1874
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810710349818147e-05,
+      "loss": 1.0894,
+      "step": 1875
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810408567789777e-05,
+      "loss": 1.263,
+      "step": 1876
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810106547691648e-05,
+      "loss": 1.1429,
+      "step": 1877
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980980428953108e-05,
+      "loss": 1.1768,
+      "step": 1878
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809501793315412e-05,
+      "loss": 1.1893,
+      "step": 1879
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809199059051987e-05,
+      "loss": 1.0988,
+      "step": 1880
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9808896086748146e-05,
+      "loss": 1.2256,
+      "step": 1881
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9808592876411245e-05,
+      "loss": 1.081,
+      "step": 1882
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980828942804864e-05,
+      "loss": 1.2198,
+      "step": 1883
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807985741667696e-05,
+      "loss": 1.1488,
+      "step": 1884
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807681817275783e-05,
+      "loss": 1.0687,
+      "step": 1885
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807377654880274e-05,
+      "loss": 1.2593,
+      "step": 1886
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807073254488554e-05,
+      "loss": 1.1925,
+      "step": 1887
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806768616108006e-05,
+      "loss": 1.1906,
+      "step": 1888
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806463739746026e-05,
+      "loss": 1.2227,
+      "step": 1889
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806158625410008e-05,
+      "loss": 1.0745,
+      "step": 1890
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980585327310736e-05,
+      "loss": 1.2953,
+      "step": 1891
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9805547682845494e-05,
+      "loss": 1.1361,
+      "step": 1892
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980524185463182e-05,
+      "loss": 1.1478,
+      "step": 1893
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804935788473765e-05,
+      "loss": 1.2305,
+      "step": 1894
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980462948437875e-05,
+      "loss": 1.3097,
+      "step": 1895
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804322942354216e-05,
+      "loss": 1.1601,
+      "step": 1896
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804016162407593e-05,
+      "loss": 1.1979,
+      "step": 1897
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980370914454633e-05,
+      "loss": 1.3045,
+      "step": 1898
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9803401888777882e-05,
+      "loss": 1.2409,
+      "step": 1899
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98030943951097e-05,
+      "loss": 1.2714,
+      "step": 1900
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802786663549244e-05,
+      "loss": 1.2811,
+      "step": 1901
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802478694103987e-05,
+      "loss": 1.1733,
+      "step": 1902
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98021704867814e-05,
+      "loss": 1.0727,
+      "step": 1903
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9801862041588966e-05,
+      "loss": 1.2867,
+      "step": 1904
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9801553358534157e-05,
+      "loss": 1.021,
+      "step": 1905
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980124443762448e-05,
+      "loss": 1.2973,
+      "step": 1906
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800935278867425e-05,
+      "loss": 1.1674,
+      "step": 1907
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800625882270493e-05,
+      "loss": 1.2352,
+      "step": 1908
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800316247841194e-05,
+      "loss": 1.1605,
+      "step": 1909
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800006375587043e-05,
+      "loss": 1.053,
+      "step": 1910
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9799696265515554e-05,
+      "loss": 1.244,
+      "step": 1911
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9799385917634256e-05,
+      "loss": 1.287,
+      "step": 1912
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9799075331950687e-05,
+      "loss": 1.0834,
+      "step": 1913
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798764508472373e-05,
+      "loss": 1.0535,
+      "step": 1914
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798453447206862e-05,
+      "loss": 1.3242,
+      "step": 1915
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798142148161703e-05,
+      "loss": 1.1676,
+      "step": 1916
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797830611344448e-05,
+      "loss": 1.2646,
+      "step": 1917
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979751883676266e-05,
+      "loss": 1.2133,
+      "step": 1918
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797206824423904e-05,
+      "loss": 1.2284,
+      "step": 1919
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979689457433575e-05,
+      "loss": 1.2465,
+      "step": 1920
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796582086505774e-05,
+      "loss": 1.1882,
+      "step": 1921
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796269360941563e-05,
+      "loss": 1.0413,
+      "step": 1922
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795956397650708e-05,
+      "loss": 1.1581,
+      "step": 1923
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795643196640794e-05,
+      "loss": 1.1832,
+      "step": 1924
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795329757919433e-05,
+      "loss": 0.9952,
+      "step": 1925
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795016081494222e-05,
+      "loss": 1.1734,
+      "step": 1926
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9794702167372776e-05,
+      "loss": 1.2008,
+      "step": 1927
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9794388015562718e-05,
+      "loss": 1.2222,
+      "step": 1928
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9794073626071664e-05,
+      "loss": 1.1711,
+      "step": 1929
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793758998907248e-05,
+      "loss": 1.099,
+      "step": 1930
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793444134077102e-05,
+      "loss": 1.1953,
+      "step": 1931
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979312903158887e-05,
+      "loss": 1.0283,
+      "step": 1932
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792813691450193e-05,
+      "loss": 1.2382,
+      "step": 1933
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792498113668733e-05,
+      "loss": 1.2965,
+      "step": 1934
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979218229825214e-05,
+      "loss": 1.2874,
+      "step": 1935
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979186624520808e-05,
+      "loss": 1.0828,
+      "step": 1936
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791549954544226e-05,
+      "loss": 1.2108,
+      "step": 1937
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791233426268245e-05,
+      "loss": 1.3084,
+      "step": 1938
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790916660387825e-05,
+      "loss": 1.0069,
+      "step": 1939
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790599656910653e-05,
+      "loss": 1.2982,
+      "step": 1940
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790282415844425e-05,
+      "loss": 1.067,
+      "step": 1941
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978996493719683e-05,
+      "loss": 1.27,
+      "step": 1942
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9789647220975578e-05,
+      "loss": 1.0996,
+      "step": 1943
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9789329267188378e-05,
+      "loss": 1.1437,
+      "step": 1944
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9789011075842947e-05,
+      "loss": 1.2079,
+      "step": 1945
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788692646947e-05,
+      "loss": 1.2565,
+      "step": 1946
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978837398050828e-05,
+      "loss": 1.2716,
+      "step": 1947
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788055076534504e-05,
+      "loss": 1.2844,
+      "step": 1948
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9787735935033415e-05,
+      "loss": 1.2876,
+      "step": 1949
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9787416556012766e-05,
+      "loss": 1.2749,
+      "step": 1950
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9787096939480296e-05,
+      "loss": 1.1175,
+      "step": 1951
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786777085443766e-05,
+      "loss": 1.2826,
+      "step": 1952
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786456993910938e-05,
+      "loss": 1.0602,
+      "step": 1953
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786136664889583e-05,
+      "loss": 1.2121,
+      "step": 1954
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9785816098387468e-05,
+      "loss": 1.2374,
+      "step": 1955
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9785495294412378e-05,
+      "loss": 1.1938,
+      "step": 1956
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9785174252972092e-05,
+      "loss": 1.2248,
+      "step": 1957
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784852974074408e-05,
+      "loss": 1.1231,
+      "step": 1958
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784531457727116e-05,
+      "loss": 1.2697,
+      "step": 1959
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978420970393802e-05,
+      "loss": 1.3699,
+      "step": 1960
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978388771271493e-05,
+      "loss": 1.1317,
+      "step": 1961
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9783565484065657e-05,
+      "loss": 1.1718,
+      "step": 1962
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978324301799803e-05,
+      "loss": 1.1891,
+      "step": 1963
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978292031451986e-05,
+      "loss": 1.242,
+      "step": 1964
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978259737363898e-05,
+      "loss": 1.2015,
+      "step": 1965
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9782274195363237e-05,
+      "loss": 1.2705,
+      "step": 1966
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978195077970047e-05,
+      "loss": 1.2716,
+      "step": 1967
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978162712665852e-05,
+      "loss": 1.0482,
+      "step": 1968
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978130323624525e-05,
+      "loss": 1.2994,
+      "step": 1969
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9780979108468513e-05,
+      "loss": 1.1673,
+      "step": 1970
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978065474333618e-05,
+      "loss": 1.196,
+      "step": 1971
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978033014085612e-05,
+      "loss": 1.19,
+      "step": 1972
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9780005301036208e-05,
+      "loss": 1.2014,
+      "step": 1973
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9779680223884335e-05,
+      "loss": 1.2361,
+      "step": 1974
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.977935490940838e-05,
+      "loss": 1.2824,
+      "step": 1975
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977902935761624e-05,
+      "loss": 1.0607,
+      "step": 1976
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977870356851582e-05,
+      "loss": 1.2573,
+      "step": 1977
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9778377542115022e-05,
+      "loss": 1.2283,
+      "step": 1978
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9778051278421758e-05,
+      "loss": 1.0883,
+      "step": 1979
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777724777443943e-05,
+      "loss": 1.1918,
+      "step": 1980
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777398039189507e-05,
+      "loss": 1.1383,
+      "step": 1981
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777071063666372e-05,
+      "loss": 1.0626,
+      "step": 1982
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9776743850882476e-05,
+      "loss": 1.0831,
+      "step": 1983
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977641640084576e-05,
+      "loss": 1.2235,
+      "step": 1984
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977608871356417e-05,
+      "loss": 1.2197,
+      "step": 1985
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775760789045656e-05,
+      "loss": 1.2328,
+      "step": 1986
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775432627298175e-05,
+      "loss": 1.2005,
+      "step": 1987
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97751042283297e-05,
+      "loss": 1.3612,
+      "step": 1988
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774775592148185e-05,
+      "loss": 1.1742,
+      "step": 1989
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774446718761616e-05,
+      "loss": 1.1632,
+      "step": 1990
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774117608177968e-05,
+      "loss": 1.3677,
+      "step": 1991
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9773788260405237e-05,
+      "loss": 1.2841,
+      "step": 1992
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9773458675451406e-05,
+      "loss": 1.0985,
+      "step": 1993
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977312885332447e-05,
+      "loss": 1.0243,
+      "step": 1994
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9772798794032445e-05,
+      "loss": 1.2965,
+      "step": 1995
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977246849758333e-05,
+      "loss": 1.1968,
+      "step": 1996
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977213796398515e-05,
+      "loss": 1.3657,
+      "step": 1997
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771807193245913e-05,
+      "loss": 1.1154,
+      "step": 1998
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771476185373658e-05,
+      "loss": 1.1744,
+      "step": 1999
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771144940376413e-05,
+      "loss": 1.1669,
+      "step": 2000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9770813458262212e-05,
+      "loss": 1.2002,
+      "step": 2001
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977048173903911e-05,
+      "loss": 1.271,
+      "step": 2002
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9770149782715143e-05,
+      "loss": 1.2043,
+      "step": 2003
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9769817589298377e-05,
+      "loss": 1.1018,
+      "step": 2004
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976948515879687e-05,
+      "loss": 1.2503,
+      "step": 2005
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976915249121869e-05,
+      "loss": 1.1818,
+      "step": 2006
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768819586571907e-05,
+      "loss": 1.0822,
+      "step": 2007
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768486444864605e-05,
+      "loss": 1.1074,
+      "step": 2008
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768153066104863e-05,
+      "loss": 1.3163,
+      "step": 2009
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9767819450300773e-05,
+      "loss": 1.07,
+      "step": 2010
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976748559746043e-05,
+      "loss": 1.2366,
+      "step": 2011
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9767151507591943e-05,
+      "loss": 1.141,
+      "step": 2012
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976681718070341e-05,
+      "loss": 1.2424,
+      "step": 2013
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9766482616802943e-05,
+      "loss": 1.2443,
+      "step": 2014
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9766147815898668e-05,
+      "loss": 1.3865,
+      "step": 2015
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765812777998704e-05,
+      "loss": 1.2841,
+      "step": 2016
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765477503111187e-05,
+      "loss": 1.1881,
+      "step": 2017
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976514199124425e-05,
+      "loss": 1.1145,
+      "step": 2018
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764806242406034e-05,
+      "loss": 1.2091,
+      "step": 2019
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976447025660469e-05,
+      "loss": 1.2719,
+      "step": 2020
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764134033848367e-05,
+      "loss": 1.2901,
+      "step": 2021
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9763797574145227e-05,
+      "loss": 1.1833,
+      "step": 2022
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9763460877503435e-05,
+      "loss": 1.138,
+      "step": 2023
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976312394393116e-05,
+      "loss": 1.2113,
+      "step": 2024
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976278677343658e-05,
+      "loss": 1.0456,
+      "step": 2025
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9762449366027873e-05,
+      "loss": 1.1371,
+      "step": 2026
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9762111721713232e-05,
+      "loss": 1.1813,
+      "step": 2027
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976177384050085e-05,
+      "loss": 1.3332,
+      "step": 2028
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9761435722398926e-05,
+      "loss": 1.1751,
+      "step": 2029
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9761097367415663e-05,
+      "loss": 1.1993,
+      "step": 2030
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9760758775559275e-05,
+      "loss": 1.0957,
+      "step": 2031
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9760419946837972e-05,
+      "loss": 1.2277,
+      "step": 2032
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9760080881259986e-05,
+      "loss": 1.2128,
+      "step": 2033
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975974157883354e-05,
+      "loss": 1.0597,
+      "step": 2034
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759402039566865e-05,
+      "loss": 1.0613,
+      "step": 2035
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759062263468207e-05,
+      "loss": 1.2395,
+      "step": 2036
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9758722250545808e-05,
+      "loss": 1.3271,
+      "step": 2037
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9758382000807915e-05,
+      "loss": 1.2505,
+      "step": 2038
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9758041514262795e-05,
+      "loss": 1.0833,
+      "step": 2039
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9757700790918704e-05,
+      "loss": 1.0843,
+      "step": 2040
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9757359830783907e-05,
+      "loss": 1.2039,
+      "step": 2041
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9757018633866688e-05,
+      "loss": 1.2722,
+      "step": 2042
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9756677200175316e-05,
+      "loss": 1.1851,
+      "step": 2043
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9756335529718086e-05,
+      "loss": 1.1619,
+      "step": 2044
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755993622503283e-05,
+      "loss": 1.2343,
+      "step": 2045
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755651478539203e-05,
+      "loss": 1.3162,
+      "step": 2046
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975530909783416e-05,
+      "loss": 1.341,
+      "step": 2047
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754966480396447e-05,
+      "loss": 1.331,
+      "step": 2048
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754623626234387e-05,
+      "loss": 1.2085,
+      "step": 2049
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754280535356302e-05,
+      "loss": 1.2237,
+      "step": 2050
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9753937207770513e-05,
+      "loss": 1.3042,
+      "step": 2051
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975359364348535e-05,
+      "loss": 1.2379,
+      "step": 2052
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975324984250916e-05,
+      "loss": 1.1855,
+      "step": 2053
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752905804850275e-05,
+      "loss": 1.1927,
+      "step": 2054
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975256153051705e-05,
+      "loss": 1.1865,
+      "step": 2055
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752217019517838e-05,
+      "loss": 1.2128,
+      "step": 2056
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751872271860998e-05,
+      "loss": 1.243,
+      "step": 2057
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751527287554898e-05,
+      "loss": 1.1221,
+      "step": 2058
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975118206660791e-05,
+      "loss": 1.2092,
+      "step": 2059
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975083660902841e-05,
+      "loss": 1.0218,
+      "step": 2060
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975049091482478e-05,
+      "loss": 1.1862,
+      "step": 2061
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975014498400541e-05,
+      "loss": 1.1648,
+      "step": 2062
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97497988165787e-05,
+      "loss": 1.2203,
+      "step": 2063
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974945241255304e-05,
+      "loss": 1.1862,
+      "step": 2064
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974910577193685e-05,
+      "loss": 1.2269,
+      "step": 2065
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974875889473853e-05,
+      "loss": 1.0454,
+      "step": 2066
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9748411780966497e-05,
+      "loss": 1.2091,
+      "step": 2067
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9748064430629185e-05,
+      "loss": 1.2515,
+      "step": 2068
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747716843735015e-05,
+      "loss": 1.1433,
+      "step": 2069
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747369020292424e-05,
+      "loss": 1.1354,
+      "step": 2070
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747020960309852e-05,
+      "loss": 1.0478,
+      "step": 2071
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9746672663795748e-05,
+      "loss": 1.0767,
+      "step": 2072
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9746324130758563e-05,
+      "loss": 1.0676,
+      "step": 2073
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9745975361206753e-05,
+      "loss": 1.2567,
+      "step": 2074
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9745626355148782e-05,
+      "loss": 1.1449,
+      "step": 2075
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974527711259312e-05,
+      "loss": 1.1813,
+      "step": 2076
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744927633548246e-05,
+      "loss": 1.157,
+      "step": 2077
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744577918022637e-05,
+      "loss": 1.0324,
+      "step": 2078
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744227966024776e-05,
+      "loss": 1.1116,
+      "step": 2079
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974387777756316e-05,
+      "loss": 1.1029,
+      "step": 2080
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9743527352646286e-05,
+      "loss": 1.3734,
+      "step": 2081
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974317669128266e-05,
+      "loss": 1.2736,
+      "step": 2082
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9742825793480785e-05,
+      "loss": 1.2156,
+      "step": 2083
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9742474659249183e-05,
+      "loss": 1.2708,
+      "step": 2084
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9742123288596375e-05,
+      "loss": 1.2434,
+      "step": 2085
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974177168153088e-05,
+      "loss": 1.1626,
+      "step": 2086
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974141983806124e-05,
+      "loss": 1.3334,
+      "step": 2087
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974106775819599e-05,
+      "loss": 1.4144,
+      "step": 2088
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974071544194367e-05,
+      "loss": 1.1369,
+      "step": 2089
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9740362889312835e-05,
+      "loss": 1.251,
+      "step": 2090
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9740010100312036e-05,
+      "loss": 1.1335,
+      "step": 2091
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739657074949835e-05,
+      "loss": 1.1845,
+      "step": 2092
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739303813234802e-05,
+      "loss": 1.2469,
+      "step": 2093
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973895031517551e-05,
+      "loss": 1.1337,
+      "step": 2094
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973859658078053e-05,
+      "loss": 1.0829,
+      "step": 2095
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9738242610058457e-05,
+      "loss": 1.295,
+      "step": 2096
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973788840301787e-05,
+      "loss": 1.0116,
+      "step": 2097
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9737533959667373e-05,
+      "loss": 1.1844,
+      "step": 2098
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973717928001556e-05,
+      "loss": 1.1381,
+      "step": 2099
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736824364071047e-05,
+      "loss": 1.2114,
+      "step": 2100
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736469211842437e-05,
+      "loss": 1.2285,
+      "step": 2101
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736113823338357e-05,
+      "loss": 1.1003,
+      "step": 2102
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9735758198567427e-05,
+      "loss": 1.2774,
+      "step": 2103
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9735402337538275e-05,
+      "loss": 1.1544,
+      "step": 2104
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973504624025954e-05,
+      "loss": 1.2605,
+      "step": 2105
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9734689906739862e-05,
+      "loss": 1.1555,
+      "step": 2106
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9734333336987886e-05,
+      "loss": 1.1126,
+      "step": 2107
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9733976531012272e-05,
+      "loss": 1.2129,
+      "step": 2108
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9733619488821673e-05,
+      "loss": 1.2347,
+      "step": 2109
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9733262210424752e-05,
+      "loss": 1.1406,
+      "step": 2110
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732904695830186e-05,
+      "loss": 1.1076,
+      "step": 2111
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732546945046643e-05,
+      "loss": 1.2182,
+      "step": 2112
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732188958082813e-05,
+      "loss": 1.2397,
+      "step": 2113
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9731830734947373e-05,
+      "loss": 1.2236,
+      "step": 2114
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9731472275649023e-05,
+      "loss": 1.0958,
+      "step": 2115
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9731113580196458e-05,
+      "loss": 1.16,
+      "step": 2116
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9730754648598388e-05,
+      "loss": 1.0195,
+      "step": 2117
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973039548086352e-05,
+      "loss": 1.2103,
+      "step": 2118
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9730036077000568e-05,
+      "loss": 1.1465,
+      "step": 2119
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9729676437018256e-05,
+      "loss": 1.2164,
+      "step": 2120
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972931656092531e-05,
+      "loss": 1.1408,
+      "step": 2121
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9728956448730466e-05,
+      "loss": 1.2599,
+      "step": 2122
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972859610044246e-05,
+      "loss": 1.416,
+      "step": 2123
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9728235516070037e-05,
+      "loss": 1.1938,
+      "step": 2124
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727874695621946e-05,
+      "loss": 1.1783,
+      "step": 2125
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727513639106948e-05,
+      "loss": 1.2573,
+      "step": 2126
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727152346533798e-05,
+      "loss": 1.1769,
+      "step": 2127
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972679081791127e-05,
+      "loss": 1.3568,
+      "step": 2128
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972642905324813e-05,
+      "loss": 1.208,
+      "step": 2129
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9726067052553167e-05,
+      "loss": 1.2344,
+      "step": 2130
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9725704815835156e-05,
+      "loss": 1.2541,
+      "step": 2131
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9725342343102897e-05,
+      "loss": 1.2697,
+      "step": 2132
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9724979634365174e-05,
+      "loss": 1.1437,
+      "step": 2133
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9724616689630798e-05,
+      "loss": 1.0869,
+      "step": 2134
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9724253508908574e-05,
+      "loss": 1.1846,
+      "step": 2135
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9723890092207317e-05,
+      "loss": 1.0926,
+      "step": 2136
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972352643953584e-05,
+      "loss": 1.1256,
+      "step": 2137
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9723162550902975e-05,
+      "loss": 1.1249,
+      "step": 2138
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9722798426317552e-05,
+      "loss": 1.1922,
+      "step": 2139
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9722434065788398e-05,
+      "loss": 1.2688,
+      "step": 2140
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972206946932437e-05,
+      "loss": 1.2161,
+      "step": 2141
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97217046369343e-05,
+      "loss": 1.1344,
+      "step": 2142
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9721339568627055e-05,
+      "loss": 1.1181,
+      "step": 2143
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9720974264411484e-05,
+      "loss": 1.2952,
+      "step": 2144
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972060872429646e-05,
+      "loss": 1.3357,
+      "step": 2145
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9720242948290847e-05,
+      "loss": 1.0328,
+      "step": 2146
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9719876936403524e-05,
+      "loss": 1.3433,
+      "step": 2147
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9719510688643374e-05,
+      "loss": 1.3035,
+      "step": 2148
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971914420501928e-05,
+      "loss": 1.2835,
+      "step": 2149
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9718777485540145e-05,
+      "loss": 1.1368,
+      "step": 2150
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9718410530214857e-05,
+      "loss": 1.1447,
+      "step": 2151
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971804333905233e-05,
+      "loss": 1.1945,
+      "step": 2152
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971767591206147e-05,
+      "loss": 1.1521,
+      "step": 2153
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971730824925119e-05,
+      "loss": 1.2015,
+      "step": 2154
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716940350630424e-05,
+      "loss": 1.2328,
+      "step": 2155
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716572216208084e-05,
+      "loss": 1.1037,
+      "step": 2156
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716203845993117e-05,
+      "loss": 1.1545,
+      "step": 2157
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9715835239994457e-05,
+      "loss": 1.3558,
+      "step": 2158
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971546639822105e-05,
+      "loss": 1.2541,
+      "step": 2159
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971509732068184e-05,
+      "loss": 1.2345,
+      "step": 2160
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9714728007385795e-05,
+      "loss": 1.2062,
+      "step": 2161
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9714358458341868e-05,
+      "loss": 1.1961,
+      "step": 2162
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971398867355903e-05,
+      "loss": 1.2624,
+      "step": 2163
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713618653046258e-05,
+      "loss": 1.118,
+      "step": 2164
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713248396812524e-05,
+      "loss": 1.1545,
+      "step": 2165
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9712877904866818e-05,
+      "loss": 1.1565,
+      "step": 2166
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971250717721813e-05,
+      "loss": 1.228,
+      "step": 2167
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971213621387546e-05,
+      "loss": 1.3493,
+      "step": 2168
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9711765014847804e-05,
+      "loss": 1.2489,
+      "step": 2169
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9711393580144168e-05,
+      "loss": 1.1298,
+      "step": 2170
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9711021909773574e-05,
+      "loss": 1.1096,
+      "step": 2171
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971065000374504e-05,
+      "loss": 1.2441,
+      "step": 2172
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9710277862067583e-05,
+      "loss": 1.2871,
+      "step": 2173
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9709905484750243e-05,
+      "loss": 1.2794,
+      "step": 2174
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970953287180205e-05,
+      "loss": 1.1503,
+      "step": 2175
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9709160023232052e-05,
+      "loss": 1.2285,
+      "step": 2176
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9708786939049292e-05,
+      "loss": 1.2891,
+      "step": 2177
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9708413619262825e-05,
+      "loss": 1.2112,
+      "step": 2178
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970804006388171e-05,
+      "loss": 1.3164,
+      "step": 2179
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9707666272915016e-05,
+      "loss": 1.4474,
+      "step": 2180
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970729224637181e-05,
+      "loss": 1.2015,
+      "step": 2181
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9706917984261168e-05,
+      "loss": 1.2321,
+      "step": 2182
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9706543486592174e-05,
+      "loss": 1.1913,
+      "step": 2183
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9706168753373912e-05,
+      "loss": 1.1991,
+      "step": 2184
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9705793784615487e-05,
+      "loss": 1.1922,
+      "step": 2185
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9705418580325984e-05,
+      "loss": 1.101,
+      "step": 2186
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970504314051452e-05,
+      "loss": 1.1396,
+      "step": 2187
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9704667465190194e-05,
+      "loss": 1.1361,
+      "step": 2188
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970429155436213e-05,
+      "loss": 1.255,
+      "step": 2189
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703915408039454e-05,
+      "loss": 1.1473,
+      "step": 2190
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703539026231288e-05,
+      "loss": 1.1591,
+      "step": 2191
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703162408946766e-05,
+      "loss": 1.1147,
+      "step": 2192
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9702785556195026e-05,
+      "loss": 1.1073,
+      "step": 2193
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970240846798522e-05,
+      "loss": 1.2507,
+      "step": 2194
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970203114432649e-05,
+      "loss": 1.2185,
+      "step": 2195
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9701653585228e-05,
+      "loss": 1.1633,
+      "step": 2196
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9701275790698906e-05,
+      "loss": 1.2541,
+      "step": 2197
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9700897760748382e-05,
+      "loss": 1.0709,
+      "step": 2198
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9700519495385597e-05,
+      "loss": 1.0944,
+      "step": 2199
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970014099461973e-05,
+      "loss": 1.1283,
+      "step": 2200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969976225845997e-05,
+      "loss": 1.2528,
+      "step": 2201
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969938328691551e-05,
+      "loss": 1.0072,
+      "step": 2202
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969900407999554e-05,
+      "loss": 1.3442,
+      "step": 2203
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9698624637709263e-05,
+      "loss": 0.9888,
+      "step": 2204
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969824496006589e-05,
+      "loss": 1.2684,
+      "step": 2205
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9697865047074633e-05,
+      "loss": 1.2927,
+      "step": 2206
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969748489874471e-05,
+      "loss": 1.1063,
+      "step": 2207
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969710451508535e-05,
+      "loss": 1.1789,
+      "step": 2208
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9696723896105778e-05,
+      "loss": 1.1497,
+      "step": 2209
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969634304181524e-05,
+      "loss": 1.1884,
+      "step": 2210
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9695961952222966e-05,
+      "loss": 1.2274,
+      "step": 2211
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9695580627338212e-05,
+      "loss": 1.2657,
+      "step": 2212
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9695199067170236e-05,
+      "loss": 1.2251,
+      "step": 2213
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9694817271728284e-05,
+      "loss": 1.184,
+      "step": 2214
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969443524102163e-05,
+      "loss": 1.0896,
+      "step": 2215
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9694052975059545e-05,
+      "loss": 1.3364,
+      "step": 2216
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96936704738513e-05,
+      "loss": 0.9603,
+      "step": 2217
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9693287737406183e-05,
+      "loss": 1.2342,
+      "step": 2218
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969290476573348e-05,
+      "loss": 1.1519,
+      "step": 2219
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9692521558842485e-05,
+      "loss": 1.1552,
+      "step": 2220
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9692138116742494e-05,
+      "loss": 1.1326,
+      "step": 2221
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9691754439442815e-05,
+      "loss": 1.1499,
+      "step": 2222
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9691370526952756e-05,
+      "loss": 1.2599,
+      "step": 2223
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969098637928164e-05,
+      "loss": 1.289,
+      "step": 2224
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9690601996438782e-05,
+      "loss": 1.1962,
+      "step": 2225
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9690217378433514e-05,
+      "loss": 1.1571,
+      "step": 2226
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9689832525275166e-05,
+      "loss": 1.2701,
+      "step": 2227
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9689447436973083e-05,
+      "loss": 1.3797,
+      "step": 2228
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9689062113536605e-05,
+      "loss": 1.2069,
+      "step": 2229
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968867655497508e-05,
+      "loss": 1.1379,
+      "step": 2230
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9688290761297873e-05,
+      "loss": 1.155,
+      "step": 2231
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968790473251434e-05,
+      "loss": 1.1043,
+      "step": 2232
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968751846863385e-05,
+      "loss": 1.1515,
+      "step": 2233
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968713196966578e-05,
+      "loss": 1.2015,
+      "step": 2234
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96867452356195e-05,
+      "loss": 1.3511,
+      "step": 2235
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9686358266504406e-05,
+      "loss": 1.0576,
+      "step": 2236
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968597106232988e-05,
+      "loss": 1.1407,
+      "step": 2237
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968558362310532e-05,
+      "loss": 1.1598,
+      "step": 2238
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9685195948840137e-05,
+      "loss": 1.1709,
+      "step": 2239
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9684808039543727e-05,
+      "loss": 1.289,
+      "step": 2240
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9684419895225506e-05,
+      "loss": 1.2331,
+      "step": 2241
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9684031515894898e-05,
+      "loss": 1.0974,
+      "step": 2242
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9683642901561324e-05,
+      "loss": 1.1438,
+      "step": 2243
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9683254052234217e-05,
+      "loss": 1.2027,
+      "step": 2244
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9682864967923006e-05,
+      "loss": 1.1446,
+      "step": 2245
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9682475648637145e-05,
+      "loss": 1.2062,
+      "step": 2246
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968208609438607e-05,
+      "loss": 1.2821,
+      "step": 2247
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9681696305179243e-05,
+      "loss": 1.3286,
+      "step": 2248
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9681306281026122e-05,
+      "loss": 1.2807,
+      "step": 2249
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968091602193616e-05,
+      "loss": 1.2119,
+      "step": 2250
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9680525527918845e-05,
+      "loss": 1.2308,
+      "step": 2251
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9680134798983642e-05,
+      "loss": 1.1817,
+      "step": 2252
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9679743835140037e-05,
+      "loss": 1.3166,
+      "step": 2253
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967935263639752e-05,
+      "loss": 1.2487,
+      "step": 2254
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9678961202765572e-05,
+      "loss": 1.1604,
+      "step": 2255
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9678569534253706e-05,
+      "loss": 1.16,
+      "step": 2256
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9678177630871418e-05,
+      "loss": 1.2052,
+      "step": 2257
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967778549262822e-05,
+      "loss": 1.2045,
+      "step": 2258
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967739311953363e-05,
+      "loss": 1.2072,
+      "step": 2259
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9677000511597175e-05,
+      "loss": 1.449,
+      "step": 2260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9676607668828372e-05,
+      "loss": 1.1973,
+      "step": 2261
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967621459123676e-05,
+      "loss": 1.0417,
+      "step": 2262
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9675821278831875e-05,
+      "loss": 0.9529,
+      "step": 2263
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9675427731623267e-05,
+      "loss": 1.201,
+      "step": 2264
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967503394962048e-05,
+      "loss": 1.0937,
+      "step": 2265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9674639932833068e-05,
+      "loss": 1.2351,
+      "step": 2266
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9674245681270604e-05,
+      "loss": 1.0973,
+      "step": 2267
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9673851194942645e-05,
+      "loss": 1.0309,
+      "step": 2268
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9673456473858766e-05,
+      "loss": 1.2046,
+      "step": 2269
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967306151802855e-05,
+      "loss": 1.2608,
+      "step": 2270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9672666327461575e-05,
+      "loss": 1.1244,
+      "step": 2271
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967227090216744e-05,
+      "loss": 1.1082,
+      "step": 2272
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967187524215573e-05,
+      "loss": 1.0577,
+      "step": 2273
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967147934743605e-05,
+      "loss": 1.2467,
+      "step": 2274
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9671083218018016e-05,
+      "loss": 1.1732,
+      "step": 2275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967068685391123e-05,
+      "loss": 1.2216,
+      "step": 2276
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9670290255125314e-05,
+      "loss": 1.244,
+      "step": 2277
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9669893421669895e-05,
+      "loss": 1.2311,
+      "step": 2278
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96694963535546e-05,
+      "loss": 1.0961,
+      "step": 2279
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9669099050789063e-05,
+      "loss": 1.2558,
+      "step": 2280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966870151338293e-05,
+      "loss": 1.1178,
+      "step": 2281
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9668303741345845e-05,
+      "loss": 1.2346,
+      "step": 2282
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9667905734687463e-05,
+      "loss": 1.2518,
+      "step": 2283
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966750749341744e-05,
+      "loss": 1.1861,
+      "step": 2284
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9667109017545442e-05,
+      "loss": 1.2645,
+      "step": 2285
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966671030708114e-05,
+      "loss": 1.3605,
+      "step": 2286
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9666311362034203e-05,
+      "loss": 1.1539,
+      "step": 2287
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966591218241432e-05,
+      "loss": 1.2113,
+      "step": 2288
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9665512768231175e-05,
+      "loss": 1.3279,
+      "step": 2289
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966511311949446e-05,
+      "loss": 1.1963,
+      "step": 2290
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966471323621387e-05,
+      "loss": 1.2661,
+      "step": 2291
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966431311839912e-05,
+      "loss": 1.314,
+      "step": 2292
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9663912766059904e-05,
+      "loss": 1.2653,
+      "step": 2293
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966351217920595e-05,
+      "loss": 1.2542,
+      "step": 2294
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9663111357846976e-05,
+      "loss": 0.8969,
+      "step": 2295
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9662710301992705e-05,
+      "loss": 1.2887,
+      "step": 2296
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9662309011652874e-05,
+      "loss": 1.2918,
+      "step": 2297
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966190748683722e-05,
+      "loss": 1.0786,
+      "step": 2298
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9661505727555482e-05,
+      "loss": 1.1416,
+      "step": 2299
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9661103733817418e-05,
+      "loss": 1.1731,
+      "step": 2300
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9660701505632773e-05,
+      "loss": 1.2475,
+      "step": 2301
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9660299043011316e-05,
+      "loss": 1.2489,
+      "step": 2302
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9659896345962815e-05,
+      "loss": 1.0558,
+      "step": 2303
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9659493414497034e-05,
+      "loss": 1.2277,
+      "step": 2304
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965909024862376e-05,
+      "loss": 1.1402,
+      "step": 2305
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9658686848352767e-05,
+      "loss": 1.2483,
+      "step": 2306
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965828321369385e-05,
+      "loss": 1.1594,
+      "step": 2307
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9657879344656804e-05,
+      "loss": 1.2362,
+      "step": 2308
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965747524125143e-05,
+      "loss": 1.1217,
+      "step": 2309
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9657070903487534e-05,
+      "loss": 1.2137,
+      "step": 2310
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9656666331374927e-05,
+      "loss": 1.2576,
+      "step": 2311
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9656261524923428e-05,
+      "loss": 1.2701,
+      "step": 2312
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965585648414286e-05,
+      "loss": 1.1762,
+      "step": 2313
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965545120904305e-05,
+      "loss": 1.0778,
+      "step": 2314
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9655045699633836e-05,
+      "loss": 1.4161,
+      "step": 2315
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9654639955925057e-05,
+      "loss": 1.198,
+      "step": 2316
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9654233977926557e-05,
+      "loss": 1.1354,
+      "step": 2317
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9653827765648194e-05,
+      "loss": 1.227,
+      "step": 2318
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965342131909982e-05,
+      "loss": 1.2807,
+      "step": 2319
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9653014638291304e-05,
+      "loss": 1.1527,
+      "step": 2320
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9652607723232507e-05,
+      "loss": 1.143,
+      "step": 2321
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965220057393331e-05,
+      "loss": 1.2369,
+      "step": 2322
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9651793190403592e-05,
+      "loss": 1.1449,
+      "step": 2323
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9651385572653235e-05,
+      "loss": 1.2005,
+      "step": 2324
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9650977720692138e-05,
+      "loss": 1.2027,
+      "step": 2325
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965056963453019e-05,
+      "loss": 0.9854,
+      "step": 2326
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96501613141773e-05,
+      "loss": 1.1001,
+      "step": 2327
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9649752759643377e-05,
+      "loss": 1.2798,
+      "step": 2328
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964934397093833e-05,
+      "loss": 1.2079,
+      "step": 2329
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9648934948072086e-05,
+      "loss": 1.147,
+      "step": 2330
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9648525691054563e-05,
+      "loss": 1.2258,
+      "step": 2331
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9648116199895703e-05,
+      "loss": 1.1999,
+      "step": 2332
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964770647460543e-05,
+      "loss": 1.1192,
+      "step": 2333
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96472965151937e-05,
+      "loss": 1.3393,
+      "step": 2334
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964688632167045e-05,
+      "loss": 1.0647,
+      "step": 2335
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964647589404564e-05,
+      "loss": 1.1664,
+      "step": 2336
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964606523232923e-05,
+      "loss": 1.3594,
+      "step": 2337
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964565433653119e-05,
+      "loss": 1.2637,
+      "step": 2338
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964524320666148e-05,
+      "loss": 1.2099,
+      "step": 2339
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9644831842730084e-05,
+      "loss": 1.18,
+      "step": 2340
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9644420244746984e-05,
+      "loss": 1.2718,
+      "step": 2341
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964400841272217e-05,
+      "loss": 1.2765,
+      "step": 2342
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9643596346665634e-05,
+      "loss": 1.2531,
+      "step": 2343
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9643184046587373e-05,
+      "loss": 1.2237,
+      "step": 2344
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9642771512497395e-05,
+      "loss": 1.0846,
+      "step": 2345
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964235874440571e-05,
+      "loss": 1.2592,
+      "step": 2346
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9641945742322337e-05,
+      "loss": 1.2005,
+      "step": 2347
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9641532506257297e-05,
+      "loss": 1.1367,
+      "step": 2348
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9641119036220616e-05,
+      "loss": 1.4008,
+      "step": 2349
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964070533222233e-05,
+      "loss": 1.3246,
+      "step": 2350
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964029139427248e-05,
+      "loss": 1.0286,
+      "step": 2351
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9639877222381106e-05,
+      "loss": 1.3188,
+      "step": 2352
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9639462816558264e-05,
+      "loss": 1.1252,
+      "step": 2353
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963904817681401e-05,
+      "loss": 1.2323,
+      "step": 2354
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96386333031584e-05,
+      "loss": 1.3057,
+      "step": 2355
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9638218195601507e-05,
+      "loss": 1.1235,
+      "step": 2356
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9637802854153403e-05,
+      "loss": 1.2521,
+      "step": 2357
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9637387278824168e-05,
+      "loss": 1.2635,
+      "step": 2358
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636971469623888e-05,
+      "loss": 1.206,
+      "step": 2359
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636555426562653e-05,
+      "loss": 1.3444,
+      "step": 2360
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963613914965055e-05,
+      "loss": 1.1838,
+      "step": 2361
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9635722638897697e-05,
+      "loss": 1.111,
+      "step": 2362
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963530589431419e-05,
+      "loss": 1.2214,
+      "step": 2363
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9634888915910144e-05,
+      "loss": 1.0791,
+      "step": 2364
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963447170369568e-05,
+      "loss": 1.0106,
+      "step": 2365
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9634054257680924e-05,
+      "loss": 1.2298,
+      "step": 2366
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9633636577876004e-05,
+      "loss": 1.2918,
+      "step": 2367
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963321866429105e-05,
+      "loss": 1.2368,
+      "step": 2368
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9632800516936215e-05,
+      "loss": 1.1689,
+      "step": 2369
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9632382135821638e-05,
+      "loss": 1.166,
+      "step": 2370
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9631963520957477e-05,
+      "loss": 1.1711,
+      "step": 2371
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9631544672353886e-05,
+      "loss": 1.3526,
+      "step": 2372
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963112559002103e-05,
+      "loss": 1.1991,
+      "step": 2373
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963070627396908e-05,
+      "loss": 1.2181,
+      "step": 2374
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9630286724208216e-05,
+      "loss": 1.3444,
+      "step": 2375
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9629866940748612e-05,
+      "loss": 1.0924,
+      "step": 2376
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9629446923600458e-05,
+      "loss": 0.963,
+      "step": 2377
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9629026672773946e-05,
+      "loss": 1.2901,
+      "step": 2378
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9628606188279273e-05,
+      "loss": 1.2137,
+      "step": 2379
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9628185470126645e-05,
+      "loss": 1.1487,
+      "step": 2380
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9627764518326274e-05,
+      "loss": 1.2885,
+      "step": 2381
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962734333288837e-05,
+      "loss": 1.2211,
+      "step": 2382
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9626921913823156e-05,
+      "loss": 1.2148,
+      "step": 2383
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962650026114086e-05,
+      "loss": 1.1895,
+      "step": 2384
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9626078374851715e-05,
+      "loss": 1.1718,
+      "step": 2385
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9625656254965954e-05,
+      "loss": 1.1928,
+      "step": 2386
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962523390149382e-05,
+      "loss": 1.1735,
+      "step": 2387
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9624811314445575e-05,
+      "loss": 1.1946,
+      "step": 2388
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9624388493831462e-05,
+      "loss": 1.3031,
+      "step": 2389
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962396543966174e-05,
+      "loss": 1.3455,
+      "step": 2390
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9623542151946683e-05,
+      "loss": 1.1174,
+      "step": 2391
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962311863069656e-05,
+      "loss": 1.1469,
+      "step": 2392
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962269487592165e-05,
+      "loss": 1.2449,
+      "step": 2393
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9622270887632234e-05,
+      "loss": 1.2215,
+      "step": 2394
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9621846665838598e-05,
+      "loss": 1.2533,
+      "step": 2395
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9621422210551047e-05,
+      "loss": 0.9549,
+      "step": 2396
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962099752177987e-05,
+      "loss": 1.2022,
+      "step": 2397
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9620572599535378e-05,
+      "loss": 1.1286,
+      "step": 2398
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962014744382788e-05,
+      "loss": 1.2861,
+      "step": 2399
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9619722054667698e-05,
+      "loss": 1.1891,
+      "step": 2400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961929643206515e-05,
+      "loss": 1.1849,
+      "step": 2401
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961887057603057e-05,
+      "loss": 1.228,
+      "step": 2402
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9618444486574287e-05,
+      "loss": 1.2155,
+      "step": 2403
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9618018163706644e-05,
+      "loss": 1.203,
+      "step": 2404
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9617591607437988e-05,
+      "loss": 1.2994,
+      "step": 2405
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9617164817778666e-05,
+      "loss": 1.319,
+      "step": 2406
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9616737794739036e-05,
+      "loss": 1.2771,
+      "step": 2407
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9616310538329463e-05,
+      "loss": 1.024,
+      "step": 2408
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961588304856031e-05,
+      "loss": 1.1697,
+      "step": 2409
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961545532544196e-05,
+      "loss": 1.2516,
+      "step": 2410
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9615027368984785e-05,
+      "loss": 1.183,
+      "step": 2411
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9614599179199172e-05,
+      "loss": 1.1755,
+      "step": 2412
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961417075609551e-05,
+      "loss": 1.1273,
+      "step": 2413
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9613742099684204e-05,
+      "loss": 1.3282,
+      "step": 2414
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9613313209975645e-05,
+      "loss": 1.2115,
+      "step": 2415
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961288408698025e-05,
+      "loss": 1.1445,
+      "step": 2416
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961245473070843e-05,
+      "loss": 1.2038,
+      "step": 2417
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.96120251411706e-05,
+      "loss": 1.1777,
+      "step": 2418
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9611595318377184e-05,
+      "loss": 1.206,
+      "step": 2419
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961116526233862e-05,
+      "loss": 1.2949,
+      "step": 2420
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9610734973065342e-05,
+      "loss": 1.3543,
+      "step": 2421
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961030445056779e-05,
+      "loss": 1.3628,
+      "step": 2422
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960987369485641e-05,
+      "loss": 1.0648,
+      "step": 2423
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960944270594166e-05,
+      "loss": 1.1193,
+      "step": 2424
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9609011483833993e-05,
+      "loss": 1.1851,
+      "step": 2425
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9608580028543875e-05,
+      "loss": 1.1842,
+      "step": 2426
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960814834008178e-05,
+      "loss": 1.2611,
+      "step": 2427
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960771641845818e-05,
+      "loss": 1.2521,
+      "step": 2428
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960728426368356e-05,
+      "loss": 1.1629,
+      "step": 2429
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9606851875768404e-05,
+      "loss": 1.2366,
+      "step": 2430
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9606419254723205e-05,
+      "loss": 1.2496,
+      "step": 2431
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9605986400558462e-05,
+      "loss": 1.3127,
+      "step": 2432
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9605553313284682e-05,
+      "loss": 1.186,
+      "step": 2433
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9605119992912368e-05,
+      "loss": 1.2146,
+      "step": 2434
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960468643945204e-05,
+      "loss": 1.2858,
+      "step": 2435
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9604252652914222e-05,
+      "loss": 1.0723,
+      "step": 2436
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9603818633309434e-05,
+      "loss": 0.9633,
+      "step": 2437
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9603384380648213e-05,
+      "loss": 1.3247,
+      "step": 2438
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9602949894941096e-05,
+      "loss": 1.2579,
+      "step": 2439
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9602515176198623e-05,
+      "loss": 1.2946,
+      "step": 2440
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960208022443135e-05,
+      "loss": 1.335,
+      "step": 2441
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9601645039649828e-05,
+      "loss": 1.1091,
+      "step": 2442
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9601209621864616e-05,
+      "loss": 1.2456,
+      "step": 2443
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9600773971086286e-05,
+      "loss": 1.0873,
+      "step": 2444
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9600338087325407e-05,
+      "loss": 1.2939,
+      "step": 2445
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959990197059255e-05,
+      "loss": 0.9583,
+      "step": 2446
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9599465620898314e-05,
+      "loss": 1.2637,
+      "step": 2447
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959902903825327e-05,
+      "loss": 1.2781,
+      "step": 2448
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959859222266803e-05,
+      "loss": 1.1217,
+      "step": 2449
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9598155174153174e-05,
+      "loss": 1.1335,
+      "step": 2450
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9597717892719326e-05,
+      "loss": 1.0984,
+      "step": 2451
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9597280378377087e-05,
+      "loss": 1.1952,
+      "step": 2452
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959684263113708e-05,
+      "loss": 1.3258,
+      "step": 2453
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9596404651009928e-05,
+      "loss": 1.192,
+      "step": 2454
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595966438006253e-05,
+      "loss": 1.2461,
+      "step": 2455
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595527992136697e-05,
+      "loss": 1.153,
+      "step": 2456
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595089313411892e-05,
+      "loss": 1.222,
+      "step": 2457
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9594650401842493e-05,
+      "loss": 1.1201,
+      "step": 2458
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959421125743914e-05,
+      "loss": 1.175,
+      "step": 2459
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9593771880212498e-05,
+      "loss": 1.1404,
+      "step": 2460
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9593332270173225e-05,
+      "loss": 1.2123,
+      "step": 2461
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9592892427331993e-05,
+      "loss": 1.2036,
+      "step": 2462
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9592452351699475e-05,
+      "loss": 1.1236,
+      "step": 2463
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9592012043286342e-05,
+      "loss": 1.2187,
+      "step": 2464
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9591571502103294e-05,
+      "loss": 1.1387,
+      "step": 2465
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959113072816101e-05,
+      "loss": 1.2586,
+      "step": 2466
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9590689721470188e-05,
+      "loss": 1.1156,
+      "step": 2467
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9590248482041533e-05,
+      "loss": 1.2807,
+      "step": 2468
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958980700988575e-05,
+      "loss": 1.1671,
+      "step": 2469
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9589365305013556e-05,
+      "loss": 1.2977,
+      "step": 2470
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9588923367435667e-05,
+      "loss": 1.0601,
+      "step": 2471
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9588481197162804e-05,
+      "loss": 1.1046,
+      "step": 2472
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9588038794205705e-05,
+      "loss": 1.2327,
+      "step": 2473
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9587596158575102e-05,
+      "loss": 1.2074,
+      "step": 2474
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9587153290281734e-05,
+      "loss": 1.0338,
+      "step": 2475
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958671018933635e-05,
+      "loss": 1.1272,
+      "step": 2476
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958626685574971e-05,
+      "loss": 1.1608,
+      "step": 2477
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9585823289532556e-05,
+      "loss": 1.2396,
+      "step": 2478
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958537949069567e-05,
+      "loss": 1.1973,
+      "step": 2479
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9584935459249807e-05,
+      "loss": 1.2867,
+      "step": 2480
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958449119520575e-05,
+      "loss": 1.1385,
+      "step": 2481
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9584046698574282e-05,
+      "loss": 1.1964,
+      "step": 2482
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958360196936618e-05,
+      "loss": 1.064,
+      "step": 2483
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958315700759225e-05,
+      "loss": 1.2471,
+      "step": 2484
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9582711813263277e-05,
+      "loss": 1.0277,
+      "step": 2485
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9582266386390075e-05,
+      "loss": 1.0597,
+      "step": 2486
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9581820726983443e-05,
+      "loss": 1.2109,
+      "step": 2487
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9581374835054205e-05,
+      "loss": 1.2052,
+      "step": 2488
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9580928710613176e-05,
+      "loss": 1.2735,
+      "step": 2489
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9580482353671184e-05,
+      "loss": 1.1159,
+      "step": 2490
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958003576423906e-05,
+      "loss": 1.1638,
+      "step": 2491
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9579588942327642e-05,
+      "loss": 1.351,
+      "step": 2492
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9579141887947772e-05,
+      "loss": 1.0294,
+      "step": 2493
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95786946011103e-05,
+      "loss": 1.0586,
+      "step": 2494
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9578247081826083e-05,
+      "loss": 1.2637,
+      "step": 2495
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9577799330105973e-05,
+      "loss": 1.2186,
+      "step": 2496
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9577351345960845e-05,
+      "loss": 1.234,
+      "step": 2497
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9576903129401563e-05,
+      "loss": 1.2511,
+      "step": 2498
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957645468043901e-05,
+      "loss": 1.245,
+      "step": 2499
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957600599908406e-05,
+      "loss": 1.1556,
+      "step": 2500
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957555708534761e-05,
+      "loss": 1.0863,
+      "step": 2501
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9575107939240548e-05,
+      "loss": 1.2547,
+      "step": 2502
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957465856077378e-05,
+      "loss": 1.2244,
+      "step": 2503
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9574208949958203e-05,
+      "loss": 1.2073,
+      "step": 2504
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9573759106804732e-05,
+      "loss": 1.1739,
+      "step": 2505
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9573309031324284e-05,
+      "loss": 1.3078,
+      "step": 2506
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957285872352778e-05,
+      "loss": 1.1815,
+      "step": 2507
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9572408183426145e-05,
+      "loss": 1.1829,
+      "step": 2508
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9571957411030318e-05,
+      "loss": 1.1517,
+      "step": 2509
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9571506406351233e-05,
+      "loss": 1.025,
+      "step": 2510
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9571055169399837e-05,
+      "loss": 1.0461,
+      "step": 2511
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957060370018708e-05,
+      "loss": 1.2237,
+      "step": 2512
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9570151998723918e-05,
+      "loss": 1.093,
+      "step": 2513
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956970006502131e-05,
+      "loss": 1.2136,
+      "step": 2514
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956924789909022e-05,
+      "loss": 1.2955,
+      "step": 2515
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9568795500941635e-05,
+      "loss": 1.087,
+      "step": 2516
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956834287058652e-05,
+      "loss": 1.2068,
+      "step": 2517
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9567890008035865e-05,
+      "loss": 1.2154,
+      "step": 2518
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956743691330065e-05,
+      "loss": 1.2421,
+      "step": 2519
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9566983586391884e-05,
+      "loss": 1.1213,
+      "step": 2520
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956653002732056e-05,
+      "loss": 1.237,
+      "step": 2521
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9566076236097695e-05,
+      "loss": 0.9828,
+      "step": 2522
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956562221273428e-05,
+      "loss": 1.2188,
+      "step": 2523
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9565167957241353e-05,
+      "loss": 1.199,
+      "step": 2524
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9564713469629928e-05,
+      "loss": 1.2882,
+      "step": 2525
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9564258749911035e-05,
+      "loss": 1.1588,
+      "step": 2526
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956380379809571e-05,
+      "loss": 1.2664,
+      "step": 2527
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9563348614194992e-05,
+      "loss": 1.0713,
+      "step": 2528
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956289319821993e-05,
+      "loss": 1.2911,
+      "step": 2529
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9562437550181573e-05,
+      "loss": 1.1977,
+      "step": 2530
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9561981670090978e-05,
+      "loss": 1.3126,
+      "step": 2531
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9561525557959207e-05,
+      "loss": 1.1597,
+      "step": 2532
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9561069213797333e-05,
+      "loss": 1.1791,
+      "step": 2533
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9560612637616428e-05,
+      "loss": 1.2116,
+      "step": 2534
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9560155829427567e-05,
+      "loss": 1.2361,
+      "step": 2535
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9559698789241844e-05,
+      "loss": 1.1499,
+      "step": 2536
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955924151707034e-05,
+      "loss": 1.2026,
+      "step": 2537
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955878401292416e-05,
+      "loss": 1.1686,
+      "step": 2538
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955832627681441e-05,
+      "loss": 1.1591,
+      "step": 2539
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955786830875218e-05,
+      "loss": 1.3846,
+      "step": 2540
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95574101087486e-05,
+      "loss": 1.0857,
+      "step": 2541
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9556951676814787e-05,
+      "loss": 1.3147,
+      "step": 2542
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9556493012961856e-05,
+      "loss": 1.2602,
+      "step": 2543
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955603411720095e-05,
+      "loss": 1.1217,
+      "step": 2544
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9555574989543197e-05,
+      "loss": 1.1197,
+      "step": 2545
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9555115629999738e-05,
+      "loss": 1.204,
+      "step": 2546
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9554656038581728e-05,
+      "loss": 1.3265,
+      "step": 2547
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9554196215300314e-05,
+      "loss": 1.1341,
+      "step": 2548
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9553736160166657e-05,
+      "loss": 1.2157,
+      "step": 2549
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9553275873191916e-05,
+      "loss": 1.124,
+      "step": 2550
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9552815354387267e-05,
+      "loss": 1.2142,
+      "step": 2551
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9552354603763882e-05,
+      "loss": 1.2759,
+      "step": 2552
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9551893621332944e-05,
+      "loss": 1.2255,
+      "step": 2553
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9551432407105642e-05,
+      "loss": 1.2526,
+      "step": 2554
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955097096109316e-05,
+      "loss": 1.1808,
+      "step": 2555
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9550509283306703e-05,
+      "loss": 1.2215,
+      "step": 2556
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9550047373757475e-05,
+      "loss": 1.2841,
+      "step": 2557
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9549585232456682e-05,
+      "loss": 1.2165,
+      "step": 2558
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9549122859415538e-05,
+      "loss": 1.1173,
+      "step": 2559
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9548660254645265e-05,
+      "loss": 1.2692,
+      "step": 2560
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954819741815709e-05,
+      "loss": 1.1305,
+      "step": 2561
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9547734349962246e-05,
+      "loss": 1.2181,
+      "step": 2562
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9547271050071965e-05,
+      "loss": 1.1752,
+      "step": 2563
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9546807518497496e-05,
+      "loss": 1.1462,
+      "step": 2564
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954634375525008e-05,
+      "loss": 1.2282,
+      "step": 2565
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9545879760340983e-05,
+      "loss": 1.2254,
+      "step": 2566
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9545415533781453e-05,
+      "loss": 1.2663,
+      "step": 2567
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954495107558276e-05,
+      "loss": 1.1638,
+      "step": 2568
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9544486385756176e-05,
+      "loss": 1.1683,
+      "step": 2569
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9544021464312977e-05,
+      "loss": 1.1824,
+      "step": 2570
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9543556311264445e-05,
+      "loss": 1.1432,
+      "step": 2571
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954309092662187e-05,
+      "loss": 1.3127,
+      "step": 2572
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9542625310396538e-05,
+      "loss": 1.0607,
+      "step": 2573
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9542159462599755e-05,
+      "loss": 1.2165,
+      "step": 2574
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954169338324283e-05,
+      "loss": 1.2594,
+      "step": 2575
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954122707233706e-05,
+      "loss": 1.1233,
+      "step": 2576
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954076052989377e-05,
+      "loss": 1.2607,
+      "step": 2577
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9540293755924285e-05,
+      "loss": 1.1751,
+      "step": 2578
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9539826750439926e-05,
+      "loss": 0.9299,
+      "step": 2579
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9539359513452026e-05,
+      "loss": 1.2483,
+      "step": 2580
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9538892044971925e-05,
+      "loss": 1.1809,
+      "step": 2581
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9538424345010968e-05,
+      "loss": 1.0313,
+      "step": 2582
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9537956413580504e-05,
+      "loss": 1.1455,
+      "step": 2583
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9537488250691884e-05,
+      "loss": 1.1277,
+      "step": 2584
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9537019856356478e-05,
+      "loss": 1.2146,
+      "step": 2585
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9536551230585643e-05,
+      "loss": 1.1782,
+      "step": 2586
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953608237339076e-05,
+      "loss": 1.2757,
+      "step": 2587
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95356132847832e-05,
+      "loss": 1.1089,
+      "step": 2588
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953514396477435e-05,
+      "loss": 1.2878,
+      "step": 2589
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9534674413375595e-05,
+      "loss": 1.1891,
+      "step": 2590
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9534204630598334e-05,
+      "loss": 1.2648,
+      "step": 2591
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953373461645397e-05,
+      "loss": 1.1511,
+      "step": 2592
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9533264370953898e-05,
+      "loss": 1.1786,
+      "step": 2593
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953279389410954e-05,
+      "loss": 1.272,
+      "step": 2594
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9532323185932306e-05,
+      "loss": 1.2099,
+      "step": 2595
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953185224643362e-05,
+      "loss": 1.0909,
+      "step": 2596
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953138107562492e-05,
+      "loss": 1.1991,
+      "step": 2597
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953090967351763e-05,
+      "loss": 1.1593,
+      "step": 2598
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9530438040123188e-05,
+      "loss": 1.2601,
+      "step": 2599
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952996617545304e-05,
+      "loss": 1.0874,
+      "step": 2600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9529494079518647e-05,
+      "loss": 1.2433,
+      "step": 2601
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9529021752331455e-05,
+      "loss": 1.0236,
+      "step": 2602
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9528549193902926e-05,
+      "loss": 1.2279,
+      "step": 2603
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9528076404244537e-05,
+      "loss": 1.0607,
+      "step": 2604
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952760338336775e-05,
+      "loss": 1.1394,
+      "step": 2605
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952713013128405e-05,
+      "loss": 1.1995,
+      "step": 2606
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9526656648004918e-05,
+      "loss": 1.2059,
+      "step": 2607
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952618293354185e-05,
+      "loss": 1.2289,
+      "step": 2608
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9525708987906334e-05,
+      "loss": 1.4091,
+      "step": 2609
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9525234811109874e-05,
+      "loss": 1.2607,
+      "step": 2610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952476040316398e-05,
+      "loss": 1.0127,
+      "step": 2611
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9524285764080166e-05,
+      "loss": 1.2303,
+      "step": 2612
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9523810893869937e-05,
+      "loss": 1.0683,
+      "step": 2613
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9523335792544835e-05,
+      "loss": 1.2682,
+      "step": 2614
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9522860460116377e-05,
+      "loss": 1.1375,
+      "step": 2615
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9522384896596102e-05,
+      "loss": 1.136,
+      "step": 2616
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952190910199555e-05,
+      "loss": 1.0911,
+      "step": 2617
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9521433076326267e-05,
+      "loss": 1.1416,
+      "step": 2618
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9520956819599804e-05,
+      "loss": 1.1015,
+      "step": 2619
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9520480331827718e-05,
+      "loss": 1.2363,
+      "step": 2620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9520003613021577e-05,
+      "loss": 1.0069,
+      "step": 2621
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951952666319294e-05,
+      "loss": 1.2088,
+      "step": 2622
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9519049482353393e-05,
+      "loss": 1.1801,
+      "step": 2623
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9518572070514507e-05,
+      "loss": 1.3201,
+      "step": 2624
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9518094427687866e-05,
+      "loss": 1.3337,
+      "step": 2625
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951761655388507e-05,
+      "loss": 1.2305,
+      "step": 2626
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9517138449117706e-05,
+      "loss": 1.1951,
+      "step": 2627
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9516660113397386e-05,
+      "loss": 1.2463,
+      "step": 2628
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951618154673571e-05,
+      "loss": 1.1583,
+      "step": 2629
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9515702749144293e-05,
+      "loss": 1.2149,
+      "step": 2630
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951522372063476e-05,
+      "loss": 0.9879,
+      "step": 2631
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9514744461218725e-05,
+      "loss": 1.1769,
+      "step": 2632
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9514264970907825e-05,
+      "loss": 1.1866,
+      "step": 2633
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9513785249713697e-05,
+      "loss": 1.1742,
+      "step": 2634
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9513305297647976e-05,
+      "loss": 1.1129,
+      "step": 2635
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512825114722314e-05,
+      "loss": 1.142,
+      "step": 2636
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512344700948363e-05,
+      "loss": 1.046,
+      "step": 2637
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9511864056337784e-05,
+      "loss": 1.0664,
+      "step": 2638
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9511383180902237e-05,
+      "loss": 1.1337,
+      "step": 2639
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951090207465339e-05,
+      "loss": 1.0798,
+      "step": 2640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951042073760292e-05,
+      "loss": 1.2645,
+      "step": 2641
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950993916976251e-05,
+      "loss": 1.1633,
+      "step": 2642
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9509457371143843e-05,
+      "loss": 1.0259,
+      "step": 2643
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950897534175861e-05,
+      "loss": 1.133,
+      "step": 2644
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9508493081618515e-05,
+      "loss": 1.1024,
+      "step": 2645
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9508010590735252e-05,
+      "loss": 1.1889,
+      "step": 2646
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9507527869120534e-05,
+      "loss": 1.0218,
+      "step": 2647
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950704491678608e-05,
+      "loss": 1.0297,
+      "step": 2648
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95065617337436e-05,
+      "loss": 1.1966,
+      "step": 2649
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9506078320004825e-05,
+      "loss": 1.025,
+      "step": 2650
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950559467558149e-05,
+      "loss": 1.3642,
+      "step": 2651
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9505110800485324e-05,
+      "loss": 1.1064,
+      "step": 2652
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950462669472807e-05,
+      "loss": 1.124,
+      "step": 2653
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950414235832148e-05,
+      "loss": 1.2099,
+      "step": 2654
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950365779127731e-05,
+      "loss": 1.2343,
+      "step": 2655
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950317299360731e-05,
+      "loss": 1.3035,
+      "step": 2656
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950268796532325e-05,
+      "loss": 1.2287,
+      "step": 2657
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95022027064369e-05,
+      "loss": 1.1915,
+      "step": 2658
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9501717216960035e-05,
+      "loss": 1.1975,
+      "step": 2659
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9501231496904435e-05,
+      "loss": 1.2197,
+      "step": 2660
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9500745546281893e-05,
+      "loss": 1.3133,
+      "step": 2661
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9500259365104192e-05,
+      "loss": 1.213,
+      "step": 2662
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949977295338314e-05,
+      "loss": 1.2245,
+      "step": 2663
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9499286311130533e-05,
+      "loss": 1.2253,
+      "step": 2664
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9498799438358186e-05,
+      "loss": 1.1393,
+      "step": 2665
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949831233507791e-05,
+      "loss": 1.2151,
+      "step": 2666
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949782500130153e-05,
+      "loss": 1.3806,
+      "step": 2667
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9497337437040867e-05,
+      "loss": 1.2664,
+      "step": 2668
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9496849642307754e-05,
+      "loss": 1.2257,
+      "step": 2669
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949636161711403e-05,
+      "loss": 1.1005,
+      "step": 2670
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9495873361471538e-05,
+      "loss": 1.3221,
+      "step": 2671
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9495384875392125e-05,
+      "loss": 1.1176,
+      "step": 2672
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9494896158887647e-05,
+      "loss": 1.1739,
+      "step": 2673
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949440721196996e-05,
+      "loss": 1.3017,
+      "step": 2674
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9493918034650934e-05,
+      "loss": 1.2058,
+      "step": 2675
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9493428626942443e-05,
+      "loss": 1.0478,
+      "step": 2676
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9492938988856354e-05,
+      "loss": 1.2113,
+      "step": 2677
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949244912040455e-05,
+      "loss": 1.326,
+      "step": 2678
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9491959021598927e-05,
+      "loss": 1.2237,
+      "step": 2679
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9491468692451373e-05,
+      "loss": 1.2427,
+      "step": 2680
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9490978132973785e-05,
+      "loss": 1.271,
+      "step": 2681
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9490487343178072e-05,
+      "loss": 1.0743,
+      "step": 2682
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948999632307614e-05,
+      "loss": 1.2927,
+      "step": 2683
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9489505072679907e-05,
+      "loss": 1.3255,
+      "step": 2684
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9489013592001293e-05,
+      "loss": 1.3002,
+      "step": 2685
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9488521881052225e-05,
+      "loss": 1.2733,
+      "step": 2686
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9488029939844634e-05,
+      "loss": 1.0325,
+      "step": 2687
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9487537768390465e-05,
+      "loss": 1.2389,
+      "step": 2688
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9487045366701652e-05,
+      "loss": 1.0745,
+      "step": 2689
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948655273479015e-05,
+      "loss": 1.2533,
+      "step": 2690
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948605987266791e-05,
+      "loss": 1.1748,
+      "step": 2691
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94855667803469e-05,
+      "loss": 1.1012,
+      "step": 2692
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9485073457839072e-05,
+      "loss": 1.1292,
+      "step": 2693
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948457990515641e-05,
+      "loss": 1.2203,
+      "step": 2694
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9484086122310887e-05,
+      "loss": 1.1788,
+      "step": 2695
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9483592109314487e-05,
+      "loss": 1.3091,
+      "step": 2696
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9483097866179194e-05,
+      "loss": 1.27,
+      "step": 2697
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9482603392917006e-05,
+      "loss": 1.0328,
+      "step": 2698
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948210868953992e-05,
+      "loss": 1.1933,
+      "step": 2699
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9481613756059944e-05,
+      "loss": 1.1058,
+      "step": 2700
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9481118592489086e-05,
+      "loss": 1.2294,
+      "step": 2701
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9480623198839362e-05,
+      "loss": 1.2287,
+      "step": 2702
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9480127575122795e-05,
+      "loss": 1.2268,
+      "step": 2703
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9479631721351412e-05,
+      "loss": 1.2219,
+      "step": 2704
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947913563753724e-05,
+      "loss": 1.173,
+      "step": 2705
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947863932369233e-05,
+      "loss": 1.0002,
+      "step": 2706
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9478142779828717e-05,
+      "loss": 1.0796,
+      "step": 2707
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9477646005958454e-05,
+      "loss": 1.0826,
+      "step": 2708
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9477149002093595e-05,
+      "loss": 1.2741,
+      "step": 2709
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9476651768246203e-05,
+      "loss": 1.2472,
+      "step": 2710
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947615430442834e-05,
+      "loss": 1.3023,
+      "step": 2711
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947565661065208e-05,
+      "loss": 1.0436,
+      "step": 2712
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9475158686929498e-05,
+      "loss": 1.2871,
+      "step": 2713
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9474660533272684e-05,
+      "loss": 1.2046,
+      "step": 2714
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9474162149693724e-05,
+      "loss": 1.3566,
+      "step": 2715
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947366353620471e-05,
+      "loss": 1.1596,
+      "step": 2716
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947316469281774e-05,
+      "loss": 1.3206,
+      "step": 2717
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9472665619544927e-05,
+      "loss": 1.3048,
+      "step": 2718
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9472166316398376e-05,
+      "loss": 1.2234,
+      "step": 2719
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9471666783390204e-05,
+      "loss": 1.2474,
+      "step": 2720
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9471167020532533e-05,
+      "loss": 1.2266,
+      "step": 2721
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9470667027837497e-05,
+      "loss": 1.1902,
+      "step": 2722
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9470166805317217e-05,
+      "loss": 1.2565,
+      "step": 2723
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9469666352983845e-05,
+      "loss": 1.2796,
+      "step": 2724
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946916567084952e-05,
+      "loss": 1.17,
+      "step": 2725
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9468664758926393e-05,
+      "loss": 1.2061,
+      "step": 2726
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9468163617226613e-05,
+      "loss": 1.1144,
+      "step": 2727
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9467662245762354e-05,
+      "loss": 1.2833,
+      "step": 2728
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9467160644545767e-05,
+      "loss": 1.1915,
+      "step": 2729
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946665881358904e-05,
+      "loss": 1.0587,
+      "step": 2730
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9466156752904344e-05,
+      "loss": 1.05,
+      "step": 2731
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9465654462503862e-05,
+      "loss": 1.2621,
+      "step": 2732
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946515194239978e-05,
+      "loss": 1.1461,
+      "step": 2733
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94646491926043e-05,
+      "loss": 1.1666,
+      "step": 2734
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9464146213129615e-05,
+      "loss": 1.3119,
+      "step": 2735
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9463643003987938e-05,
+      "loss": 1.2023,
+      "step": 2736
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9463139565191476e-05,
+      "loss": 1.2382,
+      "step": 2737
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9462635896752448e-05,
+      "loss": 1.0387,
+      "step": 2738
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9462131998683073e-05,
+      "loss": 1.225,
+      "step": 2739
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9461627870995585e-05,
+      "loss": 1.2284,
+      "step": 2740
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9461123513702208e-05,
+      "loss": 1.0492,
+      "step": 2741
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9460618926815195e-05,
+      "loss": 1.1875,
+      "step": 2742
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9460114110346775e-05,
+      "loss": 1.2055,
+      "step": 2743
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9459609064309212e-05,
+      "loss": 1.1547,
+      "step": 2744
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9459103788714756e-05,
+      "loss": 1.158,
+      "step": 2745
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945859828357567e-05,
+      "loss": 1.1931,
+      "step": 2746
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945809254890422e-05,
+      "loss": 1.2336,
+      "step": 2747
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9457586584712678e-05,
+      "loss": 1.2081,
+      "step": 2748
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9457080391013325e-05,
+      "loss": 1.0698,
+      "step": 2749
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945657396781844e-05,
+      "loss": 1.2289,
+      "step": 2750
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945606731514032e-05,
+      "loss": 1.1522,
+      "step": 2751
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9455560432991253e-05,
+      "loss": 1.2971,
+      "step": 2752
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9455053321383542e-05,
+      "loss": 1.3179,
+      "step": 2753
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9454545980329493e-05,
+      "loss": 1.1547,
+      "step": 2754
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945403840984142e-05,
+      "loss": 1.3048,
+      "step": 2755
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9453530609931635e-05,
+      "loss": 1.097,
+      "step": 2756
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9453022580612468e-05,
+      "loss": 1.0973,
+      "step": 2757
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9452514321896242e-05,
+      "loss": 1.2995,
+      "step": 2758
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945200583379529e-05,
+      "loss": 1.0434,
+      "step": 2759
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9451497116321954e-05,
+      "loss": 1.1243,
+      "step": 2760
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9450988169488577e-05,
+      "loss": 1.2332,
+      "step": 2761
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9450478993307517e-05,
+      "loss": 1.0497,
+      "step": 2762
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944996958779112e-05,
+      "loss": 1.2601,
+      "step": 2763
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9449459952951756e-05,
+      "loss": 1.2413,
+      "step": 2764
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944895008880179e-05,
+      "loss": 1.162,
+      "step": 2765
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944843999535359e-05,
+      "loss": 1.2107,
+      "step": 2766
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944792967261954e-05,
+      "loss": 1.0471,
+      "step": 2767
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9447419120612018e-05,
+      "loss": 1.3557,
+      "step": 2768
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9446908339343422e-05,
+      "loss": 1.1741,
+      "step": 2769
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9446397328826145e-05,
+      "loss": 1.3219,
+      "step": 2770
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944588608907258e-05,
+      "loss": 1.142,
+      "step": 2771
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9445374620095142e-05,
+      "loss": 1.1683,
+      "step": 2772
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944486292190624e-05,
+      "loss": 1.0757,
+      "step": 2773
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944435099451829e-05,
+      "loss": 1.2729,
+      "step": 2774
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9443838837943717e-05,
+      "loss": 1.2213,
+      "step": 2775
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9443326452194948e-05,
+      "loss": 1.2064,
+      "step": 2776
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9442813837284416e-05,
+      "loss": 1.1805,
+      "step": 2777
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9442300993224568e-05,
+      "loss": 1.4105,
+      "step": 2778
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9441787920027843e-05,
+      "loss": 1.2157,
+      "step": 2779
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944127461770669e-05,
+      "loss": 1.1882,
+      "step": 2780
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9440761086273564e-05,
+      "loss": 1.2052,
+      "step": 2781
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944024732574094e-05,
+      "loss": 1.2574,
+      "step": 2782
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9439733336121267e-05,
+      "loss": 1.1497,
+      "step": 2783
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9439219117427034e-05,
+      "loss": 1.3865,
+      "step": 2784
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943870466967071e-05,
+      "loss": 1.2401,
+      "step": 2785
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943818999286478e-05,
+      "loss": 1.2403,
+      "step": 2786
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9437675087021736e-05,
+      "loss": 1.1766,
+      "step": 2787
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9437159952154078e-05,
+      "loss": 1.3001,
+      "step": 2788
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9436644588274295e-05,
+      "loss": 1.1133,
+      "step": 2789
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9436128995394903e-05,
+      "loss": 1.1278,
+      "step": 2790
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943561317352841e-05,
+      "loss": 1.2299,
+      "step": 2791
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9435097122687337e-05,
+      "loss": 1.1944,
+      "step": 2792
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94345808428842e-05,
+      "loss": 1.1766,
+      "step": 2793
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943406433413154e-05,
+      "loss": 1.2271,
+      "step": 2794
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9433547596441877e-05,
+      "loss": 1.1391,
+      "step": 2795
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9433030629827757e-05,
+      "loss": 1.184,
+      "step": 2796
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9432513434301727e-05,
+      "loss": 1.1913,
+      "step": 2797
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943199600987633e-05,
+      "loss": 1.2058,
+      "step": 2798
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943147835656414e-05,
+      "loss": 1.1344,
+      "step": 2799
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9430960474377697e-05,
+      "loss": 1.2183,
+      "step": 2800
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9430442363329583e-05,
+      "loss": 1.26,
+      "step": 2801
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9429924023432364e-05,
+      "loss": 1.2409,
+      "step": 2802
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9429405454698624e-05,
+      "loss": 1.1081,
+      "step": 2803
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9428886657140945e-05,
+      "loss": 1.1508,
+      "step": 2804
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9428367630771915e-05,
+      "loss": 1.3041,
+      "step": 2805
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.942784837560413e-05,
+      "loss": 1.3048,
+      "step": 2806
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9427328891650194e-05,
+      "loss": 1.2395,
+      "step": 2807
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9426809178922706e-05,
+      "loss": 1.2397,
+      "step": 2808
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9426289237434286e-05,
+      "loss": 1.1247,
+      "step": 2809
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9425769067197548e-05,
+      "loss": 1.1853,
+      "step": 2810
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942524866822511e-05,
+      "loss": 1.0714,
+      "step": 2811
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9424728040529612e-05,
+      "loss": 1.2202,
+      "step": 2812
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9424207184123677e-05,
+      "loss": 1.1685,
+      "step": 2813
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942368609901995e-05,
+      "loss": 1.3175,
+      "step": 2814
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9423164785231078e-05,
+      "loss": 1.2962,
+      "step": 2815
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942264324276971e-05,
+      "loss": 1.0326,
+      "step": 2816
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.94221214716485e-05,
+      "loss": 1.1239,
+      "step": 2817
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9421599471880108e-05,
+      "loss": 1.247,
+      "step": 2818
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9421077243477208e-05,
+      "loss": 1.2294,
+      "step": 2819
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942055478645247e-05,
+      "loss": 1.1787,
+      "step": 2820
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942003210081857e-05,
+      "loss": 1.275,
+      "step": 2821
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9419509186588196e-05,
+      "loss": 1.2282,
+      "step": 2822
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9418986043774036e-05,
+      "loss": 1.1825,
+      "step": 2823
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9418462672388784e-05,
+      "loss": 1.1296,
+      "step": 2824
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941793907244514e-05,
+      "loss": 1.2262,
+      "step": 2825
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941741524395581e-05,
+      "loss": 1.0902,
+      "step": 2826
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9416891186933516e-05,
+      "loss": 1.0801,
+      "step": 2827
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941636690139096e-05,
+      "loss": 1.1912,
+      "step": 2828
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9415842387340875e-05,
+      "loss": 1.1922,
+      "step": 2829
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9415317644795984e-05,
+      "loss": 1.2735,
+      "step": 2830
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9414792673769028e-05,
+      "loss": 1.161,
+      "step": 2831
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9414267474272735e-05,
+      "loss": 1.1681,
+      "step": 2832
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9413742046319863e-05,
+      "loss": 1.1346,
+      "step": 2833
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941321638992315e-05,
+      "loss": 1.2723,
+      "step": 2834
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9412690505095363e-05,
+      "loss": 1.1338,
+      "step": 2835
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941216439184926e-05,
+      "loss": 1.1593,
+      "step": 2836
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9411638050197605e-05,
+      "loss": 1.2292,
+      "step": 2837
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9411111480153174e-05,
+      "loss": 1.2718,
+      "step": 2838
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9410584681728745e-05,
+      "loss": 1.1665,
+      "step": 2839
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.94100576549371e-05,
+      "loss": 1.1386,
+      "step": 2840
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9409530399791026e-05,
+      "loss": 1.2756,
+      "step": 2841
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940900291630333e-05,
+      "loss": 1.2446,
+      "step": 2842
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.94084752044868e-05,
+      "loss": 1.1271,
+      "step": 2843
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9407947264354242e-05,
+      "loss": 1.2581,
+      "step": 2844
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9407419095918477e-05,
+      "loss": 1.383,
+      "step": 2845
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9406890699192316e-05,
+      "loss": 1.3073,
+      "step": 2846
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9406362074188584e-05,
+      "loss": 1.2745,
+      "step": 2847
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9405833220920104e-05,
+      "loss": 1.2106,
+      "step": 2848
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9405304139399715e-05,
+      "loss": 1.1918,
+      "step": 2849
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9404774829640254e-05,
+      "loss": 1.2499,
+      "step": 2850
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9404245291654568e-05,
+      "loss": 1.2199,
+      "step": 2851
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9403715525455503e-05,
+      "loss": 1.2972,
+      "step": 2852
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9403185531055915e-05,
+      "loss": 1.1509,
+      "step": 2853
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9402655308468678e-05,
+      "loss": 1.1786,
+      "step": 2854
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940212485770664e-05,
+      "loss": 1.3298,
+      "step": 2855
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9401594178782686e-05,
+      "loss": 1.2932,
+      "step": 2856
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940106327170969e-05,
+      "loss": 1.196,
+      "step": 2857
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940053213650053e-05,
+      "loss": 1.1311,
+      "step": 2858
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9400000773168107e-05,
+      "loss": 1.2231,
+      "step": 2859
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939946918172531e-05,
+      "loss": 1.1852,
+      "step": 2860
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9398937362185037e-05,
+      "loss": 1.0735,
+      "step": 2861
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9398405314560197e-05,
+      "loss": 1.2338,
+      "step": 2862
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93978730388637e-05,
+      "loss": 1.3069,
+      "step": 2863
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939734053510846e-05,
+      "loss": 1.3218,
+      "step": 2864
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9396807803307405e-05,
+      "loss": 1.3896,
+      "step": 2865
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9396274843473455e-05,
+      "loss": 1.1978,
+      "step": 2866
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9395741655619554e-05,
+      "loss": 0.9941,
+      "step": 2867
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939520823975863e-05,
+      "loss": 1.3457,
+      "step": 2868
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9394674595903635e-05,
+      "loss": 1.255,
+      "step": 2869
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9394140724067515e-05,
+      "loss": 1.1995,
+      "step": 2870
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9393606624263228e-05,
+      "loss": 1.1312,
+      "step": 2871
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9393072296503733e-05,
+      "loss": 1.1371,
+      "step": 2872
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9392537740801997e-05,
+      "loss": 1.2676,
+      "step": 2873
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9392002957170994e-05,
+      "loss": 1.3717,
+      "step": 2874
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9391467945623698e-05,
+      "loss": 1.2187,
+      "step": 2875
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9390932706173095e-05,
+      "loss": 1.2255,
+      "step": 2876
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9390397238832173e-05,
+      "loss": 1.2324,
+      "step": 2877
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9389861543613927e-05,
+      "loss": 1.243,
+      "step": 2878
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9389325620531356e-05,
+      "loss": 1.218,
+      "step": 2879
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9388789469597464e-05,
+      "loss": 1.3388,
+      "step": 2880
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9388253090825263e-05,
+      "loss": 1.1945,
+      "step": 2881
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938771648422777e-05,
+      "loss": 1.2179,
+      "step": 2882
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9387179649818007e-05,
+      "loss": 1.1152,
+      "step": 2883
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9386642587609002e-05,
+      "loss": 1.31,
+      "step": 2884
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9386105297613782e-05,
+      "loss": 1.0995,
+      "step": 2885
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9385567779845392e-05,
+      "loss": 1.2545,
+      "step": 2886
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9385030034316873e-05,
+      "loss": 1.2277,
+      "step": 2887
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938449206104128e-05,
+      "loss": 1.1506,
+      "step": 2888
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9383953860031658e-05,
+      "loss": 1.1838,
+      "step": 2889
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9383415431301075e-05,
+      "loss": 1.1652,
+      "step": 2890
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9382876774862594e-05,
+      "loss": 1.2303,
+      "step": 2891
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938233789072929e-05,
+      "loss": 1.1536,
+      "step": 2892
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938179877891424e-05,
+      "loss": 1.2573,
+      "step": 2893
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9381259439430517e-05,
+      "loss": 0.9271,
+      "step": 2894
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9380719872291223e-05,
+      "loss": 1.1382,
+      "step": 2895
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9380180077509444e-05,
+      "loss": 1.2335,
+      "step": 2896
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937964005509828e-05,
+      "loss": 1.3264,
+      "step": 2897
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9379099805070836e-05,
+      "loss": 1.2135,
+      "step": 2898
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9378559327440222e-05,
+      "loss": 1.2696,
+      "step": 2899
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9378018622219557e-05,
+      "loss": 1.2033,
+      "step": 2900
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9377477689421958e-05,
+      "loss": 1.0516,
+      "step": 2901
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9376936529060556e-05,
+      "loss": 1.1939,
+      "step": 2902
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9376395141148475e-05,
+      "loss": 0.9936,
+      "step": 2903
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9375853525698866e-05,
+      "loss": 1.192,
+      "step": 2904
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9375311682724863e-05,
+      "loss": 1.0762,
+      "step": 2905
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9374769612239617e-05,
+      "loss": 1.2968,
+      "step": 2906
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937422731425628e-05,
+      "loss": 1.2708,
+      "step": 2907
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9373684788788018e-05,
+      "loss": 1.3023,
+      "step": 2908
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937314203584799e-05,
+      "loss": 1.1314,
+      "step": 2909
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9372599055449374e-05,
+      "loss": 1.1853,
+      "step": 2910
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937205584760534e-05,
+      "loss": 1.22,
+      "step": 2911
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9371512412329078e-05,
+      "loss": 1.3067,
+      "step": 2912
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9370968749633764e-05,
+      "loss": 1.3071,
+      "step": 2913
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9370424859532598e-05,
+      "loss": 1.1469,
+      "step": 2914
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9369880742038783e-05,
+      "loss": 1.1634,
+      "step": 2915
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9369336397165512e-05,
+      "loss": 0.9019,
+      "step": 2916
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9368791824926006e-05,
+      "loss": 1.1587,
+      "step": 2917
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9368247025333472e-05,
+      "loss": 1.2748,
+      "step": 2918
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9367701998401137e-05,
+      "loss": 1.1237,
+      "step": 2919
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9367156744142218e-05,
+      "loss": 1.3967,
+      "step": 2920
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936661126256996e-05,
+      "loss": 0.9447,
+      "step": 2921
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9366065553697586e-05,
+      "loss": 1.1606,
+      "step": 2922
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936551961753835e-05,
+      "loss": 1.2287,
+      "step": 2923
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9364973454105494e-05,
+      "loss": 1.1557,
+      "step": 2924
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9364427063412276e-05,
+      "loss": 1.1805,
+      "step": 2925
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936388044547195e-05,
+      "loss": 1.0961,
+      "step": 2926
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9363333600297784e-05,
+      "loss": 1.1499,
+      "step": 2927
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936278652790305e-05,
+      "loss": 1.0472,
+      "step": 2928
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9362239228301023e-05,
+      "loss": 1.1957,
+      "step": 2929
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936169170150498e-05,
+      "loss": 1.2889,
+      "step": 2930
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936114394752821e-05,
+      "loss": 1.2,
+      "step": 2931
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936059596638401e-05,
+      "loss": 1.3114,
+      "step": 2932
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9360047758085675e-05,
+      "loss": 1.1442,
+      "step": 2933
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9359499322646505e-05,
+      "loss": 1.2703,
+      "step": 2934
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9358950660079815e-05,
+      "loss": 1.1621,
+      "step": 2935
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9358401770398912e-05,
+      "loss": 1.1564,
+      "step": 2936
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9357852653617123e-05,
+      "loss": 1.2536,
+      "step": 2937
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9357303309747772e-05,
+      "loss": 1.2524,
+      "step": 2938
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9356753738804185e-05,
+      "loss": 1.2408,
+      "step": 2939
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9356203940799702e-05,
+      "loss": 1.1483,
+      "step": 2940
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9355653915747668e-05,
+      "loss": 1.1384,
+      "step": 2941
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9355103663661426e-05,
+      "loss": 1.1543,
+      "step": 2942
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935455318455433e-05,
+      "loss": 0.9407,
+      "step": 2943
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935400247843974e-05,
+      "loss": 1.2069,
+      "step": 2944
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935345154533102e-05,
+      "loss": 1.2821,
+      "step": 2945
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9352900385241534e-05,
+      "loss": 1.252,
+      "step": 2946
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9352348998184664e-05,
+      "loss": 1.2487,
+      "step": 2947
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9351797384173787e-05,
+      "loss": 1.1463,
+      "step": 2948
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9351245543222292e-05,
+      "loss": 1.1935,
+      "step": 2949
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935069347534357e-05,
+      "loss": 1.043,
+      "step": 2950
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9350141180551014e-05,
+      "loss": 1.2199,
+      "step": 2951
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934958865885803e-05,
+      "loss": 1.1417,
+      "step": 2952
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9349035910278027e-05,
+      "loss": 1.1312,
+      "step": 2953
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9348482934824413e-05,
+      "loss": 1.0423,
+      "step": 2954
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9347929732510614e-05,
+      "loss": 1.194,
+      "step": 2955
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9347376303350056e-05,
+      "loss": 1.2246,
+      "step": 2956
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9346822647356158e-05,
+      "loss": 1.3438,
+      "step": 2957
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934626876454236e-05,
+      "loss": 1.253,
+      "step": 2958
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9345714654922112e-05,
+      "loss": 1.1766,
+      "step": 2959
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9345160318508853e-05,
+      "loss": 1.2336,
+      "step": 2960
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9344605755316035e-05,
+      "loss": 1.1674,
+      "step": 2961
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9344050965357117e-05,
+      "loss": 1.2667,
+      "step": 2962
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9343495948645562e-05,
+      "loss": 1.2037,
+      "step": 2963
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9342940705194838e-05,
+      "loss": 1.0912,
+      "step": 2964
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934238523501842e-05,
+      "loss": 1.2097,
+      "step": 2965
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9341829538129787e-05,
+      "loss": 1.2925,
+      "step": 2966
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9341273614542427e-05,
+      "loss": 1.1611,
+      "step": 2967
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9340717464269823e-05,
+      "loss": 1.3296,
+      "step": 2968
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9340161087325483e-05,
+      "loss": 1.1634,
+      "step": 2969
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9339604483722896e-05,
+      "loss": 1.2103,
+      "step": 2970
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933904765347558e-05,
+      "loss": 1.155,
+      "step": 2971
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933849059659704e-05,
+      "loss": 1.3454,
+      "step": 2972
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9337933313100793e-05,
+      "loss": 1.2456,
+      "step": 2973
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933737580300037e-05,
+      "loss": 1.2729,
+      "step": 2974
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9336818066309297e-05,
+      "loss": 1.195,
+      "step": 2975
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9336260103041108e-05,
+      "loss": 1.281,
+      "step": 2976
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9335701913209342e-05,
+      "loss": 1.2173,
+      "step": 2977
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9335143496827546e-05,
+      "loss": 1.1554,
+      "step": 2978
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933458485390927e-05,
+      "loss": 1.3215,
+      "step": 2979
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9334025984468075e-05,
+      "loss": 1.2702,
+      "step": 2980
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9333466888517518e-05,
+      "loss": 1.2446,
+      "step": 2981
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9332907566071168e-05,
+      "loss": 1.211,
+      "step": 2982
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9332348017142598e-05,
+      "loss": 1.2842,
+      "step": 2983
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933178824174539e-05,
+      "loss": 1.183,
+      "step": 2984
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933122823989312e-05,
+      "loss": 1.3119,
+      "step": 2985
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9330668011599388e-05,
+      "loss": 1.239,
+      "step": 2986
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933010755687778e-05,
+      "loss": 1.2467,
+      "step": 2987
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9329546875741904e-05,
+      "loss": 1.2665,
+      "step": 2988
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932898596820536e-05,
+      "loss": 1.1338,
+      "step": 2989
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9328424834281763e-05,
+      "loss": 1.0842,
+      "step": 2990
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932786347398473e-05,
+      "loss": 1.097,
+      "step": 2991
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932730188732788e-05,
+      "loss": 1.2404,
+      "step": 2992
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9326740074324846e-05,
+      "loss": 1.1796,
+      "step": 2993
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932617803498926e-05,
+      "loss": 1.2642,
+      "step": 2994
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9325615769334755e-05,
+      "loss": 1.3057,
+      "step": 2995
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9325053277374986e-05,
+      "loss": 1.1221,
+      "step": 2996
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93244905591236e-05,
+      "loss": 1.2679,
+      "step": 2997
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9323927614594245e-05,
+      "loss": 1.1986,
+      "step": 2998
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932336444380059e-05,
+      "loss": 1.1993,
+      "step": 2999
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93228010467563e-05,
+      "loss": 1.2597,
+      "step": 3000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9322237423475044e-05,
+      "loss": 1.0204,
+      "step": 3001
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9321673573970502e-05,
+      "loss": 1.1884,
+      "step": 3002
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932110949825636e-05,
+      "loss": 1.1245,
+      "step": 3003
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9320545196346295e-05,
+      "loss": 1.1198,
+      "step": 3004
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9319980668254016e-05,
+      "loss": 1.1797,
+      "step": 3005
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9319415913993214e-05,
+      "loss": 1.2634,
+      "step": 3006
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9318850933577592e-05,
+      "loss": 1.0469,
+      "step": 3007
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9318285727020867e-05,
+      "loss": 1.1263,
+      "step": 3008
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9317720294336747e-05,
+      "loss": 1.1722,
+      "step": 3009
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9317154635538964e-05,
+      "loss": 1.2832,
+      "step": 3010
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9316588750641233e-05,
+      "loss": 1.1676,
+      "step": 3011
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9316022639657296e-05,
+      "loss": 1.1856,
+      "step": 3012
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9315456302600885e-05,
+      "loss": 1.1631,
+      "step": 3013
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9314889739485747e-05,
+      "loss": 1.2499,
+      "step": 3014
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931432295032563e-05,
+      "loss": 1.0917,
+      "step": 3015
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9313755935134286e-05,
+      "loss": 1.2599,
+      "step": 3016
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9313188693925475e-05,
+      "loss": 1.2995,
+      "step": 3017
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9312621226712962e-05,
+      "loss": 1.189,
+      "step": 3018
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9312053533510525e-05,
+      "loss": 1.2251,
+      "step": 3019
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9311485614331928e-05,
+      "loss": 1.1807,
+      "step": 3020
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9310917469190965e-05,
+      "loss": 1.0897,
+      "step": 3021
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9310349098101412e-05,
+      "loss": 1.2614,
+      "step": 3022
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930978050107707e-05,
+      "loss": 1.2736,
+      "step": 3023
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9309211678131733e-05,
+      "loss": 1.1484,
+      "step": 3024
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930864262927921e-05,
+      "loss": 1.0792,
+      "step": 3025
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9308073354533302e-05,
+      "loss": 1.0266,
+      "step": 3026
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9307503853907832e-05,
+      "loss": 1.1736,
+      "step": 3027
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930693412741661e-05,
+      "loss": 1.2655,
+      "step": 3028
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9306364175073474e-05,
+      "loss": 1.3587,
+      "step": 3029
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9305793996892244e-05,
+      "loss": 1.0648,
+      "step": 3030
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930522359288676e-05,
+      "loss": 1.1347,
+      "step": 3031
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9304652963070868e-05,
+      "loss": 1.3604,
+      "step": 3032
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9304082107458412e-05,
+      "loss": 1.132,
+      "step": 3033
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9303511026063244e-05,
+      "loss": 1.2382,
+      "step": 3034
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930293971889923e-05,
+      "loss": 1.2323,
+      "step": 3035
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9302368185980218e-05,
+      "loss": 1.2231,
+      "step": 3036
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9301796427320093e-05,
+      "loss": 1.1758,
+      "step": 3037
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9301224442932725e-05,
+      "loss": 1.3051,
+      "step": 3038
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930065223283199e-05,
+      "loss": 1.1478,
+      "step": 3039
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930007979703178e-05,
+      "loss": 1.078,
+      "step": 3040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9299507135545986e-05,
+      "loss": 1.1896,
+      "step": 3041
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92989342483885e-05,
+      "loss": 1.173,
+      "step": 3042
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929836113557323e-05,
+      "loss": 1.2207,
+      "step": 3043
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9297787797114078e-05,
+      "loss": 1.1772,
+      "step": 3044
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929721423302496e-05,
+      "loss": 1.1254,
+      "step": 3045
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9296640443319793e-05,
+      "loss": 1.2041,
+      "step": 3046
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9296066428012508e-05,
+      "loss": 1.2642,
+      "step": 3047
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9295492187117025e-05,
+      "loss": 1.1059,
+      "step": 3048
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9294917720647287e-05,
+      "loss": 1.0921,
+      "step": 3049
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929434302861723e-05,
+      "loss": 1.2897,
+      "step": 3050
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92937681110408e-05,
+      "loss": 1.1546,
+      "step": 3051
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929319296793195e-05,
+      "loss": 1.2519,
+      "step": 3052
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929261759930464e-05,
+      "loss": 1.1735,
+      "step": 3053
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9292042005172823e-05,
+      "loss": 1.1977,
+      "step": 3054
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9291466185550482e-05,
+      "loss": 1.0982,
+      "step": 3055
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9290890140451576e-05,
+      "loss": 1.1391,
+      "step": 3056
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929031386989009e-05,
+      "loss": 1.3536,
+      "step": 3057
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928973737388001e-05,
+      "loss": 1.2335,
+      "step": 3058
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9289160652435326e-05,
+      "loss": 1.2453,
+      "step": 3059
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9288583705570026e-05,
+      "loss": 1.0942,
+      "step": 3060
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928800653329812e-05,
+      "loss": 1.1567,
+      "step": 3061
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928742913563361e-05,
+      "loss": 1.259,
+      "step": 3062
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928685151259051e-05,
+      "loss": 1.0968,
+      "step": 3063
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9286273664182832e-05,
+      "loss": 1.2344,
+      "step": 3064
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9285695590424604e-05,
+      "loss": 1.1508,
+      "step": 3065
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9285117291329853e-05,
+      "loss": 1.1959,
+      "step": 3066
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9284538766912608e-05,
+      "loss": 1.2925,
+      "step": 3067
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9283960017186916e-05,
+      "loss": 1.2443,
+      "step": 3068
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928338104216682e-05,
+      "loss": 1.13,
+      "step": 3069
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928280184186636e-05,
+      "loss": 1.0721,
+      "step": 3070
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9282222416299604e-05,
+      "loss": 1.3145,
+      "step": 3071
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9281642765480605e-05,
+      "loss": 1.2069,
+      "step": 3072
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9281062889423436e-05,
+      "loss": 1.227,
+      "step": 3073
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9280482788142162e-05,
+      "loss": 1.0726,
+      "step": 3074
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9279902461650866e-05,
+      "loss": 1.1238,
+      "step": 3075
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9279321909963627e-05,
+      "loss": 1.2668,
+      "step": 3076
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9278741133094535e-05,
+      "loss": 1.2358,
+      "step": 3077
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9278160131057686e-05,
+      "loss": 1.3662,
+      "step": 3078
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9277578903867174e-05,
+      "loss": 1.0862,
+      "step": 3079
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9276997451537107e-05,
+      "loss": 1.2279,
+      "step": 3080
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9276415774081593e-05,
+      "loss": 1.1623,
+      "step": 3081
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927583387151475e-05,
+      "loss": 1.2315,
+      "step": 3082
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92752517438507e-05,
+      "loss": 1.2111,
+      "step": 3083
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9274669391103567e-05,
+      "loss": 1.1008,
+      "step": 3084
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9274086813287484e-05,
+      "loss": 1.0918,
+      "step": 3085
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9273504010416586e-05,
+      "loss": 1.2133,
+      "step": 3086
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927292098250502e-05,
+      "loss": 1.1983,
+      "step": 3087
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9272337729566933e-05,
+      "loss": 1.0802,
+      "step": 3088
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9271754251616475e-05,
+      "loss": 1.2254,
+      "step": 3089
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927117054866781e-05,
+      "loss": 1.2111,
+      "step": 3090
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9270586620735102e-05,
+      "loss": 1.2437,
+      "step": 3091
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927000246783252e-05,
+      "loss": 1.1117,
+      "step": 3092
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926941808997424e-05,
+      "loss": 1.1061,
+      "step": 3093
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9268833487174447e-05,
+      "loss": 0.9805,
+      "step": 3094
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926824865944732e-05,
+      "loss": 1.2904,
+      "step": 3095
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9267663606807055e-05,
+      "loss": 1.1342,
+      "step": 3096
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9267078329267853e-05,
+      "loss": 1.2336,
+      "step": 3097
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926649282684391e-05,
+      "loss": 1.1493,
+      "step": 3098
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9265907099549438e-05,
+      "loss": 1.2334,
+      "step": 3099
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926532114739865e-05,
+      "loss": 1.0293,
+      "step": 3100
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926473497040577e-05,
+      "loss": 1.2751,
+      "step": 3101
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9264148568585013e-05,
+      "loss": 1.2633,
+      "step": 3102
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9263561941950622e-05,
+      "loss": 1.1928,
+      "step": 3103
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926297509051682e-05,
+      "loss": 1.099,
+      "step": 3104
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926238801429786e-05,
+      "loss": 1.2443,
+      "step": 3105
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926180071330798e-05,
+      "loss": 1.2945,
+      "step": 3106
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9261213187561433e-05,
+      "loss": 1.1114,
+      "step": 3107
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926062543707248e-05,
+      "loss": 1.1082,
+      "step": 3108
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926003746185538e-05,
+      "loss": 1.1663,
+      "step": 3109
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9259449261924405e-05,
+      "loss": 1.2372,
+      "step": 3110
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9258860837293824e-05,
+      "loss": 1.0387,
+      "step": 3111
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9258272187977924e-05,
+      "loss": 1.1017,
+      "step": 3112
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9257683313990984e-05,
+      "loss": 1.0224,
+      "step": 3113
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9257094215347298e-05,
+      "loss": 1.0893,
+      "step": 3114
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9256504892061156e-05,
+      "loss": 1.1371,
+      "step": 3115
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9255915344146865e-05,
+      "loss": 1.2022,
+      "step": 3116
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9255325571618728e-05,
+      "loss": 1.0665,
+      "step": 3117
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925473557449106e-05,
+      "loss": 1.2153,
+      "step": 3118
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9254145352778176e-05,
+      "loss": 1.0927,
+      "step": 3119
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92535549064944e-05,
+      "loss": 1.245,
+      "step": 3120
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9252964235654058e-05,
+      "loss": 1.2203,
+      "step": 3121
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925237334027149e-05,
+      "loss": 1.2591,
+      "step": 3122
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9251782220361027e-05,
+      "loss": 1.2817,
+      "step": 3123
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9251190875937024e-05,
+      "loss": 1.1812,
+      "step": 3124
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925059930701382e-05,
+      "loss": 0.9349,
+      "step": 3125
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9250007513605776e-05,
+      "loss": 1.0766,
+      "step": 3126
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9249415495727252e-05,
+      "loss": 1.093,
+      "step": 3127
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924882325339262e-05,
+      "loss": 1.0492,
+      "step": 3128
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9248230786616244e-05,
+      "loss": 1.1393,
+      "step": 3129
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9247638095412508e-05,
+      "loss": 1.2982,
+      "step": 3130
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9247045179795788e-05,
+      "loss": 1.2492,
+      "step": 3131
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924645203978048e-05,
+      "loss": 1.2778,
+      "step": 3132
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924585867538097e-05,
+      "loss": 1.1769,
+      "step": 3133
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924526508661166e-05,
+      "loss": 1.2444,
+      "step": 3134
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9244671273486962e-05,
+      "loss": 1.2073,
+      "step": 3135
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9244077236021273e-05,
+      "loss": 1.2752,
+      "step": 3136
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924348297422902e-05,
+      "loss": 1.2377,
+      "step": 3137
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924288848812462e-05,
+      "loss": 1.1142,
+      "step": 3138
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9242293777722496e-05,
+      "loss": 1.3244,
+      "step": 3139
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9241698843037083e-05,
+      "loss": 1.1566,
+      "step": 3140
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9241103684082815e-05,
+      "loss": 1.1843,
+      "step": 3141
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9240508300874145e-05,
+      "loss": 1.0737,
+      "step": 3142
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9239912693425506e-05,
+      "loss": 1.1428,
+      "step": 3143
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9239316861751365e-05,
+      "loss": 1.1725,
+      "step": 3144
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9238720805866174e-05,
+      "loss": 1.3185,
+      "step": 3145
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92381245257844e-05,
+      "loss": 1.1195,
+      "step": 3146
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9237528021520512e-05,
+      "loss": 1.2845,
+      "step": 3147
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9236931293088982e-05,
+      "loss": 1.0391,
+      "step": 3148
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9236334340504298e-05,
+      "loss": 1.3138,
+      "step": 3149
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923573716378094e-05,
+      "loss": 1.1135,
+      "step": 3150
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9235139762933402e-05,
+      "loss": 1.3635,
+      "step": 3151
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9234542137976184e-05,
+      "loss": 1.2586,
+      "step": 3152
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9233944288923788e-05,
+      "loss": 1.12,
+      "step": 3153
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9233346215790717e-05,
+      "loss": 1.0636,
+      "step": 3154
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9232747918591488e-05,
+      "loss": 1.2502,
+      "step": 3155
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923214939734062e-05,
+      "loss": 1.2812,
+      "step": 3156
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9231550652052635e-05,
+      "loss": 1.3279,
+      "step": 3157
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9230951682742066e-05,
+      "loss": 1.1899,
+      "step": 3158
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923035248942345e-05,
+      "loss": 1.247,
+      "step": 3159
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9229753072111325e-05,
+      "loss": 1.1873,
+      "step": 3160
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9229153430820232e-05,
+      "loss": 1.241,
+      "step": 3161
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9228553565564728e-05,
+      "loss": 1.1705,
+      "step": 3162
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922795347635937e-05,
+      "loss": 1.1143,
+      "step": 3163
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922735316321872e-05,
+      "loss": 1.2451,
+      "step": 3164
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9226752626157345e-05,
+      "loss": 1.1428,
+      "step": 3165
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922615186518982e-05,
+      "loss": 0.9494,
+      "step": 3166
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9225550880330718e-05,
+      "loss": 1.093,
+      "step": 3167
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9224949671594633e-05,
+      "loss": 1.2644,
+      "step": 3168
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9224348238996146e-05,
+      "loss": 1.1755,
+      "step": 3169
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9223746582549853e-05,
+      "loss": 1.1391,
+      "step": 3170
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922314470227036e-05,
+      "loss": 1.1771,
+      "step": 3171
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9222542598172268e-05,
+      "loss": 1.1455,
+      "step": 3172
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922194027027019e-05,
+      "loss": 1.3296,
+      "step": 3173
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9221337718578744e-05,
+      "loss": 1.2295,
+      "step": 3174
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922073494311255e-05,
+      "loss": 1.2628,
+      "step": 3175
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9220131943886232e-05,
+      "loss": 1.1144,
+      "step": 3176
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9219528720914432e-05,
+      "loss": 1.2671,
+      "step": 3177
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921892527421178e-05,
+      "loss": 1.2219,
+      "step": 3178
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9218321603792928e-05,
+      "loss": 1.1565,
+      "step": 3179
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921771770967252e-05,
+      "loss": 1.2498,
+      "step": 3180
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921711359186521e-05,
+      "loss": 1.1986,
+      "step": 3181
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921650925038566e-05,
+      "loss": 1.2081,
+      "step": 3182
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9215904685248534e-05,
+      "loss": 1.0695,
+      "step": 3183
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9215299896468503e-05,
+      "loss": 1.2447,
+      "step": 3184
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9214694884060248e-05,
+      "loss": 1.292,
+      "step": 3185
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9214089648038446e-05,
+      "loss": 1.2192,
+      "step": 3186
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9213484188417788e-05,
+      "loss": 1.1351,
+      "step": 3187
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921287850521296e-05,
+      "loss": 1.2778,
+      "step": 3188
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921227259843867e-05,
+      "loss": 1.1811,
+      "step": 3189
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9211666468109612e-05,
+      "loss": 1.2062,
+      "step": 3190
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9211060114240503e-05,
+      "loss": 1.1255,
+      "step": 3191
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9210453536846053e-05,
+      "loss": 1.2705,
+      "step": 3192
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920984673594098e-05,
+      "loss": 1.0867,
+      "step": 3193
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9209239711540014e-05,
+      "loss": 1.2442,
+      "step": 3194
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9208632463657885e-05,
+      "loss": 0.9212,
+      "step": 3195
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9208024992309325e-05,
+      "loss": 1.2047,
+      "step": 3196
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920741729750908e-05,
+      "loss": 1.0617,
+      "step": 3197
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9206809379271892e-05,
+      "loss": 1.343,
+      "step": 3198
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920620123761252e-05,
+      "loss": 1.3768,
+      "step": 3199
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920559287254572e-05,
+      "loss": 1.0782,
+      "step": 3200
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920498428408625e-05,
+      "loss": 1.1636,
+      "step": 3201
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9204375472248885e-05,
+      "loss": 1.1438,
+      "step": 3202
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9203766437048395e-05,
+      "loss": 1.1715,
+      "step": 3203
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920315717849956e-05,
+      "loss": 1.108,
+      "step": 3204
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9202547696617165e-05,
+      "loss": 1.2612,
+      "step": 3205
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9201937991416003e-05,
+      "loss": 1.2681,
+      "step": 3206
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920132806291087e-05,
+      "loss": 1.1719,
+      "step": 3207
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9200717911116564e-05,
+      "loss": 1.273,
+      "step": 3208
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920010753604789e-05,
+      "loss": 1.2068,
+      "step": 3209
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9199496937719663e-05,
+      "loss": 1.1159,
+      "step": 3210
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.91988861161467e-05,
+      "loss": 1.3061,
+      "step": 3211
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9198275071343827e-05,
+      "loss": 1.0975,
+      "step": 3212
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9197663803325867e-05,
+      "loss": 1.2294,
+      "step": 3213
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9197052312107655e-05,
+      "loss": 1.2065,
+      "step": 3214
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9196440597704033e-05,
+      "loss": 1.2682,
+      "step": 3215
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9195828660129842e-05,
+      "loss": 1.2271,
+      "step": 3216
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9195216499399932e-05,
+      "loss": 1.129,
+      "step": 3217
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9194604115529163e-05,
+      "loss": 1.2415,
+      "step": 3218
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919399150853239e-05,
+      "loss": 1.0534,
+      "step": 3219
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9193378678424484e-05,
+      "loss": 1.2001,
+      "step": 3220
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9192765625220312e-05,
+      "loss": 1.3,
+      "step": 3221
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9192152348934753e-05,
+      "loss": 1.0077,
+      "step": 3222
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.919153884958269e-05,
+      "loss": 1.2932,
+      "step": 3223
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9190925127179013e-05,
+      "loss": 1.2949,
+      "step": 3224
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.919031118173861e-05,
+      "loss": 1.1646,
+      "step": 3225
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9189697013276386e-05,
+      "loss": 1.0141,
+      "step": 3226
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9189082621807235e-05,
+      "loss": 1.0466,
+      "step": 3227
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918846800734608e-05,
+      "loss": 1.3213,
+      "step": 3228
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9187853169907824e-05,
+      "loss": 1.2372,
+      "step": 3229
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9187238109507393e-05,
+      "loss": 1.2273,
+      "step": 3230
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918662282615971e-05,
+      "loss": 1.0022,
+      "step": 3231
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9186007319879714e-05,
+      "loss": 1.3831,
+      "step": 3232
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918539159068233e-05,
+      "loss": 1.1788,
+      "step": 3233
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9184775638582508e-05,
+      "loss": 1.2714,
+      "step": 3234
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918415946359519e-05,
+      "loss": 1.0218,
+      "step": 3235
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918354306573533e-05,
+      "loss": 1.1823,
+      "step": 3236
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9182926445017893e-05,
+      "loss": 1.2417,
+      "step": 3237
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9182309601457837e-05,
+      "loss": 1.2108,
+      "step": 3238
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9181692535070128e-05,
+      "loss": 1.28,
+      "step": 3239
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9181075245869744e-05,
+      "loss": 1.1722,
+      "step": 3240
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9180457733871666e-05,
+      "loss": 1.0934,
+      "step": 3241
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9179839999090874e-05,
+      "loss": 1.2663,
+      "step": 3242
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9179222041542366e-05,
+      "loss": 1.1832,
+      "step": 3243
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9178603861241133e-05,
+      "loss": 1.059,
+      "step": 3244
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917798545820218e-05,
+      "loss": 1.2517,
+      "step": 3245
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9177366832440505e-05,
+      "loss": 1.2277,
+      "step": 3246
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917674798397113e-05,
+      "loss": 1.2294,
+      "step": 3247
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917612891280907e-05,
+      "loss": 1.1701,
+      "step": 3248
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9175509618969347e-05,
+      "loss": 1.2276,
+      "step": 3249
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917489010246699e-05,
+      "loss": 1.1819,
+      "step": 3250
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917427036331703e-05,
+      "loss": 1.1089,
+      "step": 3251
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9173650401534514e-05,
+      "loss": 1.1443,
+      "step": 3252
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917303021713448e-05,
+      "loss": 1.1568,
+      "step": 3253
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9172409810131975e-05,
+      "loss": 1.0882,
+      "step": 3254
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9171789180542066e-05,
+      "loss": 1.1501,
+      "step": 3255
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9171168328379803e-05,
+      "loss": 1.2878,
+      "step": 3256
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9170547253660253e-05,
+      "loss": 1.1231,
+      "step": 3257
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9169925956398497e-05,
+      "loss": 1.3279,
+      "step": 3258
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91693044366096e-05,
+      "loss": 1.326,
+      "step": 3259
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9168682694308654e-05,
+      "loss": 1.1675,
+      "step": 3260
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9168060729510742e-05,
+      "loss": 1.2709,
+      "step": 3261
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916743854223096e-05,
+      "loss": 1.1171,
+      "step": 3262
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9166816132484404e-05,
+      "loss": 1.1404,
+      "step": 3263
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9166193500286177e-05,
+      "loss": 1.265,
+      "step": 3264
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9165570645651392e-05,
+      "loss": 1.0724,
+      "step": 3265
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9164947568595164e-05,
+      "loss": 1.1344,
+      "step": 3266
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9164324269132608e-05,
+      "loss": 1.17,
+      "step": 3267
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9163700747278857e-05,
+      "loss": 1.1784,
+      "step": 3268
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9163077003049037e-05,
+      "loss": 1.248,
+      "step": 3269
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9162453036458287e-05,
+      "loss": 1.1484,
+      "step": 3270
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9161828847521743e-05,
+      "loss": 1.2076,
+      "step": 3271
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9161204436254565e-05,
+      "loss": 1.1923,
+      "step": 3272
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916057980267189e-05,
+      "loss": 1.2631,
+      "step": 3273
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915995494678889e-05,
+      "loss": 1.2803,
+      "step": 3274
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9159329868620714e-05,
+      "loss": 1.1263,
+      "step": 3275
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9158704568182543e-05,
+      "loss": 1.2249,
+      "step": 3276
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9158079045489547e-05,
+      "loss": 1.2216,
+      "step": 3277
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9157453300556904e-05,
+      "loss": 1.1384,
+      "step": 3278
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9156827333399805e-05,
+      "loss": 1.2442,
+      "step": 3279
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915620114403343e-05,
+      "loss": 1.0429,
+      "step": 3280
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9155574732472983e-05,
+      "loss": 1.2344,
+      "step": 3281
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9154948098733663e-05,
+      "loss": 1.2238,
+      "step": 3282
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9154321242830676e-05,
+      "loss": 1.1371,
+      "step": 3283
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9153694164779234e-05,
+      "loss": 1.1419,
+      "step": 3284
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9153066864594558e-05,
+      "loss": 1.2017,
+      "step": 3285
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9152439342291865e-05,
+      "loss": 1.1207,
+      "step": 3286
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9151811597886383e-05,
+      "loss": 1.1742,
+      "step": 3287
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9151183631393352e-05,
+      "loss": 1.2864,
+      "step": 3288
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9150555442828004e-05,
+      "loss": 1.2405,
+      "step": 3289
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914992703220559e-05,
+      "loss": 1.0695,
+      "step": 3290
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9149298399541353e-05,
+      "loss": 1.2217,
+      "step": 3291
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9148669544850552e-05,
+      "loss": 1.3735,
+      "step": 3292
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9148040468148442e-05,
+      "loss": 1.2389,
+      "step": 3293
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9147411169450302e-05,
+      "loss": 1.2208,
+      "step": 3294
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9146781648771387e-05,
+      "loss": 1.1947,
+      "step": 3295
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9146151906126983e-05,
+      "loss": 1.1554,
+      "step": 3296
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9145521941532374e-05,
+      "loss": 0.9933,
+      "step": 3297
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9144891755002837e-05,
+      "loss": 1.1184,
+      "step": 3298
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9144261346553677e-05,
+      "loss": 1.2657,
+      "step": 3299
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9143630716200184e-05,
+      "loss": 1.1256,
+      "step": 3300
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9142999863957662e-05,
+      "loss": 1.1615,
+      "step": 3301
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9142368789841422e-05,
+      "loss": 1.2726,
+      "step": 3302
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914173749386678e-05,
+      "loss": 1.25,
+      "step": 3303
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9141105976049054e-05,
+      "loss": 1.2459,
+      "step": 3304
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914047423640357e-05,
+      "loss": 1.2734,
+      "step": 3305
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9139842274945655e-05,
+      "loss": 1.1109,
+      "step": 3306
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913921009169065e-05,
+      "loss": 1.2528,
+      "step": 3307
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9138577686653893e-05,
+      "loss": 1.2173,
+      "step": 3308
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913794505985073e-05,
+      "loss": 1.1426,
+      "step": 3309
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9137312211296516e-05,
+      "loss": 1.2068,
+      "step": 3310
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9136679141006603e-05,
+      "loss": 1.1644,
+      "step": 3311
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9136045848996357e-05,
+      "loss": 1.2306,
+      "step": 3312
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9135412335281152e-05,
+      "loss": 1.3253,
+      "step": 3313
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913477859987635e-05,
+      "loss": 1.1252,
+      "step": 3314
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913414464279734e-05,
+      "loss": 1.2578,
+      "step": 3315
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91335104640595e-05,
+      "loss": 1.2508,
+      "step": 3316
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913287606367822e-05,
+      "loss": 1.2227,
+      "step": 3317
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9132241441668903e-05,
+      "loss": 1.1364,
+      "step": 3318
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9131606598046936e-05,
+      "loss": 1.2034,
+      "step": 3319
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9130971532827737e-05,
+      "loss": 1.1027,
+      "step": 3320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9130336246026707e-05,
+      "loss": 1.154,
+      "step": 3321
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9129700737659273e-05,
+      "loss": 1.2753,
+      "step": 3322
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9129065007740848e-05,
+      "loss": 1.1605,
+      "step": 3323
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9128429056286865e-05,
+      "loss": 1.186,
+      "step": 3324
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9127792883312756e-05,
+      "loss": 1.1705,
+      "step": 3325
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912715648883395e-05,
+      "loss": 1.2477,
+      "step": 3326
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9126519872865906e-05,
+      "loss": 1.1042,
+      "step": 3327
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9125883035424062e-05,
+      "loss": 1.1624,
+      "step": 3328
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9125245976523876e-05,
+      "loss": 0.9719,
+      "step": 3329
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9124608696180806e-05,
+      "loss": 1.2039,
+      "step": 3330
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9123971194410317e-05,
+      "loss": 1.2707,
+      "step": 3331
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912333347122788e-05,
+      "loss": 1.0522,
+      "step": 3332
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9122695526648968e-05,
+      "loss": 1.1242,
+      "step": 3333
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912205736068907e-05,
+      "loss": 1.1511,
+      "step": 3334
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912141897336366e-05,
+      "loss": 1.1506,
+      "step": 3335
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9120780364688243e-05,
+      "loss": 1.3232,
+      "step": 3336
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912014153467831e-05,
+      "loss": 1.0641,
+      "step": 3337
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9119502483349357e-05,
+      "loss": 1.1216,
+      "step": 3338
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91188632107169e-05,
+      "loss": 1.2122,
+      "step": 3339
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9118223716796453e-05,
+      "loss": 1.2016,
+      "step": 3340
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9117584001603533e-05,
+      "loss": 1.1051,
+      "step": 3341
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911694406515366e-05,
+      "loss": 1.2011,
+      "step": 3342
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9116303907462365e-05,
+      "loss": 1.0688,
+      "step": 3343
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911566352854519e-05,
+      "loss": 1.2189,
+      "step": 3344
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9115022928417664e-05,
+      "loss": 1.2433,
+      "step": 3345
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911438210709534e-05,
+      "loss": 1.037,
+      "step": 3346
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911374106459377e-05,
+      "loss": 1.0597,
+      "step": 3347
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9113099800928502e-05,
+      "loss": 1.1888,
+      "step": 3348
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9112458316115107e-05,
+      "loss": 1.2887,
+      "step": 3349
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911181661016914e-05,
+      "loss": 1.1234,
+      "step": 3350
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911117468310619e-05,
+      "loss": 1.2142,
+      "step": 3351
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9110532534941822e-05,
+      "loss": 1.2028,
+      "step": 3352
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910989016569162e-05,
+      "loss": 1.1831,
+      "step": 3353
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9109247575371177e-05,
+      "loss": 1.1014,
+      "step": 3354
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9108604763996084e-05,
+      "loss": 1.341,
+      "step": 3355
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910796173158194e-05,
+      "loss": 1.2113,
+      "step": 3356
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910731847814435e-05,
+      "loss": 1.1851,
+      "step": 3357
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9106675003698928e-05,
+      "loss": 1.133,
+      "step": 3358
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9106031308261284e-05,
+      "loss": 1.255,
+      "step": 3359
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9105387391847036e-05,
+      "loss": 1.2478,
+      "step": 3360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910474325447182e-05,
+      "loss": 1.1798,
+      "step": 3361
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9104098896151256e-05,
+      "loss": 1.2144,
+      "step": 3362
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9103454316900987e-05,
+      "loss": 1.1502,
+      "step": 3363
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9102809516736655e-05,
+      "loss": 1.3644,
+      "step": 3364
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9102164495673906e-05,
+      "loss": 1.0878,
+      "step": 3365
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9101519253728396e-05,
+      "loss": 1.1928,
+      "step": 3366
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9100873790915776e-05,
+      "loss": 1.0927,
+      "step": 3367
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9100228107251715e-05,
+      "loss": 1.1321,
+      "step": 3368
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9099582202751885e-05,
+      "loss": 1.0456,
+      "step": 3369
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9098936077431953e-05,
+      "loss": 1.1143,
+      "step": 3370
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.90982897313076e-05,
+      "loss": 1.2764,
+      "step": 3371
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9097643164394512e-05,
+      "loss": 1.1834,
+      "step": 3372
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9096996376708382e-05,
+      "loss": 1.1895,
+      "step": 3373
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9096349368264904e-05,
+      "loss": 1.2206,
+      "step": 3374
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9095702139079774e-05,
+      "loss": 1.2382,
+      "step": 3375
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9095054689168707e-05,
+      "loss": 1.2557,
+      "step": 3376
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9094407018547406e-05,
+      "loss": 0.9818,
+      "step": 3377
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9093759127231594e-05,
+      "loss": 1.199,
+      "step": 3378
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9093111015236993e-05,
+      "loss": 1.1709,
+      "step": 3379
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909246268257933e-05,
+      "loss": 1.1533,
+      "step": 3380
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9091814129274335e-05,
+      "loss": 1.2505,
+      "step": 3381
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9091165355337754e-05,
+      "loss": 0.9819,
+      "step": 3382
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9090516360785322e-05,
+      "loss": 1.2625,
+      "step": 3383
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908986714563279e-05,
+      "loss": 1.0613,
+      "step": 3384
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9089217709895918e-05,
+      "loss": 1.2024,
+      "step": 3385
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908856805359046e-05,
+      "loss": 1.2197,
+      "step": 3386
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9087918176732188e-05,
+      "loss": 1.2116,
+      "step": 3387
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9087268079336865e-05,
+      "loss": 1.2401,
+      "step": 3388
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9086617761420272e-05,
+      "loss": 1.098,
+      "step": 3389
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908596722299819e-05,
+      "loss": 1.374,
+      "step": 3390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9085316464086403e-05,
+      "loss": 1.1525,
+      "step": 3391
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9084665484700704e-05,
+      "loss": 1.1794,
+      "step": 3392
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908401428485689e-05,
+      "loss": 1.0842,
+      "step": 3393
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9083362864570768e-05,
+      "loss": 1.1261,
+      "step": 3394
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9082711223858136e-05,
+      "loss": 1.2092,
+      "step": 3395
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9082059362734822e-05,
+      "loss": 1.1261,
+      "step": 3396
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908140728121663e-05,
+      "loss": 1.1195,
+      "step": 3397
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9080754979319395e-05,
+      "loss": 1.2854,
+      "step": 3398
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908010245705894e-05,
+      "loss": 1.2899,
+      "step": 3399
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.90794497144511e-05,
+      "loss": 1.1333,
+      "step": 3400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907879675151172e-05,
+      "loss": 1.1807,
+      "step": 3401
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9078143568256644e-05,
+      "loss": 1.1709,
+      "step": 3402
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907749016470172e-05,
+      "loss": 0.9926,
+      "step": 3403
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9076836540862804e-05,
+      "loss": 1.1627,
+      "step": 3404
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907618269675576e-05,
+      "loss": 1.3485,
+      "step": 3405
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9075528632396456e-05,
+      "loss": 1.2165,
+      "step": 3406
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907487434780076e-05,
+      "loss": 1.2199,
+      "step": 3407
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9074219842984553e-05,
+      "loss": 1.3653,
+      "step": 3408
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9073565117963714e-05,
+      "loss": 1.1968,
+      "step": 3409
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9072910172754137e-05,
+      "loss": 1.1255,
+      "step": 3410
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9072255007371715e-05,
+      "loss": 1.2069,
+      "step": 3411
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9071599621832338e-05,
+      "loss": 1.1914,
+      "step": 3412
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9070944016151923e-05,
+      "loss": 1.1026,
+      "step": 3413
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907028819034637e-05,
+      "loss": 1.2367,
+      "step": 3414
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9069632144431595e-05,
+      "loss": 1.3221,
+      "step": 3415
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9068975878423526e-05,
+      "loss": 1.0843,
+      "step": 3416
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9068319392338082e-05,
+      "loss": 1.1344,
+      "step": 3417
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9067662686191192e-05,
+      "loss": 1.1054,
+      "step": 3418
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9067005759998797e-05,
+      "loss": 1.04,
+      "step": 3419
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906634861377684e-05,
+      "loss": 1.2508,
+      "step": 3420
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9065691247541264e-05,
+      "loss": 1.1522,
+      "step": 3421
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9065033661308022e-05,
+      "loss": 1.2675,
+      "step": 3422
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906437585509307e-05,
+      "loss": 1.1158,
+      "step": 3423
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9063717828912376e-05,
+      "loss": 1.0988,
+      "step": 3424
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9063059582781905e-05,
+      "loss": 1.2915,
+      "step": 3425
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9062401116717635e-05,
+      "loss": 1.3321,
+      "step": 3426
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9061742430735538e-05,
+      "loss": 1.271,
+      "step": 3427
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.90610835248516e-05,
+      "loss": 1.0627,
+      "step": 3428
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9060424399081816e-05,
+      "loss": 1.2442,
+      "step": 3429
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9059765053442176e-05,
+      "loss": 1.1457,
+      "step": 3430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9059105487948683e-05,
+      "loss": 1.0638,
+      "step": 3431
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905844570261734e-05,
+      "loss": 1.0784,
+      "step": 3432
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9057785697464162e-05,
+      "loss": 1.1198,
+      "step": 3433
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9057125472505162e-05,
+      "loss": 1.0873,
+      "step": 3434
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9056465027756362e-05,
+      "loss": 1.2367,
+      "step": 3435
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905580436323379e-05,
+      "loss": 1.0793,
+      "step": 3436
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905514347895348e-05,
+      "loss": 1.1196,
+      "step": 3437
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905448237493147e-05,
+      "loss": 1.2443,
+      "step": 3438
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9053821051183797e-05,
+      "loss": 1.2951,
+      "step": 3439
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9053159507726514e-05,
+      "loss": 1.1806,
+      "step": 3440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9052497744575675e-05,
+      "loss": 1.2852,
+      "step": 3441
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9051835761747336e-05,
+      "loss": 1.2377,
+      "step": 3442
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9051173559257567e-05,
+      "loss": 1.1857,
+      "step": 3443
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905051113712243e-05,
+      "loss": 1.2589,
+      "step": 3444
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9049848495358006e-05,
+      "loss": 1.2252,
+      "step": 3445
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9049185633980376e-05,
+      "loss": 1.246,
+      "step": 3446
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904852255300562e-05,
+      "loss": 1.3069,
+      "step": 3447
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9047859252449837e-05,
+      "loss": 1.0189,
+      "step": 3448
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9047195732329117e-05,
+      "loss": 1.1753,
+      "step": 3449
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904653199265956e-05,
+      "loss": 1.2813,
+      "step": 3450
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904586803345728e-05,
+      "loss": 1.1049,
+      "step": 3451
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9045203854738386e-05,
+      "loss": 1.2012,
+      "step": 3452
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9044539456518992e-05,
+      "loss": 1.2509,
+      "step": 3453
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9043874838815225e-05,
+      "loss": 1.139,
+      "step": 3454
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9043210001643215e-05,
+      "loss": 1.2309,
+      "step": 3455
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9042544945019094e-05,
+      "loss": 1.1412,
+      "step": 3456
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9041879668959e-05,
+      "loss": 1.2925,
+      "step": 3457
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9041214173479075e-05,
+      "loss": 1.2291,
+      "step": 3458
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9040548458595472e-05,
+      "loss": 1.3096,
+      "step": 3459
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9039882524324346e-05,
+      "loss": 1.1962,
+      "step": 3460
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9039216370681858e-05,
+      "loss": 1.1568,
+      "step": 3461
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903854999768417e-05,
+      "loss": 1.1928,
+      "step": 3462
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9037883405347454e-05,
+      "loss": 1.3001,
+      "step": 3463
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903721659368789e-05,
+      "loss": 1.2346,
+      "step": 3464
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9036549562721657e-05,
+      "loss": 1.2419,
+      "step": 3465
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9035882312464938e-05,
+      "loss": 1.0541,
+      "step": 3466
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903521484293393e-05,
+      "loss": 1.1923,
+      "step": 3467
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9034547154144832e-05,
+      "loss": 1.2673,
+      "step": 3468
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9033879246113842e-05,
+      "loss": 1.0591,
+      "step": 3469
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903321111885717e-05,
+      "loss": 1.2061,
+      "step": 3470
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903254277239103e-05,
+      "loss": 1.1612,
+      "step": 3471
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9031874206731644e-05,
+      "loss": 1.1221,
+      "step": 3472
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903120542189523e-05,
+      "loss": 1.1557,
+      "step": 3473
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903053641789802e-05,
+      "loss": 1.192,
+      "step": 3474
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9029867194756248e-05,
+      "loss": 1.1436,
+      "step": 3475
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902919775248616e-05,
+      "loss": 1.2238,
+      "step": 3476
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902852809110399e-05,
+      "loss": 1.0817,
+      "step": 3477
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9027858210626e-05,
+      "loss": 1.1776,
+      "step": 3478
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902718811106844e-05,
+      "loss": 1.1117,
+      "step": 3479
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902651779244757e-05,
+      "loss": 1.0172,
+      "step": 3480
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9025847254779662e-05,
+      "loss": 0.9921,
+      "step": 3481
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9025176498080987e-05,
+      "loss": 1.213,
+      "step": 3482
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9024505522367818e-05,
+      "loss": 1.1442,
+      "step": 3483
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902383432765644e-05,
+      "loss": 1.2434,
+      "step": 3484
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9023162913963136e-05,
+      "loss": 1.2932,
+      "step": 3485
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9022491281304214e-05,
+      "loss": 1.1523,
+      "step": 3486
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9021819429695955e-05,
+      "loss": 1.535,
+      "step": 3487
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9021147359154674e-05,
+      "loss": 1.1789,
+      "step": 3488
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9020475069696676e-05,
+      "loss": 1.1521,
+      "step": 3489
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9019802561338278e-05,
+      "loss": 1.192,
+      "step": 3490
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9019129834095798e-05,
+      "loss": 1.0906,
+      "step": 3491
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9018456887985558e-05,
+      "loss": 1.1611,
+      "step": 3492
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9017783723023895e-05,
+      "loss": 1.0572,
+      "step": 3493
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901711033922714e-05,
+      "loss": 1.2547,
+      "step": 3494
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9016436736611637e-05,
+      "loss": 1.1273,
+      "step": 3495
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9015762915193727e-05,
+      "loss": 1.227,
+      "step": 3496
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901508887498977e-05,
+      "loss": 1.0682,
+      "step": 3497
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9014414616016116e-05,
+      "loss": 1.1764,
+      "step": 3498
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901374013828913e-05,
+      "loss": 1.1742,
+      "step": 3499
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901306544182518e-05,
+      "loss": 1.1985,
+      "step": 3500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9012390526640638e-05,
+      "loss": 1.2641,
+      "step": 3501
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9011715392751882e-05,
+      "loss": 1.2593,
+      "step": 3502
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9011040040175295e-05,
+      "loss": 1.2141,
+      "step": 3503
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9010364468927267e-05,
+      "loss": 1.2382,
+      "step": 3504
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900968867902419e-05,
+      "loss": 1.0984,
+      "step": 3505
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900901267048247e-05,
+      "loss": 1.1462,
+      "step": 3506
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9008336443318505e-05,
+      "loss": 1.2858,
+      "step": 3507
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900765999754871e-05,
+      "loss": 1.2279,
+      "step": 3508
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9006983333189493e-05,
+      "loss": 1.2152,
+      "step": 3509
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9006306450257278e-05,
+      "loss": 1.2086,
+      "step": 3510
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.90056293487685e-05,
+      "loss": 1.2334,
+      "step": 3511
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900495202873957e-05,
+      "loss": 1.1939,
+      "step": 3512
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9004274490186944e-05,
+      "loss": 1.0403,
+      "step": 3513
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9003596733127058e-05,
+      "loss": 1.2437,
+      "step": 3514
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9002918757576358e-05,
+      "loss": 1.1733,
+      "step": 3515
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9002240563551293e-05,
+      "loss": 1.0655,
+      "step": 3516
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9001562151068327e-05,
+      "loss": 1.1501,
+      "step": 3517
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900088352014392e-05,
+      "loss": 1.1678,
+      "step": 3518
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900020467079454e-05,
+      "loss": 1.2396,
+      "step": 3519
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899952560303666e-05,
+      "loss": 1.283,
+      "step": 3520
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899884631688676e-05,
+      "loss": 1.2762,
+      "step": 3521
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8998166812361325e-05,
+      "loss": 1.1886,
+      "step": 3522
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8997487089476844e-05,
+      "loss": 1.1827,
+      "step": 3523
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8996807148249817e-05,
+      "loss": 0.9684,
+      "step": 3524
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899612698869673e-05,
+      "loss": 1.2616,
+      "step": 3525
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8995446610834104e-05,
+      "loss": 1.1553,
+      "step": 3526
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8994766014678444e-05,
+      "loss": 1.2124,
+      "step": 3527
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8994085200246263e-05,
+      "loss": 1.1189,
+      "step": 3528
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8993404167554085e-05,
+      "loss": 1.171,
+      "step": 3529
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899272291661844e-05,
+      "loss": 1.1217,
+      "step": 3530
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899204144745585e-05,
+      "loss": 1.2381,
+      "step": 3531
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8991359760082864e-05,
+      "loss": 1.1251,
+      "step": 3532
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8990677854516017e-05,
+      "loss": 1.1338,
+      "step": 3533
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898999573077186e-05,
+      "loss": 1.1443,
+      "step": 3534
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8989313388866944e-05,
+      "loss": 1.2698,
+      "step": 3535
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8988630828817827e-05,
+      "loss": 1.2154,
+      "step": 3536
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898794805064108e-05,
+      "loss": 1.1696,
+      "step": 3537
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8987265054353262e-05,
+      "loss": 1.191,
+      "step": 3538
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8986581839970956e-05,
+      "loss": 1.2911,
+      "step": 3539
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898589840751073e-05,
+      "loss": 1.1065,
+      "step": 3540
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898521475698918e-05,
+      "loss": 1.196,
+      "step": 3541
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8984530888422897e-05,
+      "loss": 1.1464,
+      "step": 3542
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8983846801828468e-05,
+      "loss": 1.198,
+      "step": 3543
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8983162497222497e-05,
+      "loss": 1.0551,
+      "step": 3544
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898247797462159e-05,
+      "loss": 1.0449,
+      "step": 3545
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8981793234042362e-05,
+      "loss": 1.143,
+      "step": 3546
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8981108275501422e-05,
+      "loss": 1.1743,
+      "step": 3547
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8980423099015402e-05,
+      "loss": 1.2105,
+      "step": 3548
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8979737704600923e-05,
+      "loss": 1.0826,
+      "step": 3549
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8979052092274615e-05,
+      "loss": 1.191,
+      "step": 3550
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897836626205312e-05,
+      "loss": 1.142,
+      "step": 3551
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8977680213953084e-05,
+      "loss": 1.0873,
+      "step": 3552
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8976993947991145e-05,
+      "loss": 1.2868,
+      "step": 3553
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897630746418397e-05,
+      "loss": 1.1202,
+      "step": 3554
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8975620762548207e-05,
+      "loss": 1.3243,
+      "step": 3555
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8974933843100526e-05,
+      "loss": 1.0946,
+      "step": 3556
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8974246705857594e-05,
+      "loss": 1.1547,
+      "step": 3557
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8973559350836092e-05,
+      "loss": 1.2328,
+      "step": 3558
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8972871778052688e-05,
+      "loss": 1.2573,
+      "step": 3559
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8972183987524076e-05,
+      "loss": 1.1196,
+      "step": 3560
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8971495979266946e-05,
+      "loss": 1.2469,
+      "step": 3561
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897080775329799e-05,
+      "loss": 1.1655,
+      "step": 3562
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8970119309633916e-05,
+      "loss": 1.3114,
+      "step": 3563
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8969430648291425e-05,
+      "loss": 1.1713,
+      "step": 3564
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896874176928723e-05,
+      "loss": 1.1794,
+      "step": 3565
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896805267263805e-05,
+      "loss": 1.1653,
+      "step": 3566
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89673633583606e-05,
+      "loss": 1.0665,
+      "step": 3567
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896667382647162e-05,
+      "loss": 1.2308,
+      "step": 3568
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8965984076987835e-05,
+      "loss": 1.2177,
+      "step": 3569
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8965294109925984e-05,
+      "loss": 1.1643,
+      "step": 3570
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8964603925302813e-05,
+      "loss": 1.171,
+      "step": 3571
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8963913523135062e-05,
+      "loss": 1.2014,
+      "step": 3572
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8963222903439495e-05,
+      "loss": 1.1635,
+      "step": 3573
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896253206623287e-05,
+      "loss": 1.2456,
+      "step": 3574
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8961841011531948e-05,
+      "loss": 1.1797,
+      "step": 3575
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8961149739353502e-05,
+      "loss": 1.1534,
+      "step": 3576
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8960458249714303e-05,
+      "loss": 1.0889,
+      "step": 3577
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8959766542631137e-05,
+      "loss": 1.122,
+      "step": 3578
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8959074618120784e-05,
+      "loss": 1.2546,
+      "step": 3579
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8958382476200038e-05,
+      "loss": 1.2603,
+      "step": 3580
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8957690116885697e-05,
+      "loss": 1.2346,
+      "step": 3581
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895699754019456e-05,
+      "loss": 1.2703,
+      "step": 3582
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8956304746143433e-05,
+      "loss": 1.1821,
+      "step": 3583
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895561173474913e-05,
+      "loss": 1.0911,
+      "step": 3584
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8954918506028467e-05,
+      "loss": 1.2348,
+      "step": 3585
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895422505999827e-05,
+      "loss": 1.0977,
+      "step": 3586
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895353139667536e-05,
+      "loss": 1.2288,
+      "step": 3587
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895283751607658e-05,
+      "loss": 1.1764,
+      "step": 3588
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895214341821876e-05,
+      "loss": 1.1284,
+      "step": 3589
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8951449103118743e-05,
+      "loss": 1.2288,
+      "step": 3590
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8950754570793384e-05,
+      "loss": 1.0919,
+      "step": 3591
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8950059821259535e-05,
+      "loss": 1.225,
+      "step": 3592
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8949364854534054e-05,
+      "loss": 1.2633,
+      "step": 3593
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8948669670633812e-05,
+      "loss": 1.3047,
+      "step": 3594
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894797426957567e-05,
+      "loss": 1.3258,
+      "step": 3595
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894727865137651e-05,
+      "loss": 1.1709,
+      "step": 3596
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894658281605321e-05,
+      "loss": 1.2049,
+      "step": 3597
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894588676362265e-05,
+      "loss": 1.2363,
+      "step": 3598
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8945190494101734e-05,
+      "loss": 1.1828,
+      "step": 3599
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894449400750735e-05,
+      "loss": 1.3318,
+      "step": 3600
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89437973038564e-05,
+      "loss": 1.3169,
+      "step": 3601
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8943100383165794e-05,
+      "loss": 0.9883,
+      "step": 3602
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894240324545244e-05,
+      "loss": 1.183,
+      "step": 3603
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894170589073326e-05,
+      "loss": 1.279,
+      "step": 3604
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8941008319025174e-05,
+      "loss": 1.227,
+      "step": 3605
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894031053034511e-05,
+      "loss": 1.3268,
+      "step": 3606
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8939612524710003e-05,
+      "loss": 1.1717,
+      "step": 3607
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8938914302136792e-05,
+      "loss": 0.9768,
+      "step": 3608
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893821586264241e-05,
+      "loss": 1.1836,
+      "step": 3609
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8937517206243828e-05,
+      "loss": 1.062,
+      "step": 3610
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8936818332957983e-05,
+      "loss": 1.2149,
+      "step": 3611
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8936119242801837e-05,
+      "loss": 1.2158,
+      "step": 3612
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8935419935792358e-05,
+      "loss": 1.2848,
+      "step": 3613
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8934720411946513e-05,
+      "loss": 1.0534,
+      "step": 3614
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8934020671281285e-05,
+      "loss": 1.1941,
+      "step": 3615
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8933320713813647e-05,
+      "loss": 1.1002,
+      "step": 3616
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8932620539560587e-05,
+      "loss": 1.2879,
+      "step": 3617
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8931920148539098e-05,
+      "loss": 1.1139,
+      "step": 3618
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8931219540766174e-05,
+      "loss": 1.1176,
+      "step": 3619
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8930518716258816e-05,
+      "loss": 1.23,
+      "step": 3620
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8929817675034033e-05,
+      "loss": 1.1046,
+      "step": 3621
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8929116417108837e-05,
+      "loss": 1.1584,
+      "step": 3622
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8928414942500243e-05,
+      "loss": 1.1087,
+      "step": 3623
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892771325122528e-05,
+      "loss": 1.047,
+      "step": 3624
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892701134330097e-05,
+      "loss": 1.1532,
+      "step": 3625
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8926309218744348e-05,
+      "loss": 1.1242,
+      "step": 3626
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892560687757245e-05,
+      "loss": 1.1746,
+      "step": 3627
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8924904319802327e-05,
+      "loss": 1.1094,
+      "step": 3628
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892420154545102e-05,
+      "loss": 1.2441,
+      "step": 3629
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892349855453559e-05,
+      "loss": 1.0145,
+      "step": 3630
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892279534707309e-05,
+      "loss": 1.3279,
+      "step": 3631
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892209192308059e-05,
+      "loss": 1.2117,
+      "step": 3632
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8921388282575157e-05,
+      "loss": 1.249,
+      "step": 3633
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8920684425573865e-05,
+      "loss": 1.0822,
+      "step": 3634
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8919980352093802e-05,
+      "loss": 1.2328,
+      "step": 3635
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8919276062152043e-05,
+      "loss": 1.2053,
+      "step": 3636
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891857155576569e-05,
+      "loss": 1.1405,
+      "step": 3637
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891786683295183e-05,
+      "loss": 1.2918,
+      "step": 3638
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891716189372757e-05,
+      "loss": 1.1533,
+      "step": 3639
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8916456738110013e-05,
+      "loss": 1.186,
+      "step": 3640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8915751366116275e-05,
+      "loss": 1.2665,
+      "step": 3641
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891504577776347e-05,
+      "loss": 1.1337,
+      "step": 3642
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8914339973068725e-05,
+      "loss": 1.2349,
+      "step": 3643
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8913633952049168e-05,
+      "loss": 1.1455,
+      "step": 3644
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8912927714721922e-05,
+      "loss": 1.1313,
+      "step": 3645
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8912221261104136e-05,
+      "loss": 1.138,
+      "step": 3646
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891151459121295e-05,
+      "loss": 1.2859,
+      "step": 3647
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8910807705065514e-05,
+      "loss": 1.2797,
+      "step": 3648
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8910100602678976e-05,
+      "loss": 1.2662,
+      "step": 3649
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.89093932840705e-05,
+      "loss": 1.1547,
+      "step": 3650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8908685749257257e-05,
+      "loss": 1.2115,
+      "step": 3651
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8907977998256405e-05,
+      "loss": 1.1226,
+      "step": 3652
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890727003108513e-05,
+      "loss": 1.1936,
+      "step": 3653
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.89065618477606e-05,
+      "loss": 1.2553,
+      "step": 3654
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890585344830001e-05,
+      "loss": 1.2414,
+      "step": 3655
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8905144832720547e-05,
+      "loss": 1.3447,
+      "step": 3656
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8904436001039408e-05,
+      "loss": 1.2282,
+      "step": 3657
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8903726953273794e-05,
+      "loss": 1.1351,
+      "step": 3658
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8903017689440914e-05,
+      "loss": 1.2844,
+      "step": 3659
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8902308209557976e-05,
+      "loss": 1.2823,
+      "step": 3660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.89015985136422e-05,
+      "loss": 1.1513,
+      "step": 3661
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8900888601710804e-05,
+      "loss": 1.3353,
+      "step": 3662
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890017847378102e-05,
+      "loss": 1.2821,
+      "step": 3663
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889946812987008e-05,
+      "loss": 1.1699,
+      "step": 3664
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8898757569995218e-05,
+      "loss": 1.0956,
+      "step": 3665
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8898046794173684e-05,
+      "loss": 1.0833,
+      "step": 3666
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889733580242272e-05,
+      "loss": 1.1437,
+      "step": 3667
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8896624594759583e-05,
+      "loss": 1.3514,
+      "step": 3668
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889591317120153e-05,
+      "loss": 1.2427,
+      "step": 3669
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889520153176583e-05,
+      "loss": 1.2084,
+      "step": 3670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889448967646974e-05,
+      "loss": 1.1839,
+      "step": 3671
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8893777605330553e-05,
+      "loss": 1.0656,
+      "step": 3672
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8893065318365534e-05,
+      "loss": 1.1455,
+      "step": 3673
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8892352815591973e-05,
+      "loss": 1.1841,
+      "step": 3674
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8891640097027163e-05,
+      "loss": 1.055,
+      "step": 3675
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8890927162688398e-05,
+      "loss": 1.1958,
+      "step": 3676
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8890214012592977e-05,
+      "loss": 1.2054,
+      "step": 3677
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8889500646758203e-05,
+      "loss": 1.3127,
+      "step": 3678
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8888787065201393e-05,
+      "loss": 1.3344,
+      "step": 3679
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8888073267939865e-05,
+      "loss": 1.2004,
+      "step": 3680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8887359254990937e-05,
+      "loss": 1.1277,
+      "step": 3681
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8886645026371937e-05,
+      "loss": 1.287,
+      "step": 3682
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8885930582100195e-05,
+      "loss": 1.1791,
+      "step": 3683
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888521592219305e-05,
+      "loss": 1.1941,
+      "step": 3684
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8884501046667847e-05,
+      "loss": 1.1538,
+      "step": 3685
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888378595554193e-05,
+      "loss": 1.2613,
+      "step": 3686
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888307064883266e-05,
+      "loss": 1.062,
+      "step": 3687
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8882355126557382e-05,
+      "loss": 1.2345,
+      "step": 3688
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8881639388733468e-05,
+      "loss": 1.1252,
+      "step": 3689
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888092343537829e-05,
+      "loss": 1.1961,
+      "step": 3690
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8880207266509215e-05,
+      "loss": 1.2194,
+      "step": 3691
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887949088214363e-05,
+      "loss": 1.2627,
+      "step": 3692
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887877428229891e-05,
+      "loss": 1.0544,
+      "step": 3693
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887805746699245e-05,
+      "loss": 1.2423,
+      "step": 3694
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8877340436241645e-05,
+      "loss": 1.1496,
+      "step": 3695
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8876623190063898e-05,
+      "loss": 1.0545,
+      "step": 3696
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887590572847661e-05,
+      "loss": 1.1396,
+      "step": 3697
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887518805149719e-05,
+      "loss": 1.269,
+      "step": 3698
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887447015914306e-05,
+      "loss": 1.0734,
+      "step": 3699
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8873752051431635e-05,
+      "loss": 1.2629,
+      "step": 3700
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887303372838035e-05,
+      "loss": 1.3362,
+      "step": 3701
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8872315190006624e-05,
+      "loss": 1.0944,
+      "step": 3702
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8871596436327908e-05,
+      "loss": 0.9815,
+      "step": 3703
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8870877467361633e-05,
+      "loss": 1.3585,
+      "step": 3704
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887015828312525e-05,
+      "loss": 1.1769,
+      "step": 3705
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8869438883636212e-05,
+      "loss": 1.2953,
+      "step": 3706
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8868719268911978e-05,
+      "loss": 1.1437,
+      "step": 3707
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8867999438970007e-05,
+      "loss": 1.1831,
+      "step": 3708
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886727939382777e-05,
+      "loss": 0.977,
+      "step": 3709
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886655913350274e-05,
+      "loss": 1.1104,
+      "step": 3710
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8865838658012397e-05,
+      "loss": 1.2582,
+      "step": 3711
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8865117967374225e-05,
+      "loss": 1.2508,
+      "step": 3712
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886439706160571e-05,
+      "loss": 1.2474,
+      "step": 3713
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8863675940724344e-05,
+      "loss": 1.127,
+      "step": 3714
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886295460474763e-05,
+      "loss": 1.2135,
+      "step": 3715
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8862233053693074e-05,
+      "loss": 1.3007,
+      "step": 3716
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8861511287578188e-05,
+      "loss": 1.2621,
+      "step": 3717
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8860789306420478e-05,
+      "loss": 1.2695,
+      "step": 3718
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8860067110237478e-05,
+      "loss": 1.1913,
+      "step": 3719
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88593446990467e-05,
+      "loss": 1.0179,
+      "step": 3720
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885862207286568e-05,
+      "loss": 1.199,
+      "step": 3721
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8857899231711956e-05,
+      "loss": 1.1522,
+      "step": 3722
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885717617560307e-05,
+      "loss": 1.0553,
+      "step": 3723
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8856452904556564e-05,
+      "loss": 1.3076,
+      "step": 3724
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8855729418589994e-05,
+      "loss": 1.2537,
+      "step": 3725
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885500571772091e-05,
+      "loss": 1.0627,
+      "step": 3726
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8854281801966886e-05,
+      "loss": 1.1292,
+      "step": 3727
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8853557671345477e-05,
+      "loss": 1.3002,
+      "step": 3728
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885283332587426e-05,
+      "loss": 1.2438,
+      "step": 3729
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8852108765570814e-05,
+      "loss": 1.0864,
+      "step": 3730
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8851383990452726e-05,
+      "loss": 1.2005,
+      "step": 3731
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8850659000537575e-05,
+      "loss": 1.0112,
+      "step": 3732
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884993379584296e-05,
+      "loss": 1.1928,
+      "step": 3733
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8849208376386477e-05,
+      "loss": 1.1765,
+      "step": 3734
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8848482742185737e-05,
+      "loss": 1.4024,
+      "step": 3735
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8847756893258336e-05,
+      "loss": 1.2585,
+      "step": 3736
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88470308296219e-05,
+      "loss": 1.1948,
+      "step": 3737
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884630455129404e-05,
+      "loss": 1.2587,
+      "step": 3738
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8845578058292392e-05,
+      "loss": 1.2184,
+      "step": 3739
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8844851350634573e-05,
+      "loss": 1.2282,
+      "step": 3740
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8844124428338223e-05,
+      "loss": 1.2859,
+      "step": 3741
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8843397291420982e-05,
+      "loss": 1.2834,
+      "step": 3742
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88426699399005e-05,
+      "loss": 1.1539,
+      "step": 3743
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884194237379442e-05,
+      "loss": 1.0452,
+      "step": 3744
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8841214593120405e-05,
+      "loss": 1.3026,
+      "step": 3745
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8840486597896114e-05,
+      "loss": 1.1281,
+      "step": 3746
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8839758388139208e-05,
+      "loss": 1.2591,
+      "step": 3747
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8839029963867362e-05,
+      "loss": 1.2181,
+      "step": 3748
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883830132509826e-05,
+      "loss": 1.2327,
+      "step": 3749
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8837572471849574e-05,
+      "loss": 1.1926,
+      "step": 3750
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8836843404138992e-05,
+      "loss": 1.1939,
+      "step": 3751
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883611412198421e-05,
+      "loss": 1.2265,
+      "step": 3752
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8835384625402927e-05,
+      "loss": 1.2009,
+      "step": 3753
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883465491441284e-05,
+      "loss": 1.0491,
+      "step": 3754
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8833924989031663e-05,
+      "loss": 1.2305,
+      "step": 3755
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8833194849277105e-05,
+      "loss": 1.1929,
+      "step": 3756
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8832464495166883e-05,
+      "loss": 1.2367,
+      "step": 3757
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8831733926718725e-05,
+      "loss": 1.2586,
+      "step": 3758
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8831003143950357e-05,
+      "loss": 1.164,
+      "step": 3759
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8830272146879513e-05,
+      "loss": 1.2638,
+      "step": 3760
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882954093552394e-05,
+      "loss": 1.1956,
+      "step": 3761
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8828809509901366e-05,
+      "loss": 1.2334,
+      "step": 3762
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8828077870029554e-05,
+      "loss": 1.2494,
+      "step": 3763
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8827346015926253e-05,
+      "loss": 1.07,
+      "step": 3764
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8826613947609225e-05,
+      "loss": 1.2337,
+      "step": 3765
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8825881665096237e-05,
+      "loss": 1.0361,
+      "step": 3766
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8825149168405055e-05,
+      "loss": 1.2337,
+      "step": 3767
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8824416457553455e-05,
+      "loss": 1.28,
+      "step": 3768
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882368353255922e-05,
+      "loss": 1.2947,
+      "step": 3769
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8822950393440135e-05,
+      "loss": 1.1131,
+      "step": 3770
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8822217040213994e-05,
+      "loss": 1.0235,
+      "step": 3771
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882148347289859e-05,
+      "loss": 1.1812,
+      "step": 3772
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8820749691511723e-05,
+      "loss": 1.1449,
+      "step": 3773
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8820015696071202e-05,
+      "loss": 1.2145,
+      "step": 3774
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881928148659484e-05,
+      "loss": 1.2203,
+      "step": 3775
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8818547063100453e-05,
+      "loss": 1.0296,
+      "step": 3776
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8817812425605864e-05,
+      "loss": 1.2592,
+      "step": 3777
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88170775741289e-05,
+      "loss": 1.1979,
+      "step": 3778
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8816342508687388e-05,
+      "loss": 1.2312,
+      "step": 3779
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8815607229299177e-05,
+      "loss": 1.1554,
+      "step": 3780
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8814871735982102e-05,
+      "loss": 1.1371,
+      "step": 3781
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8814136028754016e-05,
+      "loss": 1.1505,
+      "step": 3782
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8813400107632763e-05,
+      "loss": 1.1711,
+      "step": 3783
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881266397263622e-05,
+      "loss": 1.2178,
+      "step": 3784
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8811927623782227e-05,
+      "loss": 1.2369,
+      "step": 3785
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8811191061088673e-05,
+      "loss": 1.1447,
+      "step": 3786
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8810454284573424e-05,
+      "loss": 1.2039,
+      "step": 3787
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880971729425436e-05,
+      "loss": 1.0414,
+      "step": 3788
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880898009014937e-05,
+      "loss": 1.0484,
+      "step": 3789
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880824267227633e-05,
+      "loss": 1.3357,
+      "step": 3790
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880750504065315e-05,
+      "loss": 1.1318,
+      "step": 3791
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8806767195297724e-05,
+      "loss": 1.2625,
+      "step": 3792
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880602913622796e-05,
+      "loss": 1.2772,
+      "step": 3793
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880529086346176e-05,
+      "loss": 1.1013,
+      "step": 3794
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880455237701705e-05,
+      "loss": 1.186,
+      "step": 3795
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880381367691175e-05,
+      "loss": 1.1762,
+      "step": 3796
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880307476316378e-05,
+      "loss": 1.107,
+      "step": 3797
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8802335635791076e-05,
+      "loss": 1.1336,
+      "step": 3798
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880159629481157e-05,
+      "loss": 1.1626,
+      "step": 3799
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880085674024321e-05,
+      "loss": 1.2118,
+      "step": 3800
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8800116972103935e-05,
+      "loss": 1.1361,
+      "step": 3801
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8799376990411708e-05,
+      "loss": 0.9825,
+      "step": 3802
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8798636795184473e-05,
+      "loss": 1.1588,
+      "step": 3803
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8797896386440202e-05,
+      "loss": 1.2079,
+      "step": 3804
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879715576419686e-05,
+      "loss": 1.1136,
+      "step": 3805
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8796414928472417e-05,
+      "loss": 1.2237,
+      "step": 3806
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879567387928485e-05,
+      "loss": 1.2165,
+      "step": 3807
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8794932616652152e-05,
+      "loss": 1.2659,
+      "step": 3808
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8794191140592303e-05,
+      "loss": 1.1511,
+      "step": 3809
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8793449451123296e-05,
+      "loss": 1.2432,
+      "step": 3810
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879270754826313e-05,
+      "loss": 1.1294,
+      "step": 3811
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8791965432029812e-05,
+      "loss": 1.2654,
+      "step": 3812
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8791223102441347e-05,
+      "loss": 1.0496,
+      "step": 3813
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8790480559515756e-05,
+      "loss": 1.2513,
+      "step": 3814
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878973780327105e-05,
+      "loss": 1.0757,
+      "step": 3815
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878899483372526e-05,
+      "loss": 1.0702,
+      "step": 3816
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8788251650896407e-05,
+      "loss": 1.0811,
+      "step": 3817
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8787508254802538e-05,
+      "loss": 1.2287,
+      "step": 3818
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8786764645461684e-05,
+      "loss": 1.1162,
+      "step": 3819
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8786020822891892e-05,
+      "loss": 1.1845,
+      "step": 3820
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8785276787111216e-05,
+      "loss": 1.3048,
+      "step": 3821
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878453253813771e-05,
+      "loss": 1.0748,
+      "step": 3822
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878378807598943e-05,
+      "loss": 1.1507,
+      "step": 3823
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8783043400684447e-05,
+      "loss": 1.14,
+      "step": 3824
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878229851224083e-05,
+      "loss": 1.2136,
+      "step": 3825
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8781553410676658e-05,
+      "loss": 1.1633,
+      "step": 3826
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878080809601001e-05,
+      "loss": 1.2199,
+      "step": 3827
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878006256825897e-05,
+      "loss": 1.2199,
+      "step": 3828
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8779316827441636e-05,
+      "loss": 1.1193,
+      "step": 3829
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.87785708735761e-05,
+      "loss": 1.1332,
+      "step": 3830
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8777824706680466e-05,
+      "loss": 1.2954,
+      "step": 3831
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8777078326772843e-05,
+      "loss": 1.1387,
+      "step": 3832
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877633173387134e-05,
+      "loss": 1.2365,
+      "step": 3833
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8775584927994074e-05,
+      "loss": 1.4377,
+      "step": 3834
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877483790915917e-05,
+      "loss": 1.2519,
+      "step": 3835
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877409067738476e-05,
+      "loss": 1.1391,
+      "step": 3836
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877334323268897e-05,
+      "loss": 1.2087,
+      "step": 3837
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8772595575089943e-05,
+      "loss": 1.1257,
+      "step": 3838
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8771847704605818e-05,
+      "loss": 1.1498,
+      "step": 3839
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8771099621254748e-05,
+      "loss": 1.1952,
+      "step": 3840
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8770351325054882e-05,
+      "loss": 1.0169,
+      "step": 3841
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.876960281602439e-05,
+      "loss": 1.2732,
+      "step": 3842
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.876885409418142e-05,
+      "loss": 1.1576,
+      "step": 3843
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8768105159544152e-05,
+      "loss": 1.1214,
+      "step": 3844
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8767356012130758e-05,
+      "loss": 1.3994,
+      "step": 3845
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8766606651959417e-05,
+      "loss": 1.1189,
+      "step": 3846
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876585707904831e-05,
+      "loss": 1.1589,
+      "step": 3847
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876510729341564e-05,
+      "loss": 1.1592,
+      "step": 3848
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876435729507959e-05,
+      "loss": 1.0979,
+      "step": 3849
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876360708405836e-05,
+      "loss": 1.1941,
+      "step": 3850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8762856660370165e-05,
+      "loss": 1.2489,
+      "step": 3851
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8762106024033206e-05,
+      "loss": 1.1065,
+      "step": 3852
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87613551750657e-05,
+      "loss": 1.2855,
+      "step": 3853
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8760604113485873e-05,
+      "loss": 1.282,
+      "step": 3854
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8759852839311946e-05,
+      "loss": 1.179,
+      "step": 3855
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8759101352562154e-05,
+      "loss": 1.1235,
+      "step": 3856
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8758349653254733e-05,
+      "loss": 1.1566,
+      "step": 3857
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8757597741407922e-05,
+      "loss": 1.24,
+      "step": 3858
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875684561703997e-05,
+      "loss": 1.303,
+      "step": 3859
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8756093280169126e-05,
+      "loss": 1.1713,
+      "step": 3860
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875534073081365e-05,
+      "loss": 1.2548,
+      "step": 3861
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8754587968991803e-05,
+      "loss": 1.1559,
+      "step": 3862
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8753834994721852e-05,
+      "loss": 1.2346,
+      "step": 3863
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8753081808022065e-05,
+      "loss": 1.1217,
+      "step": 3864
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8752328408910732e-05,
+      "loss": 1.0864,
+      "step": 3865
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8751574797406124e-05,
+      "loss": 1.1465,
+      "step": 3866
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8750820973526535e-05,
+      "loss": 1.277,
+      "step": 3867
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8750066937290256e-05,
+      "loss": 1.0621,
+      "step": 3868
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8749312688715587e-05,
+      "loss": 1.2286,
+      "step": 3869
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8748558227820828e-05,
+      "loss": 1.2905,
+      "step": 3870
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8747803554624287e-05,
+      "loss": 1.2032,
+      "step": 3871
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874704866914428e-05,
+      "loss": 1.1882,
+      "step": 3872
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874629357139913e-05,
+      "loss": 1.144,
+      "step": 3873
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8745538261407157e-05,
+      "loss": 1.1879,
+      "step": 3874
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8744782739186688e-05,
+      "loss": 1.264,
+      "step": 3875
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874402700475606e-05,
+      "loss": 1.0686,
+      "step": 3876
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874327105813361e-05,
+      "loss": 0.9601,
+      "step": 3877
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874251489933769e-05,
+      "loss": 1.2101,
+      "step": 3878
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874175852838664e-05,
+      "loss": 1.1621,
+      "step": 3879
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8741001945298817e-05,
+      "loss": 1.1866,
+      "step": 3880
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8740245150092585e-05,
+      "loss": 1.2748,
+      "step": 3881
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873948814278631e-05,
+      "loss": 1.2336,
+      "step": 3882
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8738730923398357e-05,
+      "loss": 1.2858,
+      "step": 3883
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8737973491947102e-05,
+      "loss": 1.1031,
+      "step": 3884
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8737215848450933e-05,
+      "loss": 1.1029,
+      "step": 3885
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8736457992928228e-05,
+      "loss": 1.2749,
+      "step": 3886
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873569992539738e-05,
+      "loss": 1.2362,
+      "step": 3887
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8734941645876786e-05,
+      "loss": 1.1532,
+      "step": 3888
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8734183154384848e-05,
+      "loss": 1.2593,
+      "step": 3889
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873342445093997e-05,
+      "loss": 1.294,
+      "step": 3890
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8732665535560564e-05,
+      "loss": 1.0344,
+      "step": 3891
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873190640826505e-05,
+      "loss": 1.1729,
+      "step": 3892
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8731147069071843e-05,
+      "loss": 1.2553,
+      "step": 3893
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8730387517999378e-05,
+      "loss": 1.0815,
+      "step": 3894
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8729627755066082e-05,
+      "loss": 1.1248,
+      "step": 3895
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8728867780290393e-05,
+      "loss": 1.3361,
+      "step": 3896
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8728107593690753e-05,
+      "loss": 0.9665,
+      "step": 3897
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872734719528561e-05,
+      "loss": 1.1878,
+      "step": 3898
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872658658509342e-05,
+      "loss": 1.1473,
+      "step": 3899
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872582576313263e-05,
+      "loss": 1.0956,
+      "step": 3900
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8725064729421717e-05,
+      "loss": 1.1841,
+      "step": 3901
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872430348397914e-05,
+      "loss": 1.2082,
+      "step": 3902
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8723542026823375e-05,
+      "loss": 1.1347,
+      "step": 3903
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8722780357972903e-05,
+      "loss": 1.1271,
+      "step": 3904
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87220184774462e-05,
+      "loss": 1.0703,
+      "step": 3905
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872125638526176e-05,
+      "loss": 1.3082,
+      "step": 3906
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872049408143808e-05,
+      "loss": 1.1728,
+      "step": 3907
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8719731565993647e-05,
+      "loss": 1.2725,
+      "step": 3908
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871896883894698e-05,
+      "loss": 1.0957,
+      "step": 3909
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8718205900316578e-05,
+      "loss": 1.0266,
+      "step": 3910
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8717442750120956e-05,
+      "loss": 1.2888,
+      "step": 3911
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871667938837864e-05,
+      "loss": 1.1391,
+      "step": 3912
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871591581510815e-05,
+      "loss": 1.1048,
+      "step": 3913
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871515203032801e-05,
+      "loss": 1.1535,
+      "step": 3914
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8714388034056764e-05,
+      "loss": 1.2074,
+      "step": 3915
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871362382631295e-05,
+      "loss": 1.2631,
+      "step": 3916
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871285940711511e-05,
+      "loss": 1.1708,
+      "step": 3917
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8712094776481798e-05,
+      "loss": 1.0536,
+      "step": 3918
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8711329934431563e-05,
+      "loss": 1.0617,
+      "step": 3919
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8710564880982975e-05,
+      "loss": 1.2707,
+      "step": 3920
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8709799616154587e-05,
+      "loss": 1.128,
+      "step": 3921
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870903413996498e-05,
+      "loss": 1.1668,
+      "step": 3922
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870826845243273e-05,
+      "loss": 1.1013,
+      "step": 3923
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870750255357641e-05,
+      "loss": 1.2458,
+      "step": 3924
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8706736443414616e-05,
+      "loss": 1.2369,
+      "step": 3925
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870597012196593e-05,
+      "loss": 1.1157,
+      "step": 3926
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8705203589248953e-05,
+      "loss": 1.001,
+      "step": 3927
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8704436845282288e-05,
+      "loss": 1.2492,
+      "step": 3928
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8703669890084536e-05,
+      "loss": 1.3697,
+      "step": 3929
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8702902723674317e-05,
+      "loss": 1.3108,
+      "step": 3930
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870213534607024e-05,
+      "loss": 1.2081,
+      "step": 3931
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870136775729093e-05,
+      "loss": 0.9999,
+      "step": 3932
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8700599957355017e-05,
+      "loss": 1.2351,
+      "step": 3933
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8699831946281127e-05,
+      "loss": 1.1674,
+      "step": 3934
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8699063724087905e-05,
+      "loss": 1.0604,
+      "step": 3935
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869829529079399e-05,
+      "loss": 1.2634,
+      "step": 3936
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869752664641802e-05,
+      "loss": 1.2421,
+      "step": 3937
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8696757790978668e-05,
+      "loss": 1.3091,
+      "step": 3938
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8695988724494577e-05,
+      "loss": 1.3192,
+      "step": 3939
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869521944698441e-05,
+      "loss": 1.2548,
+      "step": 3940
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869444995846684e-05,
+      "loss": 1.2724,
+      "step": 3941
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8693680258960543e-05,
+      "loss": 1.2685,
+      "step": 3942
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869291034848419e-05,
+      "loss": 1.205,
+      "step": 3943
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8692140227056468e-05,
+      "loss": 1.1336,
+      "step": 3944
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8691369894696064e-05,
+      "loss": 1.1856,
+      "step": 3945
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8690599351421675e-05,
+      "loss": 1.1405,
+      "step": 3946
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8689828597252e-05,
+      "loss": 1.1387,
+      "step": 3947
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8689057632205737e-05,
+      "loss": 1.062,
+      "step": 3948
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8688286456301602e-05,
+      "loss": 0.9625,
+      "step": 3949
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8687515069558303e-05,
+      "loss": 1.0397,
+      "step": 3950
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8686743471994564e-05,
+      "loss": 1.2942,
+      "step": 3951
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868597166362911e-05,
+      "loss": 1.2072,
+      "step": 3952
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8685199644480663e-05,
+      "loss": 1.0781,
+      "step": 3953
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8684427414567964e-05,
+      "loss": 1.071,
+      "step": 3954
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8683654973909754e-05,
+      "loss": 1.153,
+      "step": 3955
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8682882322524777e-05,
+      "loss": 1.2543,
+      "step": 3956
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8682109460431775e-05,
+      "loss": 1.1794,
+      "step": 3957
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8681336387649516e-05,
+      "loss": 1.1308,
+      "step": 3958
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8680563104196753e-05,
+      "loss": 1.1374,
+      "step": 3959
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867978961009225e-05,
+      "loss": 1.1555,
+      "step": 3960
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8679015905354777e-05,
+      "loss": 1.2465,
+      "step": 3961
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8678241990003116e-05,
+      "loss": 1.1826,
+      "step": 3962
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8677467864056045e-05,
+      "loss": 1.1232,
+      "step": 3963
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8676693527532344e-05,
+      "loss": 1.1392,
+      "step": 3964
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8675918980450812e-05,
+      "loss": 1.3413,
+      "step": 3965
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8675144222830242e-05,
+      "loss": 1.2755,
+      "step": 3966
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867436925468943e-05,
+      "loss": 1.1886,
+      "step": 3967
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867359407604719e-05,
+      "loss": 1.0795,
+      "step": 3968
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867281868692233e-05,
+      "loss": 1.2728,
+      "step": 3969
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8672043087333662e-05,
+      "loss": 1.0528,
+      "step": 3970
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8671267277300015e-05,
+      "loss": 1.1735,
+      "step": 3971
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8670491256840212e-05,
+      "loss": 1.1709,
+      "step": 3972
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866971502597309e-05,
+      "loss": 1.0998,
+      "step": 3973
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8668938584717473e-05,
+      "loss": 1.1218,
+      "step": 3974
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8668161933092218e-05,
+      "loss": 1.1964,
+      "step": 3975
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8667385071116157e-05,
+      "loss": 1.1752,
+      "step": 3976
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8666607998808157e-05,
+      "loss": 1.3924,
+      "step": 3977
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8665830716187064e-05,
+      "loss": 1.3889,
+      "step": 3978
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866505322327175e-05,
+      "loss": 1.1757,
+      "step": 3979
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866427552008107e-05,
+      "loss": 1.1496,
+      "step": 3980
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866349760663391e-05,
+      "loss": 1.1544,
+      "step": 3981
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8662719482949142e-05,
+      "loss": 1.0303,
+      "step": 3982
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8661941149045646e-05,
+      "loss": 1.0501,
+      "step": 3983
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866116260494231e-05,
+      "loss": 1.2457,
+      "step": 3984
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8660383850658033e-05,
+      "loss": 1.2882,
+      "step": 3985
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8659604886211705e-05,
+      "loss": 1.1677,
+      "step": 3986
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865882571162224e-05,
+      "loss": 1.1317,
+      "step": 3987
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8658046326908533e-05,
+      "loss": 1.2686,
+      "step": 3988
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8657266732089508e-05,
+      "loss": 1.2132,
+      "step": 3989
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865648692718408e-05,
+      "loss": 1.2434,
+      "step": 3990
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865570691221117e-05,
+      "loss": 1.1364,
+      "step": 3991
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8654926687189706e-05,
+      "loss": 1.1938,
+      "step": 3992
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865414625213863e-05,
+      "loss": 1.0807,
+      "step": 3993
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8653365607076873e-05,
+      "loss": 1.1815,
+      "step": 3994
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865258475202338e-05,
+      "loss": 1.1207,
+      "step": 3995
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86518036869971e-05,
+      "loss": 1.1545,
+      "step": 3996
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8651022412016993e-05,
+      "loss": 1.314,
+      "step": 3997
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8650240927102012e-05,
+      "loss": 1.1864,
+      "step": 3998
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8649459232271124e-05,
+      "loss": 1.1911,
+      "step": 3999
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8648677327543297e-05,
+      "loss": 1.1277,
+      "step": 4000
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8647895212937504e-05,
+      "loss": 1.1076,
+      "step": 4001
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8647112888472732e-05,
+      "loss": 1.1102,
+      "step": 4002
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8646330354167955e-05,
+      "loss": 1.1714,
+      "step": 4003
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8645547610042173e-05,
+      "loss": 1.2317,
+      "step": 4004
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864476465611437e-05,
+      "loss": 1.2097,
+      "step": 4005
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8643981492403557e-05,
+      "loss": 1.1577,
+      "step": 4006
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864319811892873e-05,
+      "loss": 1.2455,
+      "step": 4007
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864241453570891e-05,
+      "loss": 1.235,
+      "step": 4008
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86416307427631e-05,
+      "loss": 1.1733,
+      "step": 4009
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8640846740110327e-05,
+      "loss": 1.3917,
+      "step": 4010
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8640062527769615e-05,
+      "loss": 1.3197,
+      "step": 4011
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8639278105759998e-05,
+      "loss": 1.1091,
+      "step": 4012
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8638493474100507e-05,
+      "loss": 1.1848,
+      "step": 4013
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8637708632810185e-05,
+      "loss": 1.1451,
+      "step": 4014
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8636923581908074e-05,
+      "loss": 1.2062,
+      "step": 4015
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863613832141323e-05,
+      "loss": 1.3385,
+      "step": 4016
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8635352851344707e-05,
+      "loss": 1.2264,
+      "step": 4017
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8634567171721567e-05,
+      "loss": 1.2717,
+      "step": 4018
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8633781282562875e-05,
+      "loss": 1.1632,
+      "step": 4019
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8632995183887697e-05,
+      "loss": 1.0786,
+      "step": 4020
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8632208875715122e-05,
+      "loss": 1.045,
+      "step": 4021
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8631422358064218e-05,
+      "loss": 1.3108,
+      "step": 4022
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8630635630954083e-05,
+      "loss": 1.1491,
+      "step": 4023
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86298486944038e-05,
+      "loss": 1.0324,
+      "step": 4024
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862906154843247e-05,
+      "loss": 1.1668,
+      "step": 4025
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8628274193059193e-05,
+      "loss": 1.2145,
+      "step": 4026
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8627486628303076e-05,
+      "loss": 1.1275,
+      "step": 4027
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862669885418323e-05,
+      "loss": 1.1902,
+      "step": 4028
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8625910870718775e-05,
+      "loss": 1.1638,
+      "step": 4029
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862512267792883e-05,
+      "loss": 1.0755,
+      "step": 4030
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8624334275832522e-05,
+      "loss": 1.2605,
+      "step": 4031
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8623545664448987e-05,
+      "loss": 1.2804,
+      "step": 4032
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8622756843797356e-05,
+      "loss": 1.2739,
+      "step": 4033
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8621967813896776e-05,
+      "loss": 1.1248,
+      "step": 4034
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8621178574766397e-05,
+      "loss": 1.165,
+      "step": 4035
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8620389126425365e-05,
+      "loss": 1.1468,
+      "step": 4036
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861959946889284e-05,
+      "loss": 1.2171,
+      "step": 4037
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8618809602187987e-05,
+      "loss": 1.1556,
+      "step": 4038
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861801952632997e-05,
+      "loss": 1.0033,
+      "step": 4039
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8617229241337967e-05,
+      "loss": 1.232,
+      "step": 4040
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861643874723115e-05,
+      "loss": 1.1836,
+      "step": 4041
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86156480440287e-05,
+      "loss": 1.1848,
+      "step": 4042
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8614857131749818e-05,
+      "loss": 1.1878,
+      "step": 4043
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8614066010413686e-05,
+      "loss": 1.1483,
+      "step": 4044
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8613274680039506e-05,
+      "loss": 1.1374,
+      "step": 4045
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861248314064648e-05,
+      "loss": 1.1966,
+      "step": 4046
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8611691392253814e-05,
+      "loss": 1.1508,
+      "step": 4047
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8610899434880724e-05,
+      "loss": 1.1656,
+      "step": 4048
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861010726854643e-05,
+      "loss": 1.2399,
+      "step": 4049
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8609314893270155e-05,
+      "loss": 1.2363,
+      "step": 4050
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8608522309071128e-05,
+      "loss": 1.1878,
+      "step": 4051
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.860772951596858e-05,
+      "loss": 1.1122,
+      "step": 4052
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8606936513981745e-05,
+      "loss": 1.1449,
+      "step": 4053
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.860614330312988e-05,
+      "loss": 1.217,
+      "step": 4054
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8605349883432223e-05,
+      "loss": 1.2056,
+      "step": 4055
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8604556254908034e-05,
+      "loss": 1.2839,
+      "step": 4056
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860376241757657e-05,
+      "loss": 1.0814,
+      "step": 4057
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8602968371457094e-05,
+      "loss": 1.3529,
+      "step": 4058
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8602174116568876e-05,
+      "loss": 1.3728,
+      "step": 4059
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860137965293119e-05,
+      "loss": 1.2741,
+      "step": 4060
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8600584980563316e-05,
+      "loss": 1.2815,
+      "step": 4061
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859979009948454e-05,
+      "loss": 1.0837,
+      "step": 4062
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8598995009714145e-05,
+      "loss": 1.1803,
+      "step": 4063
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8598199711271433e-05,
+      "loss": 1.2477,
+      "step": 4064
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85974042041757e-05,
+      "loss": 1.232,
+      "step": 4065
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859660848844625e-05,
+      "loss": 1.0743,
+      "step": 4066
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8595812564102393e-05,
+      "loss": 1.2404,
+      "step": 4067
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8595016431163448e-05,
+      "loss": 1.3358,
+      "step": 4068
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8594220089648727e-05,
+      "loss": 1.0191,
+      "step": 4069
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8593423539577565e-05,
+      "loss": 1.1987,
+      "step": 4070
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859262678096928e-05,
+      "loss": 1.2525,
+      "step": 4071
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8591829813843215e-05,
+      "loss": 1.2641,
+      "step": 4072
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859103263821871e-05,
+      "loss": 1.1456,
+      "step": 4073
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8590235254115105e-05,
+      "loss": 1.2698,
+      "step": 4074
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8589437661551756e-05,
+      "loss": 1.1406,
+      "step": 4075
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8588639860548012e-05,
+      "loss": 1.1481,
+      "step": 4076
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858784185112324e-05,
+      "loss": 1.1432,
+      "step": 4077
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85870436332968e-05,
+      "loss": 1.0934,
+      "step": 4078
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8586245207088068e-05,
+      "loss": 1.1427,
+      "step": 4079
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8585446572516416e-05,
+      "loss": 1.1319,
+      "step": 4080
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858464772960122e-05,
+      "loss": 1.1377,
+      "step": 4081
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858384867836187e-05,
+      "loss": 1.1864,
+      "step": 4082
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8583049418817764e-05,
+      "loss": 1.196,
+      "step": 4083
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8582249950988285e-05,
+      "loss": 1.0486,
+      "step": 4084
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8581450274892842e-05,
+      "loss": 1.201,
+      "step": 4085
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8580650390550835e-05,
+      "loss": 1.1131,
+      "step": 4086
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857985029798168e-05,
+      "loss": 1.0449,
+      "step": 4087
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857904999720479e-05,
+      "loss": 1.128,
+      "step": 4088
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8578249488239584e-05,
+      "loss": 1.244,
+      "step": 4089
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8577448771105494e-05,
+      "loss": 1.2562,
+      "step": 4090
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8576647845821947e-05,
+      "loss": 1.1726,
+      "step": 4091
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857584671240838e-05,
+      "loss": 1.2602,
+      "step": 4092
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8575045370884232e-05,
+      "loss": 1.277,
+      "step": 4093
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8574243821268953e-05,
+      "loss": 0.9587,
+      "step": 4094
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857344206358199e-05,
+      "loss": 1.0974,
+      "step": 4095
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8572640097842804e-05,
+      "loss": 1.3013,
+      "step": 4096
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8571837924070853e-05,
+      "loss": 1.1491,
+      "step": 4097
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857103554228561e-05,
+      "loss": 1.1379,
+      "step": 4098
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857023295250653e-05,
+      "loss": 1.2382,
+      "step": 4099
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856943015475311e-05,
+      "loss": 1.2555,
+      "step": 4100
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856862714904482e-05,
+      "loss": 1.1056,
+      "step": 4101
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8567823935401145e-05,
+      "loss": 1.4098,
+      "step": 4102
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8567020513841582e-05,
+      "loss": 1.2788,
+      "step": 4103
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8566216884385625e-05,
+      "loss": 1.2343,
+      "step": 4104
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8565413047052778e-05,
+      "loss": 1.2873,
+      "step": 4105
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8564609001862547e-05,
+      "loss": 0.9559,
+      "step": 4106
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856380474883444e-05,
+      "loss": 1.2961,
+      "step": 4107
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856300028798798e-05,
+      "loss": 1.3314,
+      "step": 4108
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8562195619342684e-05,
+      "loss": 1.1129,
+      "step": 4109
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856139074291808e-05,
+      "loss": 1.0752,
+      "step": 4110
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8560585658733707e-05,
+      "loss": 1.2855,
+      "step": 4111
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855978036680909e-05,
+      "loss": 1.0598,
+      "step": 4112
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8558974867163778e-05,
+      "loss": 1.2214,
+      "step": 4113
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8558169159817316e-05,
+      "loss": 1.2934,
+      "step": 4114
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855736324478926e-05,
+      "loss": 1.1384,
+      "step": 4115
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8556557122099163e-05,
+      "loss": 1.2819,
+      "step": 4116
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8555750791766588e-05,
+      "loss": 0.9317,
+      "step": 4117
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8554944253811103e-05,
+      "loss": 1.2028,
+      "step": 4118
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8554137508252278e-05,
+      "loss": 1.2437,
+      "step": 4119
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8553330555109696e-05,
+      "loss": 1.1858,
+      "step": 4120
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8552523394402933e-05,
+      "loss": 1.1227,
+      "step": 4121
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8551716026151584e-05,
+      "loss": 1.1967,
+      "step": 4122
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8550908450375232e-05,
+      "loss": 1.2717,
+      "step": 4123
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855010066709348e-05,
+      "loss": 1.1064,
+      "step": 4124
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8549292676325935e-05,
+      "loss": 1.1586,
+      "step": 4125
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8548484478092192e-05,
+      "loss": 1.2811,
+      "step": 4126
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8547676072411874e-05,
+      "loss": 1.1613,
+      "step": 4127
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8546867459304595e-05,
+      "loss": 1.1517,
+      "step": 4128
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8546058638789982e-05,
+      "loss": 1.1015,
+      "step": 4129
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8545249610887653e-05,
+      "loss": 1.0537,
+      "step": 4130
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854444037561725e-05,
+      "loss": 1.0646,
+      "step": 4131
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8543630932998408e-05,
+      "loss": 1.2486,
+      "step": 4132
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8542821283050766e-05,
+      "loss": 1.2742,
+      "step": 4133
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8542011425793976e-05,
+      "loss": 1.3002,
+      "step": 4134
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854120136124769e-05,
+      "loss": 1.211,
+      "step": 4135
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8540391089431566e-05,
+      "loss": 1.2917,
+      "step": 4136
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8539580610365267e-05,
+      "loss": 1.1095,
+      "step": 4137
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853876992406846e-05,
+      "loss": 1.129,
+      "step": 4138
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853795903056082e-05,
+      "loss": 1.1024,
+      "step": 4139
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8537147929862023e-05,
+      "loss": 1.198,
+      "step": 4140
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853633662199175e-05,
+      "loss": 1.1616,
+      "step": 4141
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8535525106969694e-05,
+      "loss": 1.1698,
+      "step": 4142
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8534713384815548e-05,
+      "loss": 1.0938,
+      "step": 4143
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8533901455549005e-05,
+      "loss": 1.1689,
+      "step": 4144
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853308931918977e-05,
+      "loss": 1.1455,
+      "step": 4145
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8532276975757554e-05,
+      "loss": 1.2889,
+      "step": 4146
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8531464425272067e-05,
+      "loss": 1.1357,
+      "step": 4147
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853065166775303e-05,
+      "loss": 1.0537,
+      "step": 4148
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8529838703220164e-05,
+      "loss": 1.2678,
+      "step": 4149
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85290255316932e-05,
+      "loss": 1.3548,
+      "step": 4150
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8528212153191868e-05,
+      "loss": 1.1439,
+      "step": 4151
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8527398567735904e-05,
+      "loss": 1.2791,
+      "step": 4152
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852658477534506e-05,
+      "loss": 1.25,
+      "step": 4153
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8525770776039077e-05,
+      "loss": 1.2099,
+      "step": 4154
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852495656983771e-05,
+      "loss": 1.1749,
+      "step": 4155
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852414215676072e-05,
+      "loss": 1.2745,
+      "step": 4156
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8523327536827866e-05,
+      "loss": 1.3084,
+      "step": 4157
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852251271005892e-05,
+      "loss": 0.9687,
+      "step": 4158
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852169767647366e-05,
+      "loss": 1.2298,
+      "step": 4159
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852088243609185e-05,
+      "loss": 1.1663,
+      "step": 4160
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8520066988933286e-05,
+      "loss": 1.2356,
+      "step": 4161
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8519251335017753e-05,
+      "loss": 1.0353,
+      "step": 4162
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8518435474365045e-05,
+      "loss": 1.1028,
+      "step": 4163
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8517619406994956e-05,
+      "loss": 1.2187,
+      "step": 4164
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8516803132927296e-05,
+      "loss": 1.2725,
+      "step": 4165
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8515986652181873e-05,
+      "loss": 1.2771,
+      "step": 4166
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8515169964778496e-05,
+      "loss": 1.1549,
+      "step": 4167
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8514353070736987e-05,
+      "loss": 1.22,
+      "step": 4168
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851353597007717e-05,
+      "loss": 1.2,
+      "step": 4169
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851271866281887e-05,
+      "loss": 1.2822,
+      "step": 4170
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8511901148981922e-05,
+      "loss": 1.1245,
+      "step": 4171
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8511083428586172e-05,
+      "loss": 1.0578,
+      "step": 4172
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8510265501651454e-05,
+      "loss": 1.0658,
+      "step": 4173
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8509447368197617e-05,
+      "loss": 1.1086,
+      "step": 4174
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850862902824452e-05,
+      "loss": 1.1775,
+      "step": 4175
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850781048181202e-05,
+      "loss": 1.2673,
+      "step": 4176
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850699172891998e-05,
+      "loss": 1.2141,
+      "step": 4177
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850617276958827e-05,
+      "loss": 1.0842,
+      "step": 4178
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8505353603836756e-05,
+      "loss": 1.2836,
+      "step": 4179
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8504534231685332e-05,
+      "loss": 1.212,
+      "step": 4180
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8503714653153867e-05,
+      "loss": 1.0569,
+      "step": 4181
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850289486826226e-05,
+      "loss": 1.1011,
+      "step": 4182
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85020748770304e-05,
+      "loss": 1.2402,
+      "step": 4183
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8501254679478188e-05,
+      "loss": 1.2916,
+      "step": 4184
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850043427562552e-05,
+      "loss": 1.1906,
+      "step": 4185
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8499613665492317e-05,
+      "loss": 1.2658,
+      "step": 4186
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8498792849098482e-05,
+      "loss": 1.1286,
+      "step": 4187
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8497971826463943e-05,
+      "loss": 1.1003,
+      "step": 4188
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849715059760862e-05,
+      "loss": 1.2359,
+      "step": 4189
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8496329162552437e-05,
+      "loss": 1.1252,
+      "step": 4190
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8495507521315333e-05,
+      "loss": 1.1903,
+      "step": 4191
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8494685673917248e-05,
+      "loss": 1.305,
+      "step": 4192
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8493863620378123e-05,
+      "loss": 1.3074,
+      "step": 4193
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8493041360717908e-05,
+      "loss": 1.2343,
+      "step": 4194
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8492218894956555e-05,
+      "loss": 1.2834,
+      "step": 4195
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8491396223114024e-05,
+      "loss": 1.2276,
+      "step": 4196
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849057334521028e-05,
+      "loss": 1.1491,
+      "step": 4197
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848975026126529e-05,
+      "loss": 1.2274,
+      "step": 4198
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848892697129903e-05,
+      "loss": 0.9693,
+      "step": 4199
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8488103475331476e-05,
+      "loss": 1.0735,
+      "step": 4200
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8487279773382613e-05,
+      "loss": 1.0055,
+      "step": 4201
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848645586547243e-05,
+      "loss": 1.2054,
+      "step": 4202
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848563175162092e-05,
+      "loss": 1.2225,
+      "step": 4203
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8484807431848085e-05,
+      "loss": 1.0529,
+      "step": 4204
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8483982906173928e-05,
+      "loss": 1.2927,
+      "step": 4205
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8483158174618456e-05,
+      "loss": 1.2266,
+      "step": 4206
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8482333237201678e-05,
+      "loss": 1.1628,
+      "step": 4207
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8481508093943622e-05,
+      "loss": 1.1603,
+      "step": 4208
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8480682744864306e-05,
+      "loss": 1.0908,
+      "step": 4209
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8479857189983762e-05,
+      "loss": 1.1654,
+      "step": 4210
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8479031429322022e-05,
+      "loss": 1.2113,
+      "step": 4211
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8478205462899127e-05,
+      "loss": 1.2409,
+      "step": 4212
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8477379290735117e-05,
+      "loss": 1.2195,
+      "step": 4213
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8476552912850044e-05,
+      "loss": 1.2123,
+      "step": 4214
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8475726329263958e-05,
+      "loss": 1.2804,
+      "step": 4215
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8474899539996923e-05,
+      "loss": 1.2606,
+      "step": 4216
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8474072545068998e-05,
+      "loss": 1.1622,
+      "step": 4217
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8473245344500254e-05,
+      "loss": 1.2761,
+      "step": 4218
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8472417938310765e-05,
+      "loss": 1.2494,
+      "step": 4219
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8471590326520607e-05,
+      "loss": 1.225,
+      "step": 4220
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8470762509149867e-05,
+      "loss": 1.2834,
+      "step": 4221
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8469934486218634e-05,
+      "loss": 1.2801,
+      "step": 4222
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8469106257747002e-05,
+      "loss": 1.1568,
+      "step": 4223
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846827782375506e-05,
+      "loss": 1.1725,
+      "step": 4224
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8467449184262927e-05,
+      "loss": 1.1457,
+      "step": 4225
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84666203392907e-05,
+      "loss": 1.2864,
+      "step": 4226
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84657912888585e-05,
+      "loss": 1.0396,
+      "step": 4227
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846496203298644e-05,
+      "loss": 1.2362,
+      "step": 4228
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8464132571694648e-05,
+      "loss": 1.1583,
+      "step": 4229
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8463302905003247e-05,
+      "loss": 1.1516,
+      "step": 4230
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846247303293238e-05,
+      "loss": 1.27,
+      "step": 4231
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846164295550218e-05,
+      "loss": 1.2786,
+      "step": 4232
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846081267273278e-05,
+      "loss": 1.2566,
+      "step": 4233
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845998218464435e-05,
+      "loss": 1.32,
+      "step": 4234
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845915149125703e-05,
+      "loss": 1.0358,
+      "step": 4235
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8458320592590976e-05,
+      "loss": 1.2206,
+      "step": 4236
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845748948866636e-05,
+      "loss": 1.1374,
+      "step": 4237
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8456658179503345e-05,
+      "loss": 1.1628,
+      "step": 4238
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8455826665122107e-05,
+      "loss": 1.2725,
+      "step": 4239
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845499494554282e-05,
+      "loss": 1.2011,
+      "step": 4240
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8454163020785676e-05,
+      "loss": 1.1834,
+      "step": 4241
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8453330890870857e-05,
+      "loss": 1.193,
+      "step": 4242
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8452498555818556e-05,
+      "loss": 1.2983,
+      "step": 4243
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8451666015648976e-05,
+      "loss": 1.1754,
+      "step": 4244
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8450833270382312e-05,
+      "loss": 1.1061,
+      "step": 4245
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845000032003878e-05,
+      "loss": 1.0987,
+      "step": 4246
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8449167164638596e-05,
+      "loss": 1.1257,
+      "step": 4247
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844833380420197e-05,
+      "loss": 1.2225,
+      "step": 4248
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8447500238749124e-05,
+      "loss": 1.2361,
+      "step": 4249
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8446666468300292e-05,
+      "loss": 1.0713,
+      "step": 4250
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844583249287571e-05,
+      "loss": 1.1097,
+      "step": 4251
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844499831249561e-05,
+      "loss": 1.1092,
+      "step": 4252
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8444163927180232e-05,
+      "loss": 1.155,
+      "step": 4253
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8443329336949835e-05,
+      "loss": 1.0784,
+      "step": 4254
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844249454182466e-05,
+      "loss": 1.2454,
+      "step": 4255
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8441659541824975e-05,
+      "loss": 1.201,
+      "step": 4256
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844082433697104e-05,
+      "loss": 1.2192,
+      "step": 4257
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.843998892728312e-05,
+      "loss": 1.2295,
+      "step": 4258
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8439153312781487e-05,
+      "loss": 1.1866,
+      "step": 4259
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8438317493486426e-05,
+      "loss": 1.3254,
+      "step": 4260
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.843748146941821e-05,
+      "loss": 1.2334,
+      "step": 4261
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.843664524059714e-05,
+      "loss": 1.3324,
+      "step": 4262
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8435808807043498e-05,
+      "loss": 1.2719,
+      "step": 4263
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8434972168777585e-05,
+      "loss": 1.2708,
+      "step": 4264
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8434135325819703e-05,
+      "loss": 1.2296,
+      "step": 4265
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8433298278190163e-05,
+      "loss": 1.0616,
+      "step": 4266
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843246102590927e-05,
+      "loss": 1.093,
+      "step": 4267
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843162356899735e-05,
+      "loss": 1.1036,
+      "step": 4268
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8430785907474727e-05,
+      "loss": 1.3304,
+      "step": 4269
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842994804136172e-05,
+      "loss": 1.22,
+      "step": 4270
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8429109970678664e-05,
+      "loss": 1.3103,
+      "step": 4271
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8428271695445903e-05,
+      "loss": 1.1148,
+      "step": 4272
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842743321568377e-05,
+      "loss": 1.2356,
+      "step": 4273
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842659453141262e-05,
+      "loss": 1.0206,
+      "step": 4274
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8425755642652797e-05,
+      "loss": 1.2125,
+      "step": 4275
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842491654942467e-05,
+      "loss": 1.1491,
+      "step": 4276
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8424077251748593e-05,
+      "loss": 1.3364,
+      "step": 4277
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8423237749644936e-05,
+      "loss": 1.3159,
+      "step": 4278
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8422398043134068e-05,
+      "loss": 1.2498,
+      "step": 4279
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842155813223637e-05,
+      "loss": 1.159,
+      "step": 4280
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8420718016972227e-05,
+      "loss": 1.1803,
+      "step": 4281
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8419877697362015e-05,
+      "loss": 1.1575,
+      "step": 4282
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841903717342614e-05,
+      "loss": 1.0946,
+      "step": 4283
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8418196445184988e-05,
+      "loss": 1.2822,
+      "step": 4284
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841735551265897e-05,
+      "loss": 1.2353,
+      "step": 4285
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8416514375868482e-05,
+      "loss": 1.2224,
+      "step": 4286
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8415673034833945e-05,
+      "loss": 1.2259,
+      "step": 4287
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8414831489575774e-05,
+      "loss": 1.1738,
+      "step": 4288
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841398974011439e-05,
+      "loss": 1.192,
+      "step": 4289
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8413147786470217e-05,
+      "loss": 1.0537,
+      "step": 4290
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8412305628663693e-05,
+      "loss": 1.2582,
+      "step": 4291
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841146326671525e-05,
+      "loss": 1.1556,
+      "step": 4292
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841062070064533e-05,
+      "loss": 1.0428,
+      "step": 4293
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840977793047438e-05,
+      "loss": 1.239,
+      "step": 4294
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8408934956222855e-05,
+      "loss": 1.047,
+      "step": 4295
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8408091777911203e-05,
+      "loss": 1.2007,
+      "step": 4296
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8407248395559897e-05,
+      "loss": 1.08,
+      "step": 4297
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840640480918939e-05,
+      "loss": 1.1149,
+      "step": 4298
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8405561018820166e-05,
+      "loss": 1.3143,
+      "step": 4299
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8404717024472696e-05,
+      "loss": 1.3604,
+      "step": 4300
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8403872826167458e-05,
+      "loss": 1.2295,
+      "step": 4301
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8403028423924943e-05,
+      "loss": 1.2609,
+      "step": 4302
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8402183817765643e-05,
+      "loss": 1.1283,
+      "step": 4303
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840133900771005e-05,
+      "loss": 1.2128,
+      "step": 4304
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8400493993778666e-05,
+      "loss": 1.0836,
+      "step": 4305
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8399648775991996e-05,
+      "loss": 1.0917,
+      "step": 4306
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8398803354370554e-05,
+      "loss": 1.2232,
+      "step": 4307
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8397957728934855e-05,
+      "loss": 1.2713,
+      "step": 4308
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839711189970542e-05,
+      "loss": 1.17,
+      "step": 4309
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8396265866702773e-05,
+      "loss": 1.1799,
+      "step": 4310
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8395419629947448e-05,
+      "loss": 1.0181,
+      "step": 4311
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8394573189459977e-05,
+      "loss": 1.1969,
+      "step": 4312
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8393726545260903e-05,
+      "loss": 1.2275,
+      "step": 4313
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839287969737077e-05,
+      "loss": 1.1114,
+      "step": 4314
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839203264581013e-05,
+      "loss": 1.0662,
+      "step": 4315
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8391185390599537e-05,
+      "loss": 1.2486,
+      "step": 4316
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8390337931759553e-05,
+      "loss": 1.2567,
+      "step": 4317
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8389490269310744e-05,
+      "loss": 1.2671,
+      "step": 4318
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8388642403273674e-05,
+      "loss": 0.9897,
+      "step": 4319
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8387794333668928e-05,
+      "loss": 1.1882,
+      "step": 4320
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838694606051708e-05,
+      "loss": 1.1737,
+      "step": 4321
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8386097583838714e-05,
+      "loss": 1.0166,
+      "step": 4322
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8385248903654423e-05,
+      "loss": 1.0304,
+      "step": 4323
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83844000199848e-05,
+      "loss": 1.2096,
+      "step": 4324
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838355093285045e-05,
+      "loss": 1.2194,
+      "step": 4325
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838270164227197e-05,
+      "loss": 1.198,
+      "step": 4326
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8381852148269976e-05,
+      "loss": 1.1626,
+      "step": 4327
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8381002450865078e-05,
+      "loss": 0.9942,
+      "step": 4328
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8380152550077903e-05,
+      "loss": 1.2915,
+      "step": 4329
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8379302445929068e-05,
+      "loss": 1.1782,
+      "step": 4330
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8378452138439206e-05,
+      "loss": 0.8848,
+      "step": 4331
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8377601627628952e-05,
+      "loss": 1.2857,
+      "step": 4332
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837675091351894e-05,
+      "loss": 1.0552,
+      "step": 4333
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8375899996129823e-05,
+      "loss": 1.3264,
+      "step": 4334
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837504887548224e-05,
+      "loss": 1.2149,
+      "step": 4335
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8374197551596857e-05,
+      "loss": 1.2003,
+      "step": 4336
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8373346024494324e-05,
+      "loss": 1.215,
+      "step": 4337
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8372494294195306e-05,
+      "loss": 1.057,
+      "step": 4338
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837164236072048e-05,
+      "loss": 1.128,
+      "step": 4339
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8370790224090508e-05,
+      "loss": 1.2434,
+      "step": 4340
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836993788432608e-05,
+      "loss": 1.2078,
+      "step": 4341
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836908534144787e-05,
+      "loss": 1.2714,
+      "step": 4342
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8368232595476575e-05,
+      "loss": 1.1799,
+      "step": 4343
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8367379646432884e-05,
+      "loss": 1.0682,
+      "step": 4344
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8366526494337497e-05,
+      "loss": 1.1303,
+      "step": 4345
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8365673139211114e-05,
+      "loss": 1.081,
+      "step": 4346
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8364819581074447e-05,
+      "loss": 1.1395,
+      "step": 4347
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836396581994821e-05,
+      "loss": 1.1038,
+      "step": 4348
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8363111855853122e-05,
+      "loss": 1.2513,
+      "step": 4349
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8362257688809904e-05,
+      "loss": 1.1988,
+      "step": 4350
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8361403318839283e-05,
+      "loss": 1.1795,
+      "step": 4351
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8360548745961994e-05,
+      "loss": 1.0197,
+      "step": 4352
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8359693970198772e-05,
+      "loss": 1.1986,
+      "step": 4353
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8358838991570363e-05,
+      "loss": 1.0411,
+      "step": 4354
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835798381009752e-05,
+      "loss": 1.2491,
+      "step": 4355
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8357128425800983e-05,
+      "loss": 1.1441,
+      "step": 4356
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835627283870152e-05,
+      "loss": 1.0429,
+      "step": 4357
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835541704881989e-05,
+      "loss": 1.1835,
+      "step": 4358
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8354561056176857e-05,
+      "loss": 1.193,
+      "step": 4359
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8353704860793202e-05,
+      "loss": 1.3548,
+      "step": 4360
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8352848462689694e-05,
+      "loss": 1.1487,
+      "step": 4361
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835199186188712e-05,
+      "loss": 0.9803,
+      "step": 4362
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8351135058406266e-05,
+      "loss": 1.2651,
+      "step": 4363
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8350278052267922e-05,
+      "loss": 1.1225,
+      "step": 4364
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834942084349289e-05,
+      "loss": 1.1567,
+      "step": 4365
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8348563432101967e-05,
+      "loss": 1.1748,
+      "step": 4366
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834770581811596e-05,
+      "loss": 1.186,
+      "step": 4367
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8346848001555688e-05,
+      "loss": 1.161,
+      "step": 4368
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8345989982441955e-05,
+      "loss": 1.1132,
+      "step": 4369
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8345131760795598e-05,
+      "loss": 1.3296,
+      "step": 4370
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834427333663743e-05,
+      "loss": 1.0326,
+      "step": 4371
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8343414709988288e-05,
+      "loss": 1.0053,
+      "step": 4372
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834255588086901e-05,
+      "loss": 1.1744,
+      "step": 4373
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834169684930043e-05,
+      "loss": 1.0377,
+      "step": 4374
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8340837615303405e-05,
+      "loss": 1.1398,
+      "step": 4375
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833997817889878e-05,
+      "loss": 1.0255,
+      "step": 4376
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833911854010741e-05,
+      "loss": 1.2353,
+      "step": 4377
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833825869895016e-05,
+      "loss": 1.1902,
+      "step": 4378
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8337398655447894e-05,
+      "loss": 1.1356,
+      "step": 4379
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8336538409621474e-05,
+      "loss": 1.0688,
+      "step": 4380
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8335677961491793e-05,
+      "loss": 1.0658,
+      "step": 4381
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833481731107972e-05,
+      "loss": 1.2167,
+      "step": 4382
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833395645840614e-05,
+      "loss": 1.149,
+      "step": 4383
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8333095403491946e-05,
+      "loss": 1.134,
+      "step": 4384
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8332234146358034e-05,
+      "loss": 1.2426,
+      "step": 4385
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8331372687025305e-05,
+      "loss": 1.3259,
+      "step": 4386
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8330511025514662e-05,
+      "loss": 1.077,
+      "step": 4387
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8329649161847016e-05,
+      "loss": 1.186,
+      "step": 4388
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8328787096043278e-05,
+      "loss": 1.1977,
+      "step": 4389
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8327924828124377e-05,
+      "loss": 1.2017,
+      "step": 4390
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8327062358111228e-05,
+      "loss": 1.021,
+      "step": 4391
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8326199686024765e-05,
+      "loss": 1.0894,
+      "step": 4392
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8325336811885926e-05,
+      "loss": 1.1075,
+      "step": 4393
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8324473735715643e-05,
+      "loss": 1.1467,
+      "step": 4394
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832361045753486e-05,
+      "loss": 1.2227,
+      "step": 4395
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8322746977364537e-05,
+      "loss": 1.1578,
+      "step": 4396
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8321883295225617e-05,
+      "loss": 1.1745,
+      "step": 4397
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8321019411139064e-05,
+      "loss": 1.1519,
+      "step": 4398
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8320155325125843e-05,
+      "loss": 1.2049,
+      "step": 4399
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831929103720692e-05,
+      "loss": 1.1157,
+      "step": 4400
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831842654740327e-05,
+      "loss": 1.1667,
+      "step": 4401
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8317561855735867e-05,
+      "loss": 1.1775,
+      "step": 4402
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8316696962225704e-05,
+      "loss": 1.2596,
+      "step": 4403
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8315831866893762e-05,
+      "loss": 1.1449,
+      "step": 4404
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831496656976104e-05,
+      "loss": 1.295,
+      "step": 4405
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8314101070848527e-05,
+      "loss": 1.2588,
+      "step": 4406
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8313235370177235e-05,
+      "loss": 1.2405,
+      "step": 4407
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8312369467768168e-05,
+      "loss": 1.1344,
+      "step": 4408
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831150336364234e-05,
+      "loss": 1.1663,
+      "step": 4409
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831063705782077e-05,
+      "loss": 1.2216,
+      "step": 4410
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830977055032448e-05,
+      "loss": 1.1386,
+      "step": 4411
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8308903841174493e-05,
+      "loss": 1.2176,
+      "step": 4412
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8308036930391848e-05,
+      "loss": 1.1809,
+      "step": 4413
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830716981799758e-05,
+      "loss": 1.1828,
+      "step": 4414
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8306302504012732e-05,
+      "loss": 1.0509,
+      "step": 4415
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8305434988458348e-05,
+      "loss": 1.1473,
+      "step": 4416
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8304567271355482e-05,
+      "loss": 1.174,
+      "step": 4417
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8303699352725193e-05,
+      "loss": 1.3085,
+      "step": 4418
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830283123258854e-05,
+      "loss": 1.1464,
+      "step": 4419
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8301962910966592e-05,
+      "loss": 1.1846,
+      "step": 4420
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830109438788042e-05,
+      "loss": 1.2235,
+      "step": 4421
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8300225663351098e-05,
+      "loss": 1.0949,
+      "step": 4422
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8299356737399707e-05,
+      "loss": 1.1219,
+      "step": 4423
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8298487610047337e-05,
+      "loss": 1.2271,
+      "step": 4424
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829761828131508e-05,
+      "loss": 1.2478,
+      "step": 4425
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829674875122403e-05,
+      "loss": 1.2558,
+      "step": 4426
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8295879019795283e-05,
+      "loss": 1.1483,
+      "step": 4427
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8295009087049954e-05,
+      "loss": 1.1707,
+      "step": 4428
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8294138953009145e-05,
+      "loss": 1.2477,
+      "step": 4429
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829326861769398e-05,
+      "loss": 1.1508,
+      "step": 4430
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8292398081125572e-05,
+      "loss": 0.9671,
+      "step": 4431
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8291527343325052e-05,
+      "loss": 1.2275,
+      "step": 4432
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8290656404313546e-05,
+      "loss": 1.1603,
+      "step": 4433
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8289785264112193e-05,
+      "loss": 1.1362,
+      "step": 4434
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8288913922742134e-05,
+      "loss": 1.0752,
+      "step": 4435
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8288042380224508e-05,
+      "loss": 1.124,
+      "step": 4436
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8287170636580464e-05,
+      "loss": 1.1313,
+      "step": 4437
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8286298691831164e-05,
+      "loss": 1.1026,
+      "step": 4438
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8285426545997764e-05,
+      "loss": 1.0769,
+      "step": 4439
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828455419910143e-05,
+      "loss": 1.3299,
+      "step": 4440
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8283681651163324e-05,
+      "loss": 1.1644,
+      "step": 4441
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8282808902204627e-05,
+      "loss": 1.3402,
+      "step": 4442
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828193595224652e-05,
+      "loss": 1.044,
+      "step": 4443
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828106280131018e-05,
+      "loss": 1.2918,
+      "step": 4444
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8280189449416805e-05,
+      "loss": 1.1479,
+      "step": 4445
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827931589658758e-05,
+      "loss": 1.2512,
+      "step": 4446
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8278442142843703e-05,
+      "loss": 1.314,
+      "step": 4447
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8277568188206386e-05,
+      "loss": 1.0258,
+      "step": 4448
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8276694032696835e-05,
+      "loss": 1.1418,
+      "step": 4449
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8275819676336256e-05,
+      "loss": 1.0235,
+      "step": 4450
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827494511914587e-05,
+      "loss": 0.9715,
+      "step": 4451
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8274070361146906e-05,
+      "loss": 1.3143,
+      "step": 4452
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8273195402360585e-05,
+      "loss": 1.1177,
+      "step": 4453
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8272320242808143e-05,
+      "loss": 1.278,
+      "step": 4454
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827144488251082e-05,
+      "loss": 1.0618,
+      "step": 4455
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827056932148985e-05,
+      "loss": 1.1831,
+      "step": 4456
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8269693559766487e-05,
+      "loss": 1.2691,
+      "step": 4457
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8268817597361983e-05,
+      "loss": 1.3019,
+      "step": 4458
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8267941434297594e-05,
+      "loss": 1.2241,
+      "step": 4459
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826706507059458e-05,
+      "loss": 1.2832,
+      "step": 4460
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826618850627421e-05,
+      "loss": 1.1095,
+      "step": 4461
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8265311741357753e-05,
+      "loss": 1.2628,
+      "step": 4462
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826443477586649e-05,
+      "loss": 1.2361,
+      "step": 4463
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.82635576098217e-05,
+      "loss": 1.2445,
+      "step": 4464
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826268024324467e-05,
+      "loss": 1.2001,
+      "step": 4465
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8261802676156685e-05,
+      "loss": 1.1786,
+      "step": 4466
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826092490857905e-05,
+      "loss": 1.1136,
+      "step": 4467
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8260046940533063e-05,
+      "loss": 1.0664,
+      "step": 4468
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8259168772040027e-05,
+      "loss": 1.122,
+      "step": 4469
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8258290403121252e-05,
+      "loss": 1.1668,
+      "step": 4470
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825741183379806e-05,
+      "loss": 1.1,
+      "step": 4471
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8256533064091765e-05,
+      "loss": 1.2046,
+      "step": 4472
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8255654094023692e-05,
+      "loss": 1.1907,
+      "step": 4473
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8254774923615177e-05,
+      "loss": 1.3413,
+      "step": 4474
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8253895552887547e-05,
+      "loss": 1.2207,
+      "step": 4475
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825301598186215e-05,
+      "loss": 1.1523,
+      "step": 4476
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825213621056032e-05,
+      "loss": 1.0596,
+      "step": 4477
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825125623900342e-05,
+      "loss": 1.0629,
+      "step": 4478
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8250376067212794e-05,
+      "loss": 1.2471,
+      "step": 4479
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8249495695209805e-05,
+      "loss": 1.0297,
+      "step": 4480
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8248615123015816e-05,
+      "loss": 1.0418,
+      "step": 4481
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8247734350652197e-05,
+      "loss": 1.3091,
+      "step": 4482
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824685337814032e-05,
+      "loss": 1.2293,
+      "step": 4483
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8245972205501565e-05,
+      "loss": 1.2386,
+      "step": 4484
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8245090832757317e-05,
+      "loss": 1.2714,
+      "step": 4485
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824420925992896e-05,
+      "loss": 1.255,
+      "step": 4486
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824332748703789e-05,
+      "loss": 1.2319,
+      "step": 4487
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8242445514105505e-05,
+      "loss": 1.1764,
+      "step": 4488
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8241563341153203e-05,
+      "loss": 1.251,
+      "step": 4489
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82406809682024e-05,
+      "loss": 1.1203,
+      "step": 4490
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8239798395274507e-05,
+      "loss": 1.3492,
+      "step": 4491
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823891562239094e-05,
+      "loss": 1.1394,
+      "step": 4492
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8238032649573116e-05,
+      "loss": 1.1537,
+      "step": 4493
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8237149476842472e-05,
+      "loss": 1.2492,
+      "step": 4494
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8236266104220432e-05,
+      "loss": 1.2114,
+      "step": 4495
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8235382531728435e-05,
+      "loss": 1.0104,
+      "step": 4496
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8234498759387925e-05,
+      "loss": 1.1381,
+      "step": 4497
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8233614787220345e-05,
+      "loss": 1.2934,
+      "step": 4498
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8232730615247146e-05,
+      "loss": 1.2031,
+      "step": 4499
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823184624348979e-05,
+      "loss": 1.0164,
+      "step": 4500
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8230961671969735e-05,
+      "loss": 1.0891,
+      "step": 4501
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8230076900708447e-05,
+      "loss": 1.2463,
+      "step": 4502
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8229191929727395e-05,
+      "loss": 1.13,
+      "step": 4503
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8228306759048057e-05,
+      "loss": 1.2627,
+      "step": 4504
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8227421388691912e-05,
+      "loss": 1.158,
+      "step": 4505
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8226535818680445e-05,
+      "loss": 1.1255,
+      "step": 4506
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8225650049035148e-05,
+      "loss": 1.0206,
+      "step": 4507
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8224764079777514e-05,
+      "loss": 1.2856,
+      "step": 4508
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8223877910929048e-05,
+      "loss": 1.0854,
+      "step": 4509
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8222991542511247e-05,
+      "loss": 1.1432,
+      "step": 4510
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8222104974545623e-05,
+      "loss": 1.0976,
+      "step": 4511
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8221218207053694e-05,
+      "loss": 1.1146,
+      "step": 4512
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8220331240056974e-05,
+      "loss": 1.1341,
+      "step": 4513
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8219444073576993e-05,
+      "loss": 1.2141,
+      "step": 4514
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8218556707635277e-05,
+      "loss": 1.1706,
+      "step": 4515
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8217669142253358e-05,
+      "loss": 1.1761,
+      "step": 4516
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8216781377452775e-05,
+      "loss": 1.0962,
+      "step": 4517
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8215893413255073e-05,
+      "loss": 1.2206,
+      "step": 4518
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8215005249681804e-05,
+      "loss": 1.2567,
+      "step": 4519
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8214116886754513e-05,
+      "loss": 1.1661,
+      "step": 4520
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8213228324494765e-05,
+      "loss": 1.2349,
+      "step": 4521
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821233956292412e-05,
+      "loss": 1.289,
+      "step": 4522
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821145060206414e-05,
+      "loss": 1.0564,
+      "step": 4523
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8210561441936406e-05,
+      "loss": 1.2374,
+      "step": 4524
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8209672082562496e-05,
+      "loss": 1.1977,
+      "step": 4525
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8208782523963985e-05,
+      "loss": 1.1721,
+      "step": 4526
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8207892766162463e-05,
+      "loss": 1.1195,
+      "step": 4527
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820700280917952e-05,
+      "loss": 1.1939,
+      "step": 4528
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820611265303676e-05,
+      "loss": 1.3404,
+      "step": 4529
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8205222297755774e-05,
+      "loss": 1.219,
+      "step": 4530
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820433174335818e-05,
+      "loss": 1.1922,
+      "step": 4531
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8203440989865577e-05,
+      "loss": 1.1664,
+      "step": 4532
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8202550037299588e-05,
+      "loss": 1.2489,
+      "step": 4533
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820165888568183e-05,
+      "loss": 1.1997,
+      "step": 4534
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8200767535033938e-05,
+      "loss": 1.2652,
+      "step": 4535
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819987598537753e-05,
+      "loss": 1.2745,
+      "step": 4536
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8198984236734246e-05,
+      "loss": 1.0903,
+      "step": 4537
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819809228912573e-05,
+      "loss": 1.2685,
+      "step": 4538
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8197200142573625e-05,
+      "loss": 1.1635,
+      "step": 4539
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819630779709958e-05,
+      "loss": 1.2873,
+      "step": 4540
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8195415252725242e-05,
+      "loss": 1.2096,
+      "step": 4541
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8194522509472283e-05,
+      "loss": 1.2515,
+      "step": 4542
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819362956736236e-05,
+      "loss": 1.1013,
+      "step": 4543
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8192736426417146e-05,
+      "loss": 1.1723,
+      "step": 4544
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8191843086658313e-05,
+      "loss": 1.1059,
+      "step": 4545
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819094954810754e-05,
+      "loss": 1.0812,
+      "step": 4546
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8190055810786507e-05,
+      "loss": 1.2038,
+      "step": 4547
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818916187471691e-05,
+      "loss": 1.1872,
+      "step": 4548
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8188267739920433e-05,
+      "loss": 1.2552,
+      "step": 4549
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8187373406418782e-05,
+      "loss": 1.2622,
+      "step": 4550
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8186478874233655e-05,
+      "loss": 1.1925,
+      "step": 4551
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8185584143386764e-05,
+      "loss": 1.2154,
+      "step": 4552
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8184689213899816e-05,
+      "loss": 1.2077,
+      "step": 4553
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818379408579453e-05,
+      "loss": 1.1985,
+      "step": 4554
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818289875909263e-05,
+      "loss": 1.1607,
+      "step": 4555
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8182003233815847e-05,
+      "loss": 1.2749,
+      "step": 4556
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8181107509985903e-05,
+      "loss": 1.2994,
+      "step": 4557
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8180211587624543e-05,
+      "loss": 1.2303,
+      "step": 4558
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81793154667535e-05,
+      "loss": 1.1305,
+      "step": 4559
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8178419147394528e-05,
+      "loss": 1.2847,
+      "step": 4560
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8177522629569375e-05,
+      "loss": 1.0039,
+      "step": 4561
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8176625913299797e-05,
+      "loss": 1.2491,
+      "step": 4562
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8175728998607555e-05,
+      "loss": 1.0955,
+      "step": 4563
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8174831885514415e-05,
+      "loss": 1.2405,
+      "step": 4564
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817393457404214e-05,
+      "loss": 1.0679,
+      "step": 4565
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8173037064212522e-05,
+      "loss": 1.277,
+      "step": 4566
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8172139356047323e-05,
+      "loss": 1.3597,
+      "step": 4567
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817124144956834e-05,
+      "loss": 0.9393,
+      "step": 4568
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8170343344797354e-05,
+      "loss": 1.1165,
+      "step": 4569
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8169445041756165e-05,
+      "loss": 1.203,
+      "step": 4570
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8168546540466567e-05,
+      "loss": 1.1391,
+      "step": 4571
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816764784095037e-05,
+      "loss": 1.2628,
+      "step": 4572
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816674894322938e-05,
+      "loss": 1.1717,
+      "step": 4573
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8165849847325413e-05,
+      "loss": 1.166,
+      "step": 4574
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816495055326028e-05,
+      "loss": 1.2024,
+      "step": 4575
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8164051061055812e-05,
+      "loss": 1.1456,
+      "step": 4576
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8163151370733838e-05,
+      "loss": 1.198,
+      "step": 4577
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8162251482316186e-05,
+      "loss": 1.2407,
+      "step": 4578
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8161351395824688e-05,
+      "loss": 1.2508,
+      "step": 4579
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8160451111281202e-05,
+      "loss": 1.2364,
+      "step": 4580
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815955062870756e-05,
+      "loss": 1.2346,
+      "step": 4581
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815864994812562e-05,
+      "loss": 1.0636,
+      "step": 4582
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8157749069557246e-05,
+      "loss": 1.2294,
+      "step": 4583
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815684799302429e-05,
+      "loss": 1.1564,
+      "step": 4584
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815594671854862e-05,
+      "loss": 1.1371,
+      "step": 4585
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8155045246152113e-05,
+      "loss": 1.2402,
+      "step": 4586
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8154143575856634e-05,
+      "loss": 1.2201,
+      "step": 4587
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8153241707684077e-05,
+      "loss": 1.1917,
+      "step": 4588
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815233964165632e-05,
+      "loss": 1.2045,
+      "step": 4589
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8151437377795256e-05,
+      "loss": 1.2068,
+      "step": 4590
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8150534916122777e-05,
+      "loss": 1.1637,
+      "step": 4591
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814963225666079e-05,
+      "loss": 1.0853,
+      "step": 4592
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8148729399431187e-05,
+      "loss": 1.3087,
+      "step": 4593
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8147826344455893e-05,
+      "loss": 1.2097,
+      "step": 4594
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8146923091756813e-05,
+      "loss": 1.1808,
+      "step": 4595
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814601964135587e-05,
+      "loss": 1.1949,
+      "step": 4596
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8145115993274986e-05,
+      "loss": 1.3552,
+      "step": 4597
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814421214753609e-05,
+      "loss": 1.1971,
+      "step": 4598
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8143308104161117e-05,
+      "loss": 1.175,
+      "step": 4599
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8142403863172007e-05,
+      "loss": 1.0764,
+      "step": 4600
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81414994245907e-05,
+      "loss": 1.2324,
+      "step": 4601
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8140594788439142e-05,
+      "loss": 1.2922,
+      "step": 4602
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8139689954739294e-05,
+      "loss": 1.2666,
+      "step": 4603
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8138784923513107e-05,
+      "loss": 0.9513,
+      "step": 4604
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8137879694782543e-05,
+      "loss": 1.1904,
+      "step": 4605
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8136974268569575e-05,
+      "loss": 1.1555,
+      "step": 4606
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813606864489617e-05,
+      "loss": 1.1057,
+      "step": 4607
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8135162823784303e-05,
+      "loss": 1.2315,
+      "step": 4608
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8134256805255964e-05,
+      "loss": 1.3159,
+      "step": 4609
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8133350589333136e-05,
+      "loss": 0.9617,
+      "step": 4610
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8132444176037803e-05,
+      "loss": 1.2398,
+      "step": 4611
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8131537565391967e-05,
+      "loss": 1.2199,
+      "step": 4612
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8130630757417628e-05,
+      "loss": 1.2578,
+      "step": 4613
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8129723752136797e-05,
+      "loss": 1.059,
+      "step": 4614
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8128816549571472e-05,
+      "loss": 1.2046,
+      "step": 4615
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812790914974368e-05,
+      "loss": 1.2076,
+      "step": 4616
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8127001552675436e-05,
+      "loss": 1.062,
+      "step": 4617
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8126093758388764e-05,
+      "loss": 1.3487,
+      "step": 4618
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8125185766905697e-05,
+      "loss": 1.0651,
+      "step": 4619
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812427757824826e-05,
+      "loss": 1.1266,
+      "step": 4620
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8123369192438508e-05,
+      "loss": 1.1958,
+      "step": 4621
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812246060949847e-05,
+      "loss": 1.126,
+      "step": 4622
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81215518294502e-05,
+      "loss": 1.1236,
+      "step": 4623
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812064285231575e-05,
+      "loss": 1.1847,
+      "step": 4624
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8119733678117185e-05,
+      "loss": 1.1655,
+      "step": 4625
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811882430687656e-05,
+      "loss": 1.134,
+      "step": 4626
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811791473861595e-05,
+      "loss": 1.2423,
+      "step": 4627
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811700497335742e-05,
+      "loss": 1.1546,
+      "step": 4628
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811609501112305e-05,
+      "loss": 1.1949,
+      "step": 4629
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8115184851934922e-05,
+      "loss": 1.2418,
+      "step": 4630
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8114274495815123e-05,
+      "loss": 1.1602,
+      "step": 4631
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8113363942785747e-05,
+      "loss": 1.0526,
+      "step": 4632
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811245319286889e-05,
+      "loss": 1.321,
+      "step": 4633
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811154224608665e-05,
+      "loss": 1.1772,
+      "step": 4634
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8110631102461134e-05,
+      "loss": 1.2712,
+      "step": 4635
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8109719762014454e-05,
+      "loss": 1.2279,
+      "step": 4636
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8108808224768724e-05,
+      "loss": 1.265,
+      "step": 4637
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8107896490746067e-05,
+      "loss": 1.232,
+      "step": 4638
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810698455996861e-05,
+      "loss": 1.2036,
+      "step": 4639
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8106072432458478e-05,
+      "loss": 1.0404,
+      "step": 4640
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8105160108237805e-05,
+      "loss": 1.2395,
+      "step": 4641
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8104247587328733e-05,
+      "loss": 1.26,
+      "step": 4642
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8103334869753406e-05,
+      "loss": 1.1614,
+      "step": 4643
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8102421955533974e-05,
+      "loss": 1.2711,
+      "step": 4644
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8101508844692586e-05,
+      "loss": 1.2176,
+      "step": 4645
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810059553725141e-05,
+      "loss": 1.3198,
+      "step": 4646
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.80996820332326e-05,
+      "loss": 1.1539,
+      "step": 4647
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8098768332658325e-05,
+      "loss": 1.0634,
+      "step": 4648
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809785443555076e-05,
+      "loss": 1.279,
+      "step": 4649
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809694034193209e-05,
+      "loss": 1.1628,
+      "step": 4650
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8096026051824483e-05,
+      "loss": 1.1934,
+      "step": 4651
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8095111565250137e-05,
+      "loss": 1.081,
+      "step": 4652
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8094196882231235e-05,
+      "loss": 1.2239,
+      "step": 4653
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8093282002789983e-05,
+      "loss": 1.2405,
+      "step": 4654
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8092366926948578e-05,
+      "loss": 1.0506,
+      "step": 4655
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8091451654729225e-05,
+      "loss": 1.1222,
+      "step": 4656
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8090536186154143e-05,
+      "loss": 1.2025,
+      "step": 4657
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8089620521245534e-05,
+      "loss": 1.1538,
+      "step": 4658
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8088704660025626e-05,
+      "loss": 1.1564,
+      "step": 4659
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8087788602516643e-05,
+      "loss": 1.1818,
+      "step": 4660
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808687234874082e-05,
+      "loss": 1.2654,
+      "step": 4661
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8085955898720388e-05,
+      "loss": 1.249,
+      "step": 4662
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8085039252477584e-05,
+      "loss": 1.2241,
+      "step": 4663
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8084122410034655e-05,
+      "loss": 1.2614,
+      "step": 4664
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808320537141385e-05,
+      "loss": 1.1401,
+      "step": 4665
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808228813663742e-05,
+      "loss": 1.2833,
+      "step": 4666
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8081370705727632e-05,
+      "loss": 1.2176,
+      "step": 4667
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8080453078706737e-05,
+      "loss": 1.1565,
+      "step": 4668
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8079535255597014e-05,
+      "loss": 1.2111,
+      "step": 4669
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807861723642073e-05,
+      "loss": 1.2508,
+      "step": 4670
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8077699021200163e-05,
+      "loss": 1.1494,
+      "step": 4671
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.80767806099576e-05,
+      "loss": 1.1135,
+      "step": 4672
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807586200271532e-05,
+      "loss": 1.2777,
+      "step": 4673
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8074943199495622e-05,
+      "loss": 1.1485,
+      "step": 4674
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8074024200320797e-05,
+      "loss": 1.0904,
+      "step": 4675
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8073105005213154e-05,
+      "loss": 1.0921,
+      "step": 4676
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807218561419499e-05,
+      "loss": 1.1502,
+      "step": 4677
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8071266027288625e-05,
+      "loss": 1.277,
+      "step": 4678
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8070346244516367e-05,
+      "loss": 1.0879,
+      "step": 4679
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806942626590054e-05,
+      "loss": 1.2055,
+      "step": 4680
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8068506091463473e-05,
+      "loss": 1.1719,
+      "step": 4681
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806758572122749e-05,
+      "loss": 1.2966,
+      "step": 4682
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806666515521492e-05,
+      "loss": 1.245,
+      "step": 4683
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8065744393448118e-05,
+      "loss": 1.1987,
+      "step": 4684
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806482343594942e-05,
+      "loss": 1.1289,
+      "step": 4685
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806390228274117e-05,
+      "loss": 1.3132,
+      "step": 4686
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8062980933845732e-05,
+      "loss": 1.1043,
+      "step": 4687
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8062059389285455e-05,
+      "loss": 1.1801,
+      "step": 4688
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806113764908271e-05,
+      "loss": 1.1489,
+      "step": 4689
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8060215713259856e-05,
+      "loss": 0.9744,
+      "step": 4690
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8059293581839277e-05,
+      "loss": 1.3052,
+      "step": 4691
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805837125484334e-05,
+      "loss": 1.2225,
+      "step": 4692
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8057448732294432e-05,
+      "loss": 1.2228,
+      "step": 4693
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8056526014214944e-05,
+      "loss": 1.1189,
+      "step": 4694
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805560310062726e-05,
+      "loss": 1.2233,
+      "step": 4695
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8054679991553777e-05,
+      "loss": 1.163,
+      "step": 4696
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80537566870169e-05,
+      "loss": 1.0946,
+      "step": 4697
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805283318703903e-05,
+      "loss": 1.1175,
+      "step": 4698
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805190949164259e-05,
+      "loss": 1.2542,
+      "step": 4699
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805098560084998e-05,
+      "loss": 1.0474,
+      "step": 4700
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8050061514683624e-05,
+      "loss": 1.1512,
+      "step": 4701
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8049137233165955e-05,
+      "loss": 1.2151,
+      "step": 4702
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8048212756319395e-05,
+      "loss": 1.244,
+      "step": 4703
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804728808416638e-05,
+      "loss": 1.1852,
+      "step": 4704
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8046363216729354e-05,
+      "loss": 1.0354,
+      "step": 4705
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8045438154030752e-05,
+      "loss": 1.1242,
+      "step": 4706
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8044512896093027e-05,
+      "loss": 1.2526,
+      "step": 4707
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8043587442938633e-05,
+      "loss": 1.3312,
+      "step": 4708
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8042661794590023e-05,
+      "loss": 1.2709,
+      "step": 4709
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804173595106967e-05,
+      "loss": 1.0886,
+      "step": 4710
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804080991240003e-05,
+      "loss": 1.2058,
+      "step": 4711
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8039883678603583e-05,
+      "loss": 1.1671,
+      "step": 4712
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8038957249702806e-05,
+      "loss": 1.1873,
+      "step": 4713
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8038030625720173e-05,
+      "loss": 1.1748,
+      "step": 4714
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803710380667818e-05,
+      "loss": 1.0655,
+      "step": 4715
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8036176792599313e-05,
+      "loss": 1.1666,
+      "step": 4716
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803524958350607e-05,
+      "loss": 1.2175,
+      "step": 4717
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8034322179420946e-05,
+      "loss": 1.278,
+      "step": 4718
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8033394580366453e-05,
+      "loss": 1.0673,
+      "step": 4719
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8032466786365098e-05,
+      "loss": 1.3781,
+      "step": 4720
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80315387974394e-05,
+      "loss": 1.1079,
+      "step": 4721
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8030610613611874e-05,
+      "loss": 1.0343,
+      "step": 4722
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8029682234905044e-05,
+      "loss": 0.9051,
+      "step": 4723
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8028753661341442e-05,
+      "loss": 1.2656,
+      "step": 4724
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80278248929436e-05,
+      "loss": 1.1725,
+      "step": 4725
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8026895929734057e-05,
+      "loss": 1.1477,
+      "step": 4726
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8025966771735354e-05,
+      "loss": 1.1838,
+      "step": 4727
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8025037418970047e-05,
+      "loss": 1.1588,
+      "step": 4728
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8024107871460678e-05,
+      "loss": 1.2455,
+      "step": 4729
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8023178129229808e-05,
+      "loss": 1.2067,
+      "step": 4730
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8022248192300002e-05,
+      "loss": 1.3502,
+      "step": 4731
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8021318060693823e-05,
+      "loss": 1.2802,
+      "step": 4732
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8020387734433848e-05,
+      "loss": 1.2177,
+      "step": 4733
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801945721354265e-05,
+      "loss": 1.1993,
+      "step": 4734
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8018526498042805e-05,
+      "loss": 1.3438,
+      "step": 4735
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8017595587956907e-05,
+      "loss": 1.2475,
+      "step": 4736
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801666448330754e-05,
+      "loss": 1.1966,
+      "step": 4737
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8015733184117307e-05,
+      "loss": 1.148,
+      "step": 4738
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8014801690408798e-05,
+      "loss": 1.002,
+      "step": 4739
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8013870002204625e-05,
+      "loss": 1.2556,
+      "step": 4740
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8012938119527396e-05,
+      "loss": 1.0163,
+      "step": 4741
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8012006042399724e-05,
+      "loss": 1.2548,
+      "step": 4742
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8011073770844225e-05,
+      "loss": 1.0731,
+      "step": 4743
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8010141304883527e-05,
+      "loss": 1.1972,
+      "step": 4744
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800920864454026e-05,
+      "loss": 1.2046,
+      "step": 4745
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8008275789837047e-05,
+      "loss": 1.0991,
+      "step": 4746
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8007342740796538e-05,
+      "loss": 1.2333,
+      "step": 4747
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8006409497441364e-05,
+      "loss": 1.2721,
+      "step": 4748
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8005476059794183e-05,
+      "loss": 1.1662,
+      "step": 4749
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8004542427877635e-05,
+      "loss": 1.215,
+      "step": 4750
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8003608601714388e-05,
+      "loss": 1.0297,
+      "step": 4751
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8002674581327096e-05,
+      "loss": 1.1129,
+      "step": 4752
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8001740366738426e-05,
+      "loss": 1.0893,
+      "step": 4753
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8000805957971054e-05,
+      "loss": 1.196,
+      "step": 4754
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7999871355047647e-05,
+      "loss": 1.285,
+      "step": 4755
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799893655799089e-05,
+      "loss": 1.1989,
+      "step": 4756
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7998001566823466e-05,
+      "loss": 1.1569,
+      "step": 4757
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7997066381568066e-05,
+      "loss": 1.2078,
+      "step": 4758
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7996131002247382e-05,
+      "loss": 1.2828,
+      "step": 4759
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7995195428884114e-05,
+      "loss": 1.2915,
+      "step": 4760
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7994259661500967e-05,
+      "loss": 1.1099,
+      "step": 4761
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7993323700120648e-05,
+      "loss": 1.2658,
+      "step": 4762
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7992387544765874e-05,
+      "loss": 1.2099,
+      "step": 4763
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7991451195459356e-05,
+      "loss": 1.1682,
+      "step": 4764
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7990514652223818e-05,
+      "loss": 1.1789,
+      "step": 4765
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798957791508199e-05,
+      "loss": 1.1554,
+      "step": 4766
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7988640984056602e-05,
+      "loss": 1.2103,
+      "step": 4767
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7987703859170393e-05,
+      "loss": 0.8894,
+      "step": 4768
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79867665404461e-05,
+      "loss": 1.2116,
+      "step": 4769
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7985829027906475e-05,
+      "loss": 1.1827,
+      "step": 4770
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798489132157426e-05,
+      "loss": 1.1212,
+      "step": 4771
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798395342147222e-05,
+      "loss": 0.9452,
+      "step": 4772
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798301532762311e-05,
+      "loss": 1.1343,
+      "step": 4773
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7982077040049692e-05,
+      "loss": 1.2458,
+      "step": 4774
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7981138558774737e-05,
+      "loss": 1.1393,
+      "step": 4775
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7980199883821026e-05,
+      "loss": 1.166,
+      "step": 4776
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797926101521133e-05,
+      "loss": 1.1869,
+      "step": 4777
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7978321952968435e-05,
+      "loss": 1.185,
+      "step": 4778
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797738269711513e-05,
+      "loss": 1.2683,
+      "step": 4779
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797644324767421e-05,
+      "loss": 1.3071,
+      "step": 4780
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7975503604668468e-05,
+      "loss": 1.2399,
+      "step": 4781
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797456376812071e-05,
+      "loss": 1.1516,
+      "step": 4782
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797362373805374e-05,
+      "loss": 1.1768,
+      "step": 4783
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7972683514490372e-05,
+      "loss": 1.2769,
+      "step": 4784
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797174309745342e-05,
+      "loss": 1.1238,
+      "step": 4785
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7970802486965712e-05,
+      "loss": 1.2939,
+      "step": 4786
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7969861683050064e-05,
+      "loss": 1.2426,
+      "step": 4787
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7968920685729317e-05,
+      "loss": 1.2382,
+      "step": 4788
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7967979495026296e-05,
+      "loss": 1.2317,
+      "step": 4789
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7967038110963847e-05,
+      "loss": 1.0545,
+      "step": 4790
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7966096533564813e-05,
+      "loss": 1.2058,
+      "step": 4791
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7965154762852044e-05,
+      "loss": 1.1392,
+      "step": 4792
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7964212798848393e-05,
+      "loss": 1.1853,
+      "step": 4793
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796327064157672e-05,
+      "loss": 1.16,
+      "step": 4794
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7962328291059886e-05,
+      "loss": 1.0877,
+      "step": 4795
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7961385747320763e-05,
+      "loss": 1.1864,
+      "step": 4796
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796044301038222e-05,
+      "loss": 1.0445,
+      "step": 4797
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795950008026714e-05,
+      "loss": 1.3991,
+      "step": 4798
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7958556956998397e-05,
+      "loss": 1.1403,
+      "step": 4799
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795761364059888e-05,
+      "loss": 1.1768,
+      "step": 4800
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7956670131091486e-05,
+      "loss": 1.1068,
+      "step": 4801
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7955726428499107e-05,
+      "loss": 1.1336,
+      "step": 4802
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7954782532844643e-05,
+      "loss": 1.2327,
+      "step": 4803
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7953838444151004e-05,
+      "loss": 1.096,
+      "step": 4804
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7952894162441094e-05,
+      "loss": 1.3171,
+      "step": 4805
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795194968773783e-05,
+      "loss": 1.215,
+      "step": 4806
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7951005020064142e-05,
+      "loss": 1.1104,
+      "step": 4807
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7950060159442934e-05,
+      "loss": 1.2194,
+      "step": 4808
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7949115105897155e-05,
+      "loss": 1.3472,
+      "step": 4809
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7948169859449726e-05,
+      "loss": 0.9832,
+      "step": 4810
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7947224420123587e-05,
+      "loss": 1.1704,
+      "step": 4811
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7946278787941687e-05,
+      "loss": 1.1149,
+      "step": 4812
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7945332962926966e-05,
+      "loss": 1.072,
+      "step": 4813
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7944386945102387e-05,
+      "loss": 1.1788,
+      "step": 4814
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7943440734490893e-05,
+      "loss": 1.173,
+      "step": 4815
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794249433111546e-05,
+      "loss": 1.1814,
+      "step": 4816
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7941547734999043e-05,
+      "loss": 1.142,
+      "step": 4817
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794060094616462e-05,
+      "loss": 1.2165,
+      "step": 4818
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7939653964635163e-05,
+      "loss": 1.1556,
+      "step": 4819
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7938706790433655e-05,
+      "loss": 1.221,
+      "step": 4820
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793775942358308e-05,
+      "loss": 1.15,
+      "step": 4821
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7936811864106425e-05,
+      "loss": 0.9768,
+      "step": 4822
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793586411202669e-05,
+      "loss": 1.1582,
+      "step": 4823
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793491616736687e-05,
+      "loss": 1.1269,
+      "step": 4824
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7933968030149972e-05,
+      "loss": 1.1679,
+      "step": 4825
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7933019700399006e-05,
+      "loss": 1.233,
+      "step": 4826
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793207117813698e-05,
+      "loss": 1.2528,
+      "step": 4827
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793112246338691e-05,
+      "loss": 1.2077,
+      "step": 4828
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7930173556171824e-05,
+      "loss": 1.1284,
+      "step": 4829
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792922445651475e-05,
+      "loss": 1.2977,
+      "step": 4830
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7928275164438715e-05,
+      "loss": 1.098,
+      "step": 4831
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792732567996676e-05,
+      "loss": 1.1841,
+      "step": 4832
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7926376003121922e-05,
+      "loss": 1.2066,
+      "step": 4833
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792542613392725e-05,
+      "loss": 1.0484,
+      "step": 4834
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7924476072405795e-05,
+      "loss": 1.1887,
+      "step": 4835
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792352581858061e-05,
+      "loss": 1.0725,
+      "step": 4836
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7922575372474755e-05,
+      "loss": 1.103,
+      "step": 4837
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7921624734111292e-05,
+      "loss": 1.1317,
+      "step": 4838
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79206739035133e-05,
+      "loss": 1.1898,
+      "step": 4839
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7919722880703843e-05,
+      "loss": 1.3799,
+      "step": 4840
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7918771665706e-05,
+      "loss": 1.0376,
+      "step": 4841
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7917820258542863e-05,
+      "loss": 1.2095,
+      "step": 4842
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791686865923751e-05,
+      "loss": 1.1917,
+      "step": 4843
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7915916867813037e-05,
+      "loss": 1.3788,
+      "step": 4844
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7914964884292543e-05,
+      "loss": 1.1922,
+      "step": 4845
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7914012708699126e-05,
+      "loss": 1.166,
+      "step": 4846
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7913060341055895e-05,
+      "loss": 1.196,
+      "step": 4847
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7912107781385963e-05,
+      "loss": 1.2649,
+      "step": 4848
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7911155029712444e-05,
+      "loss": 1.3121,
+      "step": 4849
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7910202086058458e-05,
+      "loss": 1.1829,
+      "step": 4850
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790924895044713e-05,
+      "loss": 1.117,
+      "step": 4851
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790829562290159e-05,
+      "loss": 1.0147,
+      "step": 4852
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7907342103444975e-05,
+      "loss": 1.2144,
+      "step": 4853
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790638839210042e-05,
+      "loss": 1.1133,
+      "step": 4854
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790543448889107e-05,
+      "loss": 1.0954,
+      "step": 4855
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7904480393840074e-05,
+      "loss": 1.1421,
+      "step": 4856
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7903526106970585e-05,
+      "loss": 1.2443,
+      "step": 4857
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790257162830576e-05,
+      "loss": 1.2082,
+      "step": 4858
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7901616957868766e-05,
+      "loss": 1.1645,
+      "step": 4859
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7900662095682762e-05,
+      "loss": 1.263,
+      "step": 4860
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7899707041770925e-05,
+      "loss": 1.1454,
+      "step": 4861
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7898751796156433e-05,
+      "loss": 1.2076,
+      "step": 4862
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789779635886246e-05,
+      "loss": 1.2343,
+      "step": 4863
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7896840729912198e-05,
+      "loss": 1.2501,
+      "step": 4864
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7895884909328835e-05,
+      "loss": 1.2328,
+      "step": 4865
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789492889713557e-05,
+      "loss": 1.2724,
+      "step": 4866
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7893972693355595e-05,
+      "loss": 1.2504,
+      "step": 4867
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7893016298012117e-05,
+      "loss": 1.3152,
+      "step": 4868
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7892059711128346e-05,
+      "loss": 1.1047,
+      "step": 4869
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.78911029327275e-05,
+      "loss": 1.1328,
+      "step": 4870
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7890145962832786e-05,
+      "loss": 1.169,
+      "step": 4871
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788918880146744e-05,
+      "loss": 1.2325,
+      "step": 4872
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788823144865468e-05,
+      "loss": 1.3383,
+      "step": 4873
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7887273904417742e-05,
+      "loss": 1.2686,
+      "step": 4874
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7886316168779862e-05,
+      "loss": 1.1315,
+      "step": 4875
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7885358241764282e-05,
+      "loss": 1.1547,
+      "step": 4876
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7884400123394243e-05,
+      "loss": 1.1333,
+      "step": 4877
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7883441813693006e-05,
+      "loss": 1.086,
+      "step": 4878
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7882483312683816e-05,
+      "loss": 1.3296,
+      "step": 4879
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788152462038994e-05,
+      "loss": 1.1954,
+      "step": 4880
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7880565736834642e-05,
+      "loss": 1.2637,
+      "step": 4881
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7879606662041186e-05,
+      "loss": 1.1838,
+      "step": 4882
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.787864739603285e-05,
+      "loss": 1.2556,
+      "step": 4883
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7877687938832915e-05,
+      "loss": 1.0493,
+      "step": 4884
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7876728290464658e-05,
+      "loss": 1.1653,
+      "step": 4885
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.787576845095137e-05,
+      "loss": 1.2447,
+      "step": 4886
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7874808420316345e-05,
+      "loss": 1.1719,
+      "step": 4887
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.787384819858288e-05,
+      "loss": 1.1338,
+      "step": 4888
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.787288778577427e-05,
+      "loss": 1.3006,
+      "step": 4889
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7871927181913832e-05,
+      "loss": 1.1922,
+      "step": 4890
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.787096638702487e-05,
+      "loss": 1.1535,
+      "step": 4891
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78700054011307e-05,
+      "loss": 1.0762,
+      "step": 4892
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7869044224254648e-05,
+      "loss": 1.1705,
+      "step": 4893
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786808285642003e-05,
+      "loss": 1.1094,
+      "step": 4894
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7867121297650184e-05,
+      "loss": 1.1163,
+      "step": 4895
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786615954796844e-05,
+      "loss": 1.2695,
+      "step": 4896
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7865197607398133e-05,
+      "loss": 1.1958,
+      "step": 4897
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7864235475962616e-05,
+      "loss": 1.0678,
+      "step": 4898
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786327315368523e-05,
+      "loss": 1.0927,
+      "step": 4899
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7862310640589328e-05,
+      "loss": 1.1125,
+      "step": 4900
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786134793669827e-05,
+      "loss": 1.1665,
+      "step": 4901
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7860385042035418e-05,
+      "loss": 1.1898,
+      "step": 4902
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7859421956624135e-05,
+      "loss": 1.0441,
+      "step": 4903
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7858458680487798e-05,
+      "loss": 1.3148,
+      "step": 4904
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785749521364978e-05,
+      "loss": 1.364,
+      "step": 4905
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7856531556133457e-05,
+      "loss": 1.228,
+      "step": 4906
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785556770796222e-05,
+      "loss": 1.1547,
+      "step": 4907
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785460366915946e-05,
+      "loss": 1.2189,
+      "step": 4908
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7853639439748564e-05,
+      "loss": 1.0826,
+      "step": 4909
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785267501975294e-05,
+      "loss": 1.2938,
+      "step": 4910
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7851710409195987e-05,
+      "loss": 1.2558,
+      "step": 4911
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785074560810111e-05,
+      "loss": 1.3677,
+      "step": 4912
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7849780616491726e-05,
+      "loss": 1.1879,
+      "step": 4913
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7848815434391254e-05,
+      "loss": 1.1838,
+      "step": 4914
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784785006182311e-05,
+      "loss": 1.1834,
+      "step": 4915
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784688449881073e-05,
+      "loss": 1.2172,
+      "step": 4916
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784591874537754e-05,
+      "loss": 1.2352,
+      "step": 4917
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784495280154697e-05,
+      "loss": 1.1442,
+      "step": 4918
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784398666734247e-05,
+      "loss": 1.1756,
+      "step": 4919
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784302034278748e-05,
+      "loss": 1.0384,
+      "step": 4920
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7842053827905457e-05,
+      "loss": 1.1404,
+      "step": 4921
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784108712271985e-05,
+      "loss": 1.1515,
+      "step": 4922
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7840120227254115e-05,
+      "loss": 1.2761,
+      "step": 4923
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783915314153172e-05,
+      "loss": 1.2855,
+      "step": 4924
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783818586557613e-05,
+      "loss": 1.263,
+      "step": 4925
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7837218399410822e-05,
+      "loss": 1.1703,
+      "step": 4926
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783625074305927e-05,
+      "loss": 1.1732,
+      "step": 4927
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7835282896544963e-05,
+      "loss": 1.2093,
+      "step": 4928
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783431485989138e-05,
+      "loss": 1.279,
+      "step": 4929
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7833346633122013e-05,
+      "loss": 1.1201,
+      "step": 4930
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7832378216260365e-05,
+      "loss": 1.0789,
+      "step": 4931
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7831409609329927e-05,
+      "loss": 1.4208,
+      "step": 4932
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7830440812354216e-05,
+      "loss": 1.2144,
+      "step": 4933
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782947182535673e-05,
+      "loss": 1.3143,
+      "step": 4934
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782850264836099e-05,
+      "loss": 1.1904,
+      "step": 4935
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782753328139051e-05,
+      "loss": 1.1482,
+      "step": 4936
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782656372446882e-05,
+      "loss": 1.1225,
+      "step": 4937
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7825593977619443e-05,
+      "loss": 1.2917,
+      "step": 4938
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782462404086592e-05,
+      "loss": 1.2618,
+      "step": 4939
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782365391423178e-05,
+      "loss": 1.2021,
+      "step": 4940
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7822683597740568e-05,
+      "loss": 1.1552,
+      "step": 4941
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782171309141583e-05,
+      "loss": 1.2686,
+      "step": 4942
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782074239528112e-05,
+      "loss": 1.2688,
+      "step": 4943
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781977150935999e-05,
+      "loss": 1.1702,
+      "step": 4944
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7818800433676e-05,
+      "loss": 1.111,
+      "step": 4945
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781782916825272e-05,
+      "loss": 1.1936,
+      "step": 4946
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781685771311372e-05,
+      "loss": 1.1792,
+      "step": 4947
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781588606828257e-05,
+      "loss": 1.3039,
+      "step": 4948
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7814914233782848e-05,
+      "loss": 1.3005,
+      "step": 4949
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7813942209638148e-05,
+      "loss": 1.0665,
+      "step": 4950
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7812969995872044e-05,
+      "loss": 1.1096,
+      "step": 4951
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781199759250814e-05,
+      "loss": 1.0724,
+      "step": 4952
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7811024999570023e-05,
+      "loss": 1.0811,
+      "step": 4953
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7810052217081306e-05,
+      "loss": 1.1018,
+      "step": 4954
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7809079245065586e-05,
+      "loss": 1.104,
+      "step": 4955
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7808106083546478e-05,
+      "loss": 1.1163,
+      "step": 4956
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7807132732547603e-05,
+      "loss": 1.1433,
+      "step": 4957
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7806159192092575e-05,
+      "loss": 1.063,
+      "step": 4958
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7805185462205018e-05,
+      "loss": 0.9597,
+      "step": 4959
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7804211542908568e-05,
+      "loss": 1.18,
+      "step": 4960
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780323743422685e-05,
+      "loss": 1.1668,
+      "step": 4961
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7802263136183514e-05,
+      "loss": 1.2204,
+      "step": 4962
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780128864880219e-05,
+      "loss": 1.1633,
+      "step": 4963
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780031397210654e-05,
+      "loss": 1.207,
+      "step": 4964
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7799339106120205e-05,
+      "loss": 1.1257,
+      "step": 4965
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7798364050866853e-05,
+      "loss": 1.2757,
+      "step": 4966
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7797388806370132e-05,
+      "loss": 1.1498,
+      "step": 4967
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779641337265372e-05,
+      "loss": 1.1629,
+      "step": 4968
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7795437749741283e-05,
+      "loss": 1.1475,
+      "step": 4969
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77944619376565e-05,
+      "loss": 1.0919,
+      "step": 4970
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7793485936423045e-05,
+      "loss": 1.2676,
+      "step": 4971
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7792509746064608e-05,
+      "loss": 1.1399,
+      "step": 4972
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7791533366604876e-05,
+      "loss": 1.1169,
+      "step": 4973
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7790556798067543e-05,
+      "loss": 1.1265,
+      "step": 4974
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7789580040476305e-05,
+      "loss": 1.169,
+      "step": 4975
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7788603093854872e-05,
+      "loss": 1.2093,
+      "step": 4976
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7787625958226947e-05,
+      "loss": 1.2216,
+      "step": 4977
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778664863361624e-05,
+      "loss": 1.064,
+      "step": 4978
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7785671120046472e-05,
+      "loss": 1.1565,
+      "step": 4979
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7784693417541364e-05,
+      "loss": 1.2669,
+      "step": 4980
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7783715526124637e-05,
+      "loss": 1.2843,
+      "step": 4981
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778273744582003e-05,
+      "loss": 1.0614,
+      "step": 4982
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7781759176651273e-05,
+      "loss": 1.1013,
+      "step": 4983
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77807807186421e-05,
+      "loss": 1.0415,
+      "step": 4984
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777980207181627e-05,
+      "loss": 1.1262,
+      "step": 4985
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7778823236197515e-05,
+      "loss": 1.0934,
+      "step": 4986
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7777844211809602e-05,
+      "loss": 1.0749,
+      "step": 4987
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7776864998676284e-05,
+      "loss": 1.1954,
+      "step": 4988
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777588559682132e-05,
+      "loss": 1.2794,
+      "step": 4989
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7774906006268482e-05,
+      "loss": 1.2646,
+      "step": 4990
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777392622704154e-05,
+      "loss": 1.182,
+      "step": 4991
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7772946259164272e-05,
+      "loss": 1.0552,
+      "step": 4992
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777196610266046e-05,
+      "loss": 1.1665,
+      "step": 4993
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7770985757553883e-05,
+      "loss": 1.0284,
+      "step": 4994
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7770005223868336e-05,
+      "loss": 1.2822,
+      "step": 4995
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7769024501627617e-05,
+      "loss": 1.1975,
+      "step": 4996
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7768043590855514e-05,
+      "loss": 1.2953,
+      "step": 4997
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7767062491575843e-05,
+      "loss": 1.1339,
+      "step": 4998
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7766081203812408e-05,
+      "loss": 1.1741,
+      "step": 4999
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776509972758902e-05,
+      "loss": 1.1782,
+      "step": 5000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7764118062929497e-05,
+      "loss": 1.1073,
+      "step": 5001
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7763136209857666e-05,
+      "loss": 1.3399,
+      "step": 5002
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7762154168397345e-05,
+      "loss": 1.1678,
+      "step": 5003
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7761171938572373e-05,
+      "loss": 1.0737,
+      "step": 5004
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776018952040658e-05,
+      "loss": 1.2144,
+      "step": 5005
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775920691392381e-05,
+      "loss": 1.2394,
+      "step": 5006
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7758224119147908e-05,
+      "loss": 1.1614,
+      "step": 5007
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7757241136102723e-05,
+      "loss": 1.1073,
+      "step": 5008
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775625796481211e-05,
+      "loss": 1.0518,
+      "step": 5009
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7755274605299924e-05,
+      "loss": 1.1558,
+      "step": 5010
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7754291057590033e-05,
+      "loss": 1.0965,
+      "step": 5011
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77533073217063e-05,
+      "loss": 1.0076,
+      "step": 5012
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7752323397672603e-05,
+      "loss": 1.1224,
+      "step": 5013
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775133928551281e-05,
+      "loss": 1.1009,
+      "step": 5014
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7750354985250817e-05,
+      "loss": 1.2232,
+      "step": 5015
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7749370496910497e-05,
+      "loss": 1.1572,
+      "step": 5016
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7748385820515744e-05,
+      "loss": 1.1213,
+      "step": 5017
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7747400956090457e-05,
+      "loss": 1.3047,
+      "step": 5018
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774641590365853e-05,
+      "loss": 1.2586,
+      "step": 5019
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7745430663243874e-05,
+      "loss": 1.2578,
+      "step": 5020
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774444523487039e-05,
+      "loss": 1.221,
+      "step": 5021
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7743459618562e-05,
+      "loss": 1.0642,
+      "step": 5022
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7742473814342616e-05,
+      "loss": 1.146,
+      "step": 5023
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7741487822236163e-05,
+      "loss": 1.1417,
+      "step": 5024
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7740501642266568e-05,
+      "loss": 1.0913,
+      "step": 5025
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773951527445776e-05,
+      "loss": 1.1254,
+      "step": 5026
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7738528718833677e-05,
+      "loss": 1.3149,
+      "step": 5027
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7737541975418263e-05,
+      "loss": 1.0102,
+      "step": 5028
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7736555044235458e-05,
+      "loss": 1.0261,
+      "step": 5029
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7735567925309217e-05,
+      "loss": 1.1477,
+      "step": 5030
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773458061866349e-05,
+      "loss": 1.2588,
+      "step": 5031
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7733593124322238e-05,
+      "loss": 1.0925,
+      "step": 5032
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773260544230942e-05,
+      "loss": 1.1974,
+      "step": 5033
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773161757264902e-05,
+      "loss": 0.8927,
+      "step": 5034
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773062951536499e-05,
+      "loss": 1.3152,
+      "step": 5035
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7729641270481323e-05,
+      "loss": 1.2931,
+      "step": 5036
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7728652838021988e-05,
+      "loss": 1.2002,
+      "step": 5037
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772766421801098e-05,
+      "loss": 1.2058,
+      "step": 5038
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772667541047229e-05,
+      "loss": 1.1512,
+      "step": 5039
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772568641542991e-05,
+      "loss": 1.0049,
+      "step": 5040
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7724697232907838e-05,
+      "loss": 1.0291,
+      "step": 5041
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7723707862930086e-05,
+      "loss": 0.9991,
+      "step": 5042
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772271830552066e-05,
+      "loss": 1.2116,
+      "step": 5043
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772172856070357e-05,
+      "loss": 1.129,
+      "step": 5044
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7720738628502836e-05,
+      "loss": 1.1919,
+      "step": 5045
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771974850894248e-05,
+      "loss": 1.1544,
+      "step": 5046
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7718758202046532e-05,
+      "loss": 1.1384,
+      "step": 5047
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771776770783902e-05,
+      "loss": 1.0901,
+      "step": 5048
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771677702634399e-05,
+      "loss": 1.1332,
+      "step": 5049
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771578615758547e-05,
+      "loss": 1.2097,
+      "step": 5050
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7714795101587512e-05,
+      "loss": 1.2831,
+      "step": 5051
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7713803858374164e-05,
+      "loss": 1.1107,
+      "step": 5052
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7712812427969485e-05,
+      "loss": 1.1985,
+      "step": 5053
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7711820810397532e-05,
+      "loss": 1.2198,
+      "step": 5054
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7710829005682364e-05,
+      "loss": 1.2861,
+      "step": 5055
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7709837013848056e-05,
+      "loss": 1.1558,
+      "step": 5056
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.770884483491867e-05,
+      "loss": 1.241,
+      "step": 5057
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77078524689183e-05,
+      "loss": 1.0824,
+      "step": 5058
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7706859915871016e-05,
+      "loss": 1.1328,
+      "step": 5059
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7705867175800905e-05,
+      "loss": 1.1675,
+      "step": 5060
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.770487424873206e-05,
+      "loss": 1.1528,
+      "step": 5061
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7703881134688576e-05,
+      "loss": 1.2177,
+      "step": 5062
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7702887833694554e-05,
+      "loss": 1.144,
+      "step": 5063
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77018943457741e-05,
+      "loss": 1.0616,
+      "step": 5064
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7700900670951317e-05,
+      "loss": 1.2789,
+      "step": 5065
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7699906809250324e-05,
+      "loss": 1.1483,
+      "step": 5066
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769891276069524e-05,
+      "loss": 1.1924,
+      "step": 5067
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7697918525310177e-05,
+      "loss": 1.2682,
+      "step": 5068
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7696924103119277e-05,
+      "loss": 1.231,
+      "step": 5069
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7695929494146662e-05,
+      "loss": 1.2562,
+      "step": 5070
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7694934698416472e-05,
+      "loss": 1.2735,
+      "step": 5071
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7693939715952847e-05,
+      "loss": 1.1235,
+      "step": 5072
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7692944546779932e-05,
+      "loss": 1.1903,
+      "step": 5073
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7691949190921877e-05,
+      "loss": 1.3722,
+      "step": 5074
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7690953648402835e-05,
+      "loss": 1.2318,
+      "step": 5075
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768995791924697e-05,
+      "loss": 1.2012,
+      "step": 5076
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7688962003478438e-05,
+      "loss": 1.2801,
+      "step": 5077
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768796590112141e-05,
+      "loss": 1.081,
+      "step": 5078
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7686969612200063e-05,
+      "loss": 1.1475,
+      "step": 5079
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7685973136738568e-05,
+      "loss": 1.2068,
+      "step": 5080
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768497647476111e-05,
+      "loss": 1.1648,
+      "step": 5081
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768397962629187e-05,
+      "loss": 1.1459,
+      "step": 5082
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7682982591355043e-05,
+      "loss": 1.0702,
+      "step": 5083
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7681985369974827e-05,
+      "loss": 1.2013,
+      "step": 5084
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7680987962175415e-05,
+      "loss": 1.1766,
+      "step": 5085
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7679990367981015e-05,
+      "loss": 1.2192,
+      "step": 5086
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7678992587415834e-05,
+      "loss": 1.1743,
+      "step": 5087
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767799462050409e-05,
+      "loss": 1.2173,
+      "step": 5088
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767699646726999e-05,
+      "loss": 1.1376,
+      "step": 5089
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767599812773777e-05,
+      "loss": 1.1454,
+      "step": 5090
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7674999601931645e-05,
+      "loss": 1.2653,
+      "step": 5091
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7674000889875854e-05,
+      "loss": 1.2373,
+      "step": 5092
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7673001991594625e-05,
+      "loss": 1.2145,
+      "step": 5093
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7672002907112203e-05,
+      "loss": 1.1628,
+      "step": 5094
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.767100363645284e-05,
+      "loss": 1.001,
+      "step": 5095
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7670004179640773e-05,
+      "loss": 1.1061,
+      "step": 5096
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7669004536700263e-05,
+      "loss": 1.3005,
+      "step": 5097
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7668004707655568e-05,
+      "loss": 1.3004,
+      "step": 5098
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7667004692530946e-05,
+      "loss": 1.2895,
+      "step": 5099
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766600449135067e-05,
+      "loss": 1.2756,
+      "step": 5100
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766500410413901e-05,
+      "loss": 1.0756,
+      "step": 5101
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766400353092024e-05,
+      "loss": 1.1381,
+      "step": 5102
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766300277171865e-05,
+      "loss": 1.2533,
+      "step": 5103
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766200182655851e-05,
+      "loss": 0.9823,
+      "step": 5104
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7661000695464125e-05,
+      "loss": 1.1393,
+      "step": 5105
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7659999378459783e-05,
+      "loss": 1.3189,
+      "step": 5106
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7658997875569782e-05,
+      "loss": 1.0236,
+      "step": 5107
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7657996186818427e-05,
+      "loss": 1.2387,
+      "step": 5108
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7656994312230023e-05,
+      "loss": 1.0515,
+      "step": 5109
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765599225182889e-05,
+      "loss": 1.0534,
+      "step": 5110
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7654990005639338e-05,
+      "loss": 1.1652,
+      "step": 5111
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765398757368569e-05,
+      "loss": 0.9494,
+      "step": 5112
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7652984955992275e-05,
+      "loss": 1.3109,
+      "step": 5113
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765198215258342e-05,
+      "loss": 1.2244,
+      "step": 5114
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7650979163483464e-05,
+      "loss": 1.1427,
+      "step": 5115
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7649975988716745e-05,
+      "loss": 1.0479,
+      "step": 5116
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7648972628307607e-05,
+      "loss": 1.2198,
+      "step": 5117
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7647969082280397e-05,
+      "loss": 1.2162,
+      "step": 5118
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7646965350659466e-05,
+      "loss": 1.2714,
+      "step": 5119
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764596143346918e-05,
+      "loss": 1.2288,
+      "step": 5120
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764495733073389e-05,
+      "loss": 1.2769,
+      "step": 5121
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764395304247797e-05,
+      "loss": 1.114,
+      "step": 5122
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7642948568725796e-05,
+      "loss": 1.1113,
+      "step": 5123
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7641943909501734e-05,
+      "loss": 1.0988,
+      "step": 5124
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7640939064830166e-05,
+      "loss": 1.1435,
+      "step": 5125
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7639934034735476e-05,
+      "loss": 1.1528,
+      "step": 5126
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7638928819242062e-05,
+      "loss": 1.1982,
+      "step": 5127
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7637923418374305e-05,
+      "loss": 1.1706,
+      "step": 5128
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763691783215661e-05,
+      "loss": 1.233,
+      "step": 5129
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7635912060613377e-05,
+      "loss": 1.2279,
+      "step": 5130
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763490610376902e-05,
+      "loss": 1.1637,
+      "step": 5131
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763389996164794e-05,
+      "loss": 1.2879,
+      "step": 5132
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763289363427456e-05,
+      "loss": 1.3021,
+      "step": 5133
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7631887121673302e-05,
+      "loss": 1.3103,
+      "step": 5134
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7630880423868585e-05,
+      "loss": 1.1671,
+      "step": 5135
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7629873540884843e-05,
+      "loss": 1.2222,
+      "step": 5136
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762886647274651e-05,
+      "loss": 1.1712,
+      "step": 5137
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762785921947802e-05,
+      "loss": 1.2401,
+      "step": 5138
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762685178110382e-05,
+      "loss": 0.9546,
+      "step": 5139
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762584415764836e-05,
+      "loss": 1.1666,
+      "step": 5140
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7624836349136088e-05,
+      "loss": 1.2798,
+      "step": 5141
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762382835559146e-05,
+      "loss": 1.1313,
+      "step": 5142
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762282017703894e-05,
+      "loss": 0.9908,
+      "step": 5143
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762181181350299e-05,
+      "loss": 1.2703,
+      "step": 5144
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7620803265008085e-05,
+      "loss": 1.058,
+      "step": 5145
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7619794531578698e-05,
+      "loss": 1.2246,
+      "step": 5146
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7618785613239305e-05,
+      "loss": 1.206,
+      "step": 5147
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761777651001439e-05,
+      "loss": 1.1191,
+      "step": 5148
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7616767221928443e-05,
+      "loss": 1.2353,
+      "step": 5149
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7615757749005954e-05,
+      "loss": 1.1752,
+      "step": 5150
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7614748091271426e-05,
+      "loss": 1.1977,
+      "step": 5151
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761373824874935e-05,
+      "loss": 1.1177,
+      "step": 5152
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7612728221464243e-05,
+      "loss": 1.1205,
+      "step": 5153
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7611718009440604e-05,
+      "loss": 1.2896,
+      "step": 5154
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7610707612702963e-05,
+      "loss": 1.218,
+      "step": 5155
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7609697031275824e-05,
+      "loss": 1.1788,
+      "step": 5156
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760868626518372e-05,
+      "loss": 1.1023,
+      "step": 5157
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760767531445118e-05,
+      "loss": 1.0203,
+      "step": 5158
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7606664179102725e-05,
+      "loss": 1.43,
+      "step": 5159
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7605652859162908e-05,
+      "loss": 1.1959,
+      "step": 5160
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760464135465626e-05,
+      "loss": 1.2471,
+      "step": 5161
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760362966560733e-05,
+      "loss": 1.1151,
+      "step": 5162
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760261779204067e-05,
+      "loss": 1.3018,
+      "step": 5163
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7601605733980837e-05,
+      "loss": 1.086,
+      "step": 5164
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760059349145239e-05,
+      "loss": 1.2541,
+      "step": 5165
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7599581064479886e-05,
+      "loss": 1.2109,
+      "step": 5166
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7598568453087902e-05,
+      "loss": 1.0335,
+      "step": 5167
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759755565730101e-05,
+      "loss": 1.1283,
+      "step": 5168
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759654267714378e-05,
+      "loss": 1.0483,
+      "step": 5169
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7595529512640807e-05,
+      "loss": 1.1213,
+      "step": 5170
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7594516163816666e-05,
+      "loss": 1.2334,
+      "step": 5171
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7593502630695954e-05,
+      "loss": 1.3429,
+      "step": 5172
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7592488913303265e-05,
+      "loss": 1.2369,
+      "step": 5173
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7591475011663197e-05,
+      "loss": 1.1749,
+      "step": 5174
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7590460925800358e-05,
+      "loss": 1.1957,
+      "step": 5175
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7589446655739352e-05,
+      "loss": 1.2787,
+      "step": 5176
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7588432201504798e-05,
+      "loss": 1.1568,
+      "step": 5177
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7587417563121314e-05,
+      "loss": 1.0633,
+      "step": 5178
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7586402740613514e-05,
+      "loss": 1.123,
+      "step": 5179
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7585387734006033e-05,
+      "loss": 1.0089,
+      "step": 5180
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75843725433235e-05,
+      "loss": 1.2438,
+      "step": 5181
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7583357168590552e-05,
+      "loss": 1.1824,
+      "step": 5182
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7582341609831824e-05,
+      "loss": 1.074,
+      "step": 5183
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7581325867071967e-05,
+      "loss": 1.1544,
+      "step": 5184
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758030994033562e-05,
+      "loss": 1.1852,
+      "step": 5185
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7579293829647454e-05,
+      "loss": 1.1013,
+      "step": 5186
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757827753503211e-05,
+      "loss": 1.1424,
+      "step": 5187
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7577261056514257e-05,
+      "loss": 1.2061,
+      "step": 5188
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7576244394118562e-05,
+      "loss": 1.2856,
+      "step": 5189
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75752275478697e-05,
+      "loss": 1.1697,
+      "step": 5190
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757421051779234e-05,
+      "loss": 1.1502,
+      "step": 5191
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7573193303911167e-05,
+      "loss": 1.1039,
+      "step": 5192
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7572175906250866e-05,
+      "loss": 1.1438,
+      "step": 5193
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7571158324836123e-05,
+      "loss": 1.1246,
+      "step": 5194
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7570140559691635e-05,
+      "loss": 1.2048,
+      "step": 5195
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7569122610842097e-05,
+      "loss": 1.2296,
+      "step": 5196
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7568104478312214e-05,
+      "loss": 1.2531,
+      "step": 5197
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756708616212669e-05,
+      "loss": 1.216,
+      "step": 5198
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756606766231024e-05,
+      "loss": 1.1971,
+      "step": 5199
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756504897888758e-05,
+      "loss": 1.0739,
+      "step": 5200
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756403011188343e-05,
+      "loss": 1.2288,
+      "step": 5201
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7563011061322513e-05,
+      "loss": 1.1384,
+      "step": 5202
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756199182722956e-05,
+      "loss": 1.2045,
+      "step": 5203
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75609724096293e-05,
+      "loss": 1.1839,
+      "step": 5204
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755995280854648e-05,
+      "loss": 1.073,
+      "step": 5205
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755893302400584e-05,
+      "loss": 1.316,
+      "step": 5206
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7557913056032126e-05,
+      "loss": 1.0322,
+      "step": 5207
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7556892904650086e-05,
+      "loss": 1.2177,
+      "step": 5208
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7555872569884478e-05,
+      "loss": 1.2095,
+      "step": 5209
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7554852051760068e-05,
+      "loss": 1.173,
+      "step": 5210
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755383135030161e-05,
+      "loss": 1.2394,
+      "step": 5211
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755281046553389e-05,
+      "loss": 1.1801,
+      "step": 5212
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7551789397481666e-05,
+      "loss": 1.2426,
+      "step": 5213
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7550768146169722e-05,
+      "loss": 1.0616,
+      "step": 5214
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754974671162284e-05,
+      "loss": 1.33,
+      "step": 5215
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754872509386581e-05,
+      "loss": 1.1215,
+      "step": 5216
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754770329292342e-05,
+      "loss": 1.2135,
+      "step": 5217
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754668130882047e-05,
+      "loss": 1.1477,
+      "step": 5218
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7545659141581762e-05,
+      "loss": 1.1352,
+      "step": 5219
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7544636791232094e-05,
+      "loss": 1.1629,
+      "step": 5220
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754361425779628e-05,
+      "loss": 1.1672,
+      "step": 5221
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754259154129913e-05,
+      "loss": 1.2279,
+      "step": 5222
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754156864176547e-05,
+      "loss": 0.9942,
+      "step": 5223
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7540545559220116e-05,
+      "loss": 1.1339,
+      "step": 5224
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75395222936879e-05,
+      "loss": 1.0484,
+      "step": 5225
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7538498845193645e-05,
+      "loss": 1.0791,
+      "step": 5226
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7537475213762196e-05,
+      "loss": 1.1007,
+      "step": 5227
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753645139941839e-05,
+      "loss": 1.2489,
+      "step": 5228
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753542740218707e-05,
+      "loss": 1.2473,
+      "step": 5229
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7534403222093093e-05,
+      "loss": 1.2102,
+      "step": 5230
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7533378859161304e-05,
+      "loss": 0.9524,
+      "step": 5231
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753235431341657e-05,
+      "loss": 1.1006,
+      "step": 5232
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7531329584883747e-05,
+      "loss": 1.1414,
+      "step": 5233
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75303046735877e-05,
+      "loss": 0.9966,
+      "step": 5234
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7529279579553303e-05,
+      "loss": 1.0638,
+      "step": 5235
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.752825430280544e-05,
+      "loss": 1.2282,
+      "step": 5236
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.752722884336898e-05,
+      "loss": 1.1635,
+      "step": 5237
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7526203201268814e-05,
+      "loss": 1.3135,
+      "step": 5238
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7525177376529828e-05,
+      "loss": 1.1949,
+      "step": 5239
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.752415136917692e-05,
+      "loss": 1.2848,
+      "step": 5240
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7523125179234987e-05,
+      "loss": 1.1156,
+      "step": 5241
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.752209880672893e-05,
+      "loss": 1.1583,
+      "step": 5242
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7521072251683657e-05,
+      "loss": 1.1412,
+      "step": 5243
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7520045514124077e-05,
+      "loss": 1.0824,
+      "step": 5244
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751901859407511e-05,
+      "loss": 1.1935,
+      "step": 5245
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7517991491561676e-05,
+      "loss": 1.2625,
+      "step": 5246
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7516964206608697e-05,
+      "loss": 1.0305,
+      "step": 5247
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75159367392411e-05,
+      "loss": 1.1891,
+      "step": 5248
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7514909089483825e-05,
+      "loss": 1.2023,
+      "step": 5249
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7513881257361806e-05,
+      "loss": 1.1348,
+      "step": 5250
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7512853242899987e-05,
+      "loss": 1.1068,
+      "step": 5251
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751182504612331e-05,
+      "loss": 1.0895,
+      "step": 5252
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7510796667056736e-05,
+      "loss": 1.0659,
+      "step": 5253
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7509768105725215e-05,
+      "loss": 1.2444,
+      "step": 5254
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7508739362153708e-05,
+      "loss": 1.1093,
+      "step": 5255
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750771043636718e-05,
+      "loss": 1.1596,
+      "step": 5256
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7506681328390592e-05,
+      "loss": 1.2862,
+      "step": 5257
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750565203824893e-05,
+      "loss": 1.2638,
+      "step": 5258
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7504622565967166e-05,
+      "loss": 1.3876,
+      "step": 5259
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7503592911570283e-05,
+      "loss": 1.2432,
+      "step": 5260
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7502563075083267e-05,
+      "loss": 1.199,
+      "step": 5261
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7501533056531106e-05,
+      "loss": 1.2164,
+      "step": 5262
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7500502855938802e-05,
+      "loss": 1.207,
+      "step": 5263
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7499472473331353e-05,
+      "loss": 1.0881,
+      "step": 5264
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749844190873376e-05,
+      "loss": 1.2441,
+      "step": 5265
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7497411162171034e-05,
+      "loss": 1.2407,
+      "step": 5266
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749638023366819e-05,
+      "loss": 1.3109,
+      "step": 5267
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7495349123250242e-05,
+      "loss": 1.1514,
+      "step": 5268
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7494317830942214e-05,
+      "loss": 1.2001,
+      "step": 5269
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7493286356769135e-05,
+      "loss": 1.1171,
+      "step": 5270
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7492254700756032e-05,
+      "loss": 1.2978,
+      "step": 5271
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749122286292794e-05,
+      "loss": 1.2614,
+      "step": 5272
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.74901908433099e-05,
+      "loss": 1.3485,
+      "step": 5273
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7489158641926955e-05,
+      "loss": 1.2364,
+      "step": 5274
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748812625880416e-05,
+      "loss": 1.0072,
+      "step": 5275
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7487093693966553e-05,
+      "loss": 1.1546,
+      "step": 5276
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748606094743921e-05,
+      "loss": 1.1419,
+      "step": 5277
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7485028019247176e-05,
+      "loss": 0.9622,
+      "step": 5278
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748399490941553e-05,
+      "loss": 1.1843,
+      "step": 5279
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7482961617969338e-05,
+      "loss": 1.1282,
+      "step": 5280
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748192814493367e-05,
+      "loss": 1.3234,
+      "step": 5281
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748089449033361e-05,
+      "loss": 1.1566,
+      "step": 5282
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7479860654194244e-05,
+      "loss": 1.2884,
+      "step": 5283
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7478826636540653e-05,
+      "loss": 1.1786,
+      "step": 5284
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747779243739794e-05,
+      "loss": 1.0222,
+      "step": 5285
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747675805679119e-05,
+      "loss": 1.2006,
+      "step": 5286
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7475723494745516e-05,
+      "loss": 1.1311,
+      "step": 5287
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7474688751286014e-05,
+      "loss": 1.1845,
+      "step": 5288
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.74736538264378e-05,
+      "loss": 1.1187,
+      "step": 5289
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7472618720225988e-05,
+      "loss": 1.3046,
+      "step": 5290
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7471583432675692e-05,
+      "loss": 1.119,
+      "step": 5291
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747054796381204e-05,
+      "loss": 1.1415,
+      "step": 5292
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746951231366016e-05,
+      "loss": 1.0605,
+      "step": 5293
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7468476482245186e-05,
+      "loss": 1.1468,
+      "step": 5294
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7467440469592248e-05,
+      "loss": 1.2745,
+      "step": 5295
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746640427572649e-05,
+      "loss": 1.147,
+      "step": 5296
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7465367900673064e-05,
+      "loss": 1.3253,
+      "step": 5297
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7464331344457108e-05,
+      "loss": 1.0617,
+      "step": 5298
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7463294607103787e-05,
+      "loss": 1.2295,
+      "step": 5299
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746225768863825e-05,
+      "loss": 1.2258,
+      "step": 5300
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7461220589085668e-05,
+      "loss": 1.168,
+      "step": 5301
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.74601833084712e-05,
+      "loss": 1.2091,
+      "step": 5302
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745914584682003e-05,
+      "loss": 1.1481,
+      "step": 5303
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745810820415732e-05,
+      "loss": 1.0219,
+      "step": 5304
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745707038050826e-05,
+      "loss": 1.0364,
+      "step": 5305
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7456032375898034e-05,
+      "loss": 1.1201,
+      "step": 5306
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745499419035183e-05,
+      "loss": 1.1079,
+      "step": 5307
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7453955823894836e-05,
+      "loss": 1.1511,
+      "step": 5308
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7452917276552263e-05,
+      "loss": 1.0388,
+      "step": 5309
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.74518785483493e-05,
+      "loss": 1.1399,
+      "step": 5310
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7450839639311164e-05,
+      "loss": 1.1063,
+      "step": 5311
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744980054946306e-05,
+      "loss": 1.1005,
+      "step": 5312
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7448761278830203e-05,
+      "loss": 1.0827,
+      "step": 5313
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744772182743782e-05,
+      "loss": 1.2155,
+      "step": 5314
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7446682195311128e-05,
+      "loss": 1.1876,
+      "step": 5315
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7445642382475362e-05,
+      "loss": 1.2534,
+      "step": 5316
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7444602388955752e-05,
+      "loss": 1.0297,
+      "step": 5317
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7443562214777533e-05,
+      "loss": 1.2627,
+      "step": 5318
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7442521859965956e-05,
+      "loss": 1.0385,
+      "step": 5319
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7441481324546254e-05,
+      "loss": 1.1004,
+      "step": 5320
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7440440608543688e-05,
+      "loss": 1.0964,
+      "step": 5321
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743939971198351e-05,
+      "loss": 1.111,
+      "step": 5322
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7438358634890978e-05,
+      "loss": 1.1672,
+      "step": 5323
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7437317377291358e-05,
+      "loss": 1.2126,
+      "step": 5324
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7436275939209918e-05,
+      "loss": 1.1388,
+      "step": 5325
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743523432067193e-05,
+      "loss": 1.106,
+      "step": 5326
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7434192521702674e-05,
+      "loss": 1.1661,
+      "step": 5327
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7433150542327425e-05,
+      "loss": 1.2737,
+      "step": 5328
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7432108382571474e-05,
+      "loss": 1.1325,
+      "step": 5329
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743106604246011e-05,
+      "loss": 1.0684,
+      "step": 5330
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7430023522018627e-05,
+      "loss": 1.1506,
+      "step": 5331
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742898082127232e-05,
+      "loss": 1.186,
+      "step": 5332
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.74279379402465e-05,
+      "loss": 1.1306,
+      "step": 5333
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742689487896647e-05,
+      "loss": 1.2844,
+      "step": 5334
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7425851637457546e-05,
+      "loss": 1.2293,
+      "step": 5335
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7424808215745036e-05,
+      "loss": 1.0275,
+      "step": 5336
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742376461385427e-05,
+      "loss": 1.2691,
+      "step": 5337
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7422720831810572e-05,
+      "loss": 1.0159,
+      "step": 5338
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7421676869639264e-05,
+      "loss": 1.1721,
+      "step": 5339
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7420632727365686e-05,
+      "loss": 1.1812,
+      "step": 5340
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7419588405015176e-05,
+      "loss": 1.2025,
+      "step": 5341
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7418543902613074e-05,
+      "loss": 1.0751,
+      "step": 5342
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741749922018473e-05,
+      "loss": 1.2364,
+      "step": 5343
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741645435775549e-05,
+      "loss": 1.2561,
+      "step": 5344
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741540931535072e-05,
+      "loss": 1.242,
+      "step": 5345
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7414364092995773e-05,
+      "loss": 1.22,
+      "step": 5346
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741331869071601e-05,
+      "loss": 1.1501,
+      "step": 5347
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7412273108536805e-05,
+      "loss": 1.1314,
+      "step": 5348
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7411227346483534e-05,
+      "loss": 1.1866,
+      "step": 5349
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7410181404581567e-05,
+      "loss": 1.0742,
+      "step": 5350
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7409135282856297e-05,
+      "loss": 1.1552,
+      "step": 5351
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7408088981333095e-05,
+      "loss": 1.3785,
+      "step": 5352
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7407042500037362e-05,
+      "loss": 1.1669,
+      "step": 5353
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7405995838994495e-05,
+      "loss": 1.0618,
+      "step": 5354
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740494899822988e-05,
+      "loss": 1.209,
+      "step": 5355
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740390197776894e-05,
+      "loss": 1.2874,
+      "step": 5356
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740285477763707e-05,
+      "loss": 1.0323,
+      "step": 5357
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7401807397859682e-05,
+      "loss": 1.301,
+      "step": 5358
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.74007598384622e-05,
+      "loss": 1.1461,
+      "step": 5359
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7399712099470042e-05,
+      "loss": 1.168,
+      "step": 5360
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739866418090863e-05,
+      "loss": 1.1045,
+      "step": 5361
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7397616082803398e-05,
+      "loss": 0.9558,
+      "step": 5362
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739656780517978e-05,
+      "loss": 1.0009,
+      "step": 5363
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7395519348063214e-05,
+      "loss": 1.1076,
+      "step": 5364
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739447071147914e-05,
+      "loss": 1.2698,
+      "step": 5365
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7393421895453012e-05,
+      "loss": 1.2558,
+      "step": 5366
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7392372900010277e-05,
+      "loss": 1.0685,
+      "step": 5367
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7391323725176392e-05,
+      "loss": 1.2819,
+      "step": 5368
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7390274370976816e-05,
+      "loss": 1.1763,
+      "step": 5369
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7389224837437018e-05,
+      "loss": 1.204,
+      "step": 5370
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7388175124582462e-05,
+      "loss": 1.1326,
+      "step": 5371
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738712523243862e-05,
+      "loss": 1.2625,
+      "step": 5372
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738607516103098e-05,
+      "loss": 1.1571,
+      "step": 5373
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7385024910385014e-05,
+      "loss": 1.2326,
+      "step": 5374
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7383974480526213e-05,
+      "loss": 1.099,
+      "step": 5375
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7382923871480065e-05,
+      "loss": 1.323,
+      "step": 5376
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738187308327207e-05,
+      "loss": 1.2583,
+      "step": 5377
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7380822115927723e-05,
+      "loss": 1.1804,
+      "step": 5378
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.737977096947253e-05,
+      "loss": 1.3381,
+      "step": 5379
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7378719643932e-05,
+      "loss": 1.2214,
+      "step": 5380
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7377668139331638e-05,
+      "loss": 1.1229,
+      "step": 5381
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7376616455696972e-05,
+      "loss": 1.2479,
+      "step": 5382
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7375564593053518e-05,
+      "loss": 1.2529,
+      "step": 5383
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7374512551426804e-05,
+      "loss": 1.165,
+      "step": 5384
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7373460330842354e-05,
+      "loss": 1.2529,
+      "step": 5385
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7372407931325708e-05,
+      "loss": 1.0129,
+      "step": 5386
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7371355352902404e-05,
+      "loss": 1.1698,
+      "step": 5387
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7370302595597984e-05,
+      "loss": 1.081,
+      "step": 5388
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7369249659437996e-05,
+      "loss": 1.0568,
+      "step": 5389
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7368196544447987e-05,
+      "loss": 1.2184,
+      "step": 5390
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736714325065352e-05,
+      "loss": 1.1717,
+      "step": 5391
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7366089778080152e-05,
+      "loss": 1.0839,
+      "step": 5392
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7365036126753448e-05,
+      "loss": 1.1004,
+      "step": 5393
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7363982296698978e-05,
+      "loss": 1.2048,
+      "step": 5394
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7362928287942314e-05,
+      "loss": 1.1287,
+      "step": 5395
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7361874100509034e-05,
+      "loss": 1.1145,
+      "step": 5396
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7360819734424718e-05,
+      "loss": 1.0973,
+      "step": 5397
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7359765189714957e-05,
+      "loss": 1.031,
+      "step": 5398
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7358710466405337e-05,
+      "loss": 1.2492,
+      "step": 5399
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7357655564521457e-05,
+      "loss": 1.2327,
+      "step": 5400
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7356600484088914e-05,
+      "loss": 1.1154,
+      "step": 5401
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735554522513331e-05,
+      "loss": 1.1171,
+      "step": 5402
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735448978768026e-05,
+      "loss": 1.2497,
+      "step": 5403
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7353434171755366e-05,
+      "loss": 1.1839,
+      "step": 5404
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7352378377384254e-05,
+      "loss": 1.2617,
+      "step": 5405
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735132240459254e-05,
+      "loss": 1.2406,
+      "step": 5406
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7350266253405853e-05,
+      "loss": 1.2683,
+      "step": 5407
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7349209923849818e-05,
+      "loss": 1.2085,
+      "step": 5408
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734815341595007e-05,
+      "loss": 0.8396,
+      "step": 5409
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734709672973225e-05,
+      "loss": 1.0556,
+      "step": 5410
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7346039865222e-05,
+      "loss": 1.1155,
+      "step": 5411
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7344982822444968e-05,
+      "loss": 0.8953,
+      "step": 5412
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73439256014268e-05,
+      "loss": 1.2541,
+      "step": 5413
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734286820219316e-05,
+      "loss": 1.0729,
+      "step": 5414
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73418106247697e-05,
+      "loss": 1.1499,
+      "step": 5415
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7340752869182087e-05,
+      "loss": 1.2491,
+      "step": 5416
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7339694935455994e-05,
+      "loss": 1.2207,
+      "step": 5417
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7338636823617086e-05,
+      "loss": 1.3009,
+      "step": 5418
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7337578533691047e-05,
+      "loss": 1.0515,
+      "step": 5419
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7336520065703557e-05,
+      "loss": 1.2567,
+      "step": 5420
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73354614196803e-05,
+      "loss": 1.2339,
+      "step": 5421
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7334402595646965e-05,
+      "loss": 1.2005,
+      "step": 5422
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733334359362925e-05,
+      "loss": 1.1399,
+      "step": 5423
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7332284413652852e-05,
+      "loss": 1.2574,
+      "step": 5424
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7331225055743478e-05,
+      "loss": 1.1936,
+      "step": 5425
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733016551992683e-05,
+      "loss": 1.0982,
+      "step": 5426
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732910580622862e-05,
+      "loss": 1.1046,
+      "step": 5427
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7328045914674566e-05,
+      "loss": 1.2036,
+      "step": 5428
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732698584529039e-05,
+      "loss": 1.1192,
+      "step": 5429
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7325925598101817e-05,
+      "loss": 1.2978,
+      "step": 5430
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7324865173134576e-05,
+      "loss": 1.1382,
+      "step": 5431
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7323804570414395e-05,
+      "loss": 1.1717,
+      "step": 5432
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7322743789967017e-05,
+      "loss": 1.3793,
+      "step": 5433
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7321682831818184e-05,
+      "loss": 1.1052,
+      "step": 5434
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732062169599364e-05,
+      "loss": 1.15,
+      "step": 5435
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7319560382519137e-05,
+      "loss": 1.0787,
+      "step": 5436
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7318498891420426e-05,
+      "loss": 1.1419,
+      "step": 5437
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7317437222723274e-05,
+      "loss": 1.1979,
+      "step": 5438
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731637537645344e-05,
+      "loss": 1.2984,
+      "step": 5439
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731531335263669e-05,
+      "loss": 1.1263,
+      "step": 5440
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73142511512988e-05,
+      "loss": 1.321,
+      "step": 5441
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7313188772465546e-05,
+      "loss": 1.0992,
+      "step": 5442
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7312126216162707e-05,
+      "loss": 1.1943,
+      "step": 5443
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7311063482416068e-05,
+      "loss": 0.9361,
+      "step": 5444
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731000057125142e-05,
+      "loss": 1.2799,
+      "step": 5445
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730893748269456e-05,
+      "loss": 1.1527,
+      "step": 5446
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7307874216771276e-05,
+      "loss": 1.1278,
+      "step": 5447
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7306810773507376e-05,
+      "loss": 1.141,
+      "step": 5448
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7305747152928673e-05,
+      "loss": 1.0922,
+      "step": 5449
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7304683355060966e-05,
+      "loss": 1.252,
+      "step": 5450
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7303619379930083e-05,
+      "loss": 1.0752,
+      "step": 5451
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730255522756183e-05,
+      "loss": 1.1108,
+      "step": 5452
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730149089798204e-05,
+      "loss": 1.0846,
+      "step": 5453
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730042639121654e-05,
+      "loss": 1.1599,
+      "step": 5454
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729936170729116e-05,
+      "loss": 1.2619,
+      "step": 5455
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7298296846231738e-05,
+      "loss": 1.1856,
+      "step": 5456
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7297231808064118e-05,
+      "loss": 1.271,
+      "step": 5457
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729616659281414e-05,
+      "loss": 1.1628,
+      "step": 5458
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7295101200507656e-05,
+      "loss": 1.1591,
+      "step": 5459
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729403563117052e-05,
+      "loss": 1.2789,
+      "step": 5460
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729296988482859e-05,
+      "loss": 1.1019,
+      "step": 5461
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729190396150773e-05,
+      "loss": 1.1994,
+      "step": 5462
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7290837861233805e-05,
+      "loss": 1.1334,
+      "step": 5463
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7289771584032686e-05,
+      "loss": 1.0973,
+      "step": 5464
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728870512993025e-05,
+      "loss": 1.0485,
+      "step": 5465
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7287638498952375e-05,
+      "loss": 1.1697,
+      "step": 5466
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7286571691124945e-05,
+      "loss": 1.086,
+      "step": 5467
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728550470647385e-05,
+      "loss": 1.1251,
+      "step": 5468
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7284437545024984e-05,
+      "loss": 1.1367,
+      "step": 5469
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728337020680424e-05,
+      "loss": 1.2629,
+      "step": 5470
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728230269183752e-05,
+      "loss": 1.3166,
+      "step": 5471
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728123500015073e-05,
+      "loss": 1.1094,
+      "step": 5472
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728016713176978e-05,
+      "loss": 1.1979,
+      "step": 5473
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7279099086720584e-05,
+      "loss": 1.1184,
+      "step": 5474
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7278030865029062e-05,
+      "loss": 1.202,
+      "step": 5475
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727696246672113e-05,
+      "loss": 1.1401,
+      "step": 5476
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7275893891822724e-05,
+      "loss": 1.1838,
+      "step": 5477
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7274825140359767e-05,
+      "loss": 1.2864,
+      "step": 5478
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7273756212358204e-05,
+      "loss": 1.1516,
+      "step": 5479
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7272687107843966e-05,
+      "loss": 1.1231,
+      "step": 5480
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7271617826842998e-05,
+      "loss": 1.2127,
+      "step": 5481
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7270548369381254e-05,
+      "loss": 1.0357,
+      "step": 5482
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7269478735484682e-05,
+      "loss": 1.3117,
+      "step": 5483
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726840892517924e-05,
+      "loss": 1.2084,
+      "step": 5484
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726733893849089e-05,
+      "loss": 1.1718,
+      "step": 5485
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7266268775445595e-05,
+      "loss": 1.0353,
+      "step": 5486
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7265198436069326e-05,
+      "loss": 1.3369,
+      "step": 5487
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7264127920388057e-05,
+      "loss": 1.1067,
+      "step": 5488
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7263057228427764e-05,
+      "loss": 1.1411,
+      "step": 5489
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7261986360214435e-05,
+      "loss": 1.2223,
+      "step": 5490
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7260915315774053e-05,
+      "loss": 1.0413,
+      "step": 5491
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7259844095132612e-05,
+      "loss": 1.2918,
+      "step": 5492
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.72587726983161e-05,
+      "loss": 1.0348,
+      "step": 5493
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7257701125350527e-05,
+      "loss": 1.1336,
+      "step": 5494
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725662937626189e-05,
+      "loss": 1.3573,
+      "step": 5495
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7255557451076202e-05,
+      "loss": 1.0761,
+      "step": 5496
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725448534981947e-05,
+      "loss": 1.2145,
+      "step": 5497
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725341307251771e-05,
+      "loss": 1.1784,
+      "step": 5498
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725234061919695e-05,
+      "loss": 1.2745,
+      "step": 5499
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7251267989883212e-05,
+      "loss": 1.2102,
+      "step": 5500
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7250195184602522e-05,
+      "loss": 1.2185,
+      "step": 5501
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724912220338092e-05,
+      "loss": 1.2777,
+      "step": 5502
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724804904624444e-05,
+      "loss": 1.168,
+      "step": 5503
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7246975713219124e-05,
+      "loss": 1.3296,
+      "step": 5504
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724590220433102e-05,
+      "loss": 1.1411,
+      "step": 5505
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7244828519606183e-05,
+      "loss": 1.2988,
+      "step": 5506
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724375465907066e-05,
+      "loss": 1.1072,
+      "step": 5507
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7242680622750514e-05,
+      "loss": 1.2624,
+      "step": 5508
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7241606410671813e-05,
+      "loss": 0.9889,
+      "step": 5509
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.724053202286062e-05,
+      "loss": 1.1624,
+      "step": 5510
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723945745934301e-05,
+      "loss": 1.2535,
+      "step": 5511
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7238382720145052e-05,
+      "loss": 1.3042,
+      "step": 5512
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7237307805292838e-05,
+      "loss": 1.0807,
+      "step": 5513
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7236232714812443e-05,
+      "loss": 1.1415,
+      "step": 5514
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7235157448729967e-05,
+      "loss": 1.339,
+      "step": 5515
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7234082007071494e-05,
+      "loss": 1.0255,
+      "step": 5516
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7233006389863125e-05,
+      "loss": 1.164,
+      "step": 5517
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7231930597130962e-05,
+      "loss": 1.3163,
+      "step": 5518
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7230854628901115e-05,
+      "loss": 1.2197,
+      "step": 5519
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722977848519969e-05,
+      "loss": 1.1708,
+      "step": 5520
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.72287021660528e-05,
+      "loss": 1.2195,
+      "step": 5521
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722762567148657e-05,
+      "loss": 1.2519,
+      "step": 5522
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722654900152712e-05,
+      "loss": 1.1191,
+      "step": 5523
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722547215620058e-05,
+      "loss": 1.1442,
+      "step": 5524
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722439513553308e-05,
+      "loss": 1.1943,
+      "step": 5525
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7223317939550756e-05,
+      "loss": 1.1752,
+      "step": 5526
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7222240568279746e-05,
+      "loss": 1.0845,
+      "step": 5527
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7221163021746198e-05,
+      "loss": 0.9775,
+      "step": 5528
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722008529997626e-05,
+      "loss": 1.161,
+      "step": 5529
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7219007402996092e-05,
+      "loss": 1.097,
+      "step": 5530
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7217929330831834e-05,
+      "loss": 1.1213,
+      "step": 5531
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7216851083509663e-05,
+      "loss": 1.1325,
+      "step": 5532
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7215772661055743e-05,
+      "loss": 1.1433,
+      "step": 5533
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.721469406349624e-05,
+      "loss": 1.13,
+      "step": 5534
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7213615290857326e-05,
+      "loss": 1.0182,
+      "step": 5535
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7212536343165187e-05,
+      "loss": 1.0406,
+      "step": 5536
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7211457220446e-05,
+      "loss": 1.2908,
+      "step": 5537
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7210377922725957e-05,
+      "loss": 1.2023,
+      "step": 5538
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7209298450031247e-05,
+      "loss": 1.0815,
+      "step": 5539
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7208218802388065e-05,
+      "loss": 1.0907,
+      "step": 5540
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7207138979822608e-05,
+      "loss": 1.2304,
+      "step": 5541
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7206058982361087e-05,
+      "loss": 1.1048,
+      "step": 5542
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7204978810029705e-05,
+      "loss": 1.147,
+      "step": 5543
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7203898462854677e-05,
+      "loss": 1.1768,
+      "step": 5544
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7202817940862216e-05,
+      "loss": 1.1474,
+      "step": 5545
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720173724407855e-05,
+      "loss": 1.2674,
+      "step": 5546
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.72006563725299e-05,
+      "loss": 1.2317,
+      "step": 5547
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7199575326242497e-05,
+      "loss": 1.1435,
+      "step": 5548
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7198494105242575e-05,
+      "loss": 1.1759,
+      "step": 5549
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7197412709556368e-05,
+      "loss": 1.1676,
+      "step": 5550
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7196331139210124e-05,
+      "loss": 1.0947,
+      "step": 5551
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.719524939423008e-05,
+      "loss": 1.1245,
+      "step": 5552
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71941674746425e-05,
+      "loss": 1.1973,
+      "step": 5553
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7193085380473633e-05,
+      "loss": 1.2934,
+      "step": 5554
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7192003111749737e-05,
+      "loss": 1.3072,
+      "step": 5555
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7190920668497074e-05,
+      "loss": 1.0782,
+      "step": 5556
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718983805074192e-05,
+      "loss": 1.1638,
+      "step": 5557
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7188755258510536e-05,
+      "loss": 1.1578,
+      "step": 5558
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7187672291829205e-05,
+      "loss": 1.2088,
+      "step": 5559
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7186589150724203e-05,
+      "loss": 1.0658,
+      "step": 5560
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7185505835221822e-05,
+      "loss": 1.1388,
+      "step": 5561
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7184422345348344e-05,
+      "loss": 1.2481,
+      "step": 5562
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7183338681130065e-05,
+      "loss": 1.1688,
+      "step": 5563
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718225484259328e-05,
+      "loss": 1.1217,
+      "step": 5564
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7181170829764292e-05,
+      "loss": 1.2166,
+      "step": 5565
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7180086642669406e-05,
+      "loss": 1.0313,
+      "step": 5566
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7179002281334934e-05,
+      "loss": 1.1487,
+      "step": 5567
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717791774578719e-05,
+      "loss": 1.2434,
+      "step": 5568
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7176833036052495e-05,
+      "loss": 1.2123,
+      "step": 5569
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7175748152157163e-05,
+      "loss": 1.1817,
+      "step": 5570
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717466309412753e-05,
+      "loss": 1.318,
+      "step": 5571
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7173577861989924e-05,
+      "loss": 1.2097,
+      "step": 5572
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717249245577068e-05,
+      "loss": 1.2814,
+      "step": 5573
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7171406875496134e-05,
+      "loss": 1.2206,
+      "step": 5574
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7170321121192635e-05,
+      "loss": 1.146,
+      "step": 5575
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716923519288653e-05,
+      "loss": 1.1883,
+      "step": 5576
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716814909060417e-05,
+      "loss": 1.2371,
+      "step": 5577
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716706281437192e-05,
+      "loss": 1.1222,
+      "step": 5578
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7165976364216123e-05,
+      "loss": 1.3152,
+      "step": 5579
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7164889740163158e-05,
+      "loss": 1.2344,
+      "step": 5580
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7163802942239388e-05,
+      "loss": 1.2724,
+      "step": 5581
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716271597047119e-05,
+      "loss": 1.0178,
+      "step": 5582
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716162882488494e-05,
+      "loss": 1.2125,
+      "step": 5583
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716054150550702e-05,
+      "loss": 1.29,
+      "step": 5584
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715945401236382e-05,
+      "loss": 1.1774,
+      "step": 5585
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7158366345481722e-05,
+      "loss": 1.0395,
+      "step": 5586
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7157278504887126e-05,
+      "loss": 1.1566,
+      "step": 5587
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715619049060643e-05,
+      "loss": 1.1436,
+      "step": 5588
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7155102302666034e-05,
+      "loss": 1.1125,
+      "step": 5589
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7154013941092352e-05,
+      "loss": 1.2945,
+      "step": 5590
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7152925405911786e-05,
+      "loss": 1.0592,
+      "step": 5591
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7151836697150758e-05,
+      "loss": 1.1364,
+      "step": 5592
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7150747814835687e-05,
+      "loss": 1.0416,
+      "step": 5593
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7149658758992995e-05,
+      "loss": 1.3028,
+      "step": 5594
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714856952964911e-05,
+      "loss": 0.9573,
+      "step": 5595
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7147480126830468e-05,
+      "loss": 1.1236,
+      "step": 5596
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7146390550563506e-05,
+      "loss": 1.2421,
+      "step": 5597
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714530080087466e-05,
+      "loss": 1.2268,
+      "step": 5598
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7144210877790372e-05,
+      "loss": 1.0799,
+      "step": 5599
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7143120781337102e-05,
+      "loss": 1.214,
+      "step": 5600
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7142030511541298e-05,
+      "loss": 1.1209,
+      "step": 5601
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7140940068429413e-05,
+      "loss": 1.1681,
+      "step": 5602
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7139849452027917e-05,
+      "loss": 1.2673,
+      "step": 5603
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713875866236327e-05,
+      "loss": 1.1322,
+      "step": 5604
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713766769946195e-05,
+      "loss": 1.2732,
+      "step": 5605
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713657656335042e-05,
+      "loss": 1.0944,
+      "step": 5606
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7135485254055167e-05,
+      "loss": 1.0583,
+      "step": 5607
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7134393771602672e-05,
+      "loss": 1.3239,
+      "step": 5608
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713330211601942e-05,
+      "loss": 1.1688,
+      "step": 5609
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7132210287331905e-05,
+      "loss": 1.0433,
+      "step": 5610
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7131118285566623e-05,
+      "loss": 1.2178,
+      "step": 5611
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713002611075007e-05,
+      "loss": 1.0768,
+      "step": 5612
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7128933762908754e-05,
+      "loss": 1.1795,
+      "step": 5613
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712784124206918e-05,
+      "loss": 1.1396,
+      "step": 5614
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7126748548257858e-05,
+      "loss": 1.1791,
+      "step": 5615
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712565568150131e-05,
+      "loss": 1.2256,
+      "step": 5616
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7124562641826058e-05,
+      "loss": 1.2944,
+      "step": 5617
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7123469429258617e-05,
+      "loss": 1.1328,
+      "step": 5618
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7122376043825527e-05,
+      "loss": 1.1785,
+      "step": 5619
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7121282485553315e-05,
+      "loss": 1.1356,
+      "step": 5620
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7120188754468522e-05,
+      "loss": 1.2706,
+      "step": 5621
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7119094850597682e-05,
+      "loss": 1.188,
+      "step": 5622
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711800077396735e-05,
+      "loss": 1.2171,
+      "step": 5623
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7116906524604074e-05,
+      "loss": 1.2967,
+      "step": 5624
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7115812102534405e-05,
+      "loss": 1.1701,
+      "step": 5625
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7114717507784906e-05,
+      "loss": 1.1748,
+      "step": 5626
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7113622740382133e-05,
+      "loss": 1.1126,
+      "step": 5627
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711252780035266e-05,
+      "loss": 1.2124,
+      "step": 5628
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711143268772305e-05,
+      "loss": 1.2349,
+      "step": 5629
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7110337402519885e-05,
+      "loss": 1.2789,
+      "step": 5630
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7109241944769742e-05,
+      "loss": 1.309,
+      "step": 5631
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7108146314499205e-05,
+      "loss": 1.0864,
+      "step": 5632
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710705051173486e-05,
+      "loss": 1.2249,
+      "step": 5633
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71059545365033e-05,
+      "loss": 1.0926,
+      "step": 5634
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7104858388831123e-05,
+      "loss": 1.3355,
+      "step": 5635
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7103762068744926e-05,
+      "loss": 1.1845,
+      "step": 5636
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7102665576271316e-05,
+      "loss": 1.2648,
+      "step": 5637
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7101568911436902e-05,
+      "loss": 1.0903,
+      "step": 5638
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710047207426829e-05,
+      "loss": 1.1358,
+      "step": 5639
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7099375064792107e-05,
+      "loss": 1.0878,
+      "step": 5640
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709827788303497e-05,
+      "loss": 1.1441,
+      "step": 5641
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7097180529023502e-05,
+      "loss": 1.1704,
+      "step": 5642
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7096083002784337e-05,
+      "loss": 1.1608,
+      "step": 5643
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7094985304344105e-05,
+      "loss": 1.1415,
+      "step": 5644
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7093887433729445e-05,
+      "loss": 1.0397,
+      "step": 5645
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7092789390967e-05,
+      "loss": 1.2478,
+      "step": 5646
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7091691176083414e-05,
+      "loss": 1.0804,
+      "step": 5647
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7090592789105343e-05,
+      "loss": 1.2116,
+      "step": 5648
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7089494230059434e-05,
+      "loss": 1.2517,
+      "step": 5649
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708839549897235e-05,
+      "loss": 1.009,
+      "step": 5650
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7087296595870752e-05,
+      "loss": 1.266,
+      "step": 5651
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7086197520781312e-05,
+      "loss": 1.102,
+      "step": 5652
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7085098273730696e-05,
+      "loss": 1.2049,
+      "step": 5653
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708399885474558e-05,
+      "loss": 1.2162,
+      "step": 5654
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708289926385265e-05,
+      "loss": 1.0243,
+      "step": 5655
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7081799501078583e-05,
+      "loss": 1.0237,
+      "step": 5656
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7080699566450066e-05,
+      "loss": 1.1964,
+      "step": 5657
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7079599459993797e-05,
+      "loss": 1.1281,
+      "step": 5658
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707849918173647e-05,
+      "loss": 1.1576,
+      "step": 5659
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7077398731704778e-05,
+      "loss": 1.1487,
+      "step": 5660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707629810992544e-05,
+      "loss": 1.0173,
+      "step": 5661
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7075197316425157e-05,
+      "loss": 1.3647,
+      "step": 5662
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7074096351230643e-05,
+      "loss": 1.0766,
+      "step": 5663
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707299521436861e-05,
+      "loss": 1.1272,
+      "step": 5664
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7071893905865788e-05,
+      "loss": 1.0393,
+      "step": 5665
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7070792425748897e-05,
+      "loss": 1.0889,
+      "step": 5666
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706969077404467e-05,
+      "loss": 1.2187,
+      "step": 5667
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7068588950779838e-05,
+      "loss": 1.1808,
+      "step": 5668
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706748695598114e-05,
+      "loss": 1.1412,
+      "step": 5669
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7066384789675318e-05,
+      "loss": 1.0466,
+      "step": 5670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706528245188912e-05,
+      "loss": 1.1931,
+      "step": 5671
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7064179942649293e-05,
+      "loss": 1.2424,
+      "step": 5672
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7063077261982594e-05,
+      "loss": 1.1664,
+      "step": 5673
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7061974409915783e-05,
+      "loss": 1.021,
+      "step": 5674
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7060871386475623e-05,
+      "loss": 1.1857,
+      "step": 5675
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7059768191688877e-05,
+      "loss": 1.262,
+      "step": 5676
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705866482558232e-05,
+      "loss": 1.1335,
+      "step": 5677
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7057561288182727e-05,
+      "loss": 1.141,
+      "step": 5678
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7056457579516874e-05,
+      "loss": 1.1114,
+      "step": 5679
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705535369961155e-05,
+      "loss": 1.2629,
+      "step": 5680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705424964849354e-05,
+      "loss": 1.2085,
+      "step": 5681
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7053145426189643e-05,
+      "loss": 0.9125,
+      "step": 5682
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705204103272664e-05,
+      "loss": 1.3319,
+      "step": 5683
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7050936468131344e-05,
+      "loss": 1.1917,
+      "step": 5684
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704983173243056e-05,
+      "loss": 1.1801,
+      "step": 5685
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7048726825651087e-05,
+      "loss": 1.0957,
+      "step": 5686
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7047621747819744e-05,
+      "loss": 1.1393,
+      "step": 5687
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704651649896335e-05,
+      "loss": 1.3856,
+      "step": 5688
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704541107910872e-05,
+      "loss": 1.2374,
+      "step": 5689
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704430548828269e-05,
+      "loss": 1.1662,
+      "step": 5690
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7043199726512075e-05,
+      "loss": 1.07,
+      "step": 5691
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7042093793823718e-05,
+      "loss": 1.3042,
+      "step": 5692
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7040987690244458e-05,
+      "loss": 1.1511,
+      "step": 5693
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703988141580113e-05,
+      "loss": 1.1491,
+      "step": 5694
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7038774970520585e-05,
+      "loss": 1.101,
+      "step": 5695
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703766835442967e-05,
+      "loss": 1.1693,
+      "step": 5696
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7036561567555245e-05,
+      "loss": 1.1996,
+      "step": 5697
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703545460992416e-05,
+      "loss": 1.1817,
+      "step": 5698
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7034347481563288e-05,
+      "loss": 1.277,
+      "step": 5699
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7033240182499484e-05,
+      "loss": 1.1531,
+      "step": 5700
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7032132712759624e-05,
+      "loss": 1.2208,
+      "step": 5701
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7031025072370592e-05,
+      "loss": 1.2022,
+      "step": 5702
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7029917261359254e-05,
+      "loss": 1.1066,
+      "step": 5703
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7028809279752495e-05,
+      "loss": 0.9131,
+      "step": 5704
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7027701127577208e-05,
+      "loss": 1.1684,
+      "step": 5705
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7026592804860283e-05,
+      "loss": 1.2546,
+      "step": 5706
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7025484311628616e-05,
+      "loss": 1.187,
+      "step": 5707
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7024375647909106e-05,
+      "loss": 1.0033,
+      "step": 5708
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7023266813728656e-05,
+      "loss": 1.1546,
+      "step": 5709
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7022157809114176e-05,
+      "loss": 1.042,
+      "step": 5710
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7021048634092578e-05,
+      "loss": 0.9238,
+      "step": 5711
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7019939288690773e-05,
+      "loss": 1.025,
+      "step": 5712
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.701882977293569e-05,
+      "loss": 1.1574,
+      "step": 5713
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.701772008685425e-05,
+      "loss": 1.2812,
+      "step": 5714
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7016610230473385e-05,
+      "loss": 1.1671,
+      "step": 5715
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.701550020382002e-05,
+      "loss": 1.2835,
+      "step": 5716
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7014390006921096e-05,
+      "loss": 1.0439,
+      "step": 5717
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7013279639803563e-05,
+      "loss": 1.1574,
+      "step": 5718
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7012169102494352e-05,
+      "loss": 1.18,
+      "step": 5719
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7011058395020417e-05,
+      "loss": 1.2332,
+      "step": 5720
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7009947517408717e-05,
+      "loss": 1.2953,
+      "step": 5721
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7008836469686207e-05,
+      "loss": 1.2361,
+      "step": 5722
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7007725251879844e-05,
+      "loss": 1.2511,
+      "step": 5723
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7006613864016605e-05,
+      "loss": 1.2557,
+      "step": 5724
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7005502306123447e-05,
+      "loss": 1.2234,
+      "step": 5725
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7004390578227355e-05,
+      "loss": 1.1427,
+      "step": 5726
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7003278680355298e-05,
+      "loss": 1.0759,
+      "step": 5727
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7002166612534266e-05,
+      "loss": 1.1739,
+      "step": 5728
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7001054374791245e-05,
+      "loss": 1.227,
+      "step": 5729
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699994196715322e-05,
+      "loss": 1.1857,
+      "step": 5730
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699882938964719e-05,
+      "loss": 1.2364,
+      "step": 5731
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6997716642300156e-05,
+      "loss": 1.0662,
+      "step": 5732
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6996603725139117e-05,
+      "loss": 0.9811,
+      "step": 5733
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699549063819108e-05,
+      "loss": 1.098,
+      "step": 5734
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699437738148306e-05,
+      "loss": 1.0707,
+      "step": 5735
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6993263955042073e-05,
+      "loss": 1.1496,
+      "step": 5736
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6992150358895134e-05,
+      "loss": 1.1639,
+      "step": 5737
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699103659306927e-05,
+      "loss": 1.1709,
+      "step": 5738
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6989922657591506e-05,
+      "loss": 1.0808,
+      "step": 5739
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698880855248888e-05,
+      "loss": 1.3563,
+      "step": 5740
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698769427778842e-05,
+      "loss": 1.2005,
+      "step": 5741
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6986579833517168e-05,
+      "loss": 1.2223,
+      "step": 5742
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6985465219702172e-05,
+      "loss": 1.0969,
+      "step": 5743
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698435043637048e-05,
+      "loss": 1.2121,
+      "step": 5744
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6983235483549144e-05,
+      "loss": 1.2385,
+      "step": 5745
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6982120361265216e-05,
+      "loss": 1.1474,
+      "step": 5746
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6981005069545765e-05,
+      "loss": 1.1523,
+      "step": 5747
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6979889608417847e-05,
+      "loss": 1.0783,
+      "step": 5748
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6978773977908537e-05,
+      "loss": 1.0665,
+      "step": 5749
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697765817804491e-05,
+      "loss": 1.2409,
+      "step": 5750
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6976542208854036e-05,
+      "loss": 1.1834,
+      "step": 5751
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6975426070362998e-05,
+      "loss": 1.1117,
+      "step": 5752
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697430976259889e-05,
+      "loss": 1.0351,
+      "step": 5753
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6973193285588786e-05,
+      "loss": 1.0504,
+      "step": 5754
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6972076639359795e-05,
+      "loss": 1.151,
+      "step": 5755
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6970959823939006e-05,
+      "loss": 1.2508,
+      "step": 5756
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696984283935352e-05,
+      "loss": 1.1537,
+      "step": 5757
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696872568563045e-05,
+      "loss": 1.1934,
+      "step": 5758
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69676083627969e-05,
+      "loss": 1.1444,
+      "step": 5759
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6966490870879984e-05,
+      "loss": 1.2438,
+      "step": 5760
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6965373209906827e-05,
+      "loss": 1.128,
+      "step": 5761
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6964255379904546e-05,
+      "loss": 1.2592,
+      "step": 5762
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6963137380900265e-05,
+      "loss": 1.2987,
+      "step": 5763
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6962019212921118e-05,
+      "loss": 1.1568,
+      "step": 5764
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6960900875994243e-05,
+      "loss": 1.2013,
+      "step": 5765
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695978237014677e-05,
+      "loss": 1.1283,
+      "step": 5766
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695866369540585e-05,
+      "loss": 1.0842,
+      "step": 5767
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6957544851798628e-05,
+      "loss": 1.1161,
+      "step": 5768
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695642583935225e-05,
+      "loss": 1.1905,
+      "step": 5769
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6955306658093876e-05,
+      "loss": 1.1739,
+      "step": 5770
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6954187308050666e-05,
+      "loss": 1.2354,
+      "step": 5771
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695306778924978e-05,
+      "loss": 1.0667,
+      "step": 5772
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6951948101718388e-05,
+      "loss": 1.1132,
+      "step": 5773
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695082824548366e-05,
+      "loss": 1.2751,
+      "step": 5774
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694970822057277e-05,
+      "loss": 1.2316,
+      "step": 5775
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6948588027012903e-05,
+      "loss": 0.935,
+      "step": 5776
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694746766483124e-05,
+      "loss": 1.1984,
+      "step": 5777
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6946347134054967e-05,
+      "loss": 1.2166,
+      "step": 5778
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6945226434711277e-05,
+      "loss": 1.0925,
+      "step": 5779
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694410556682737e-05,
+      "loss": 1.2124,
+      "step": 5780
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694298453043044e-05,
+      "loss": 1.2599,
+      "step": 5781
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6941863325547694e-05,
+      "loss": 1.0942,
+      "step": 5782
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6940741952206342e-05,
+      "loss": 1.0919,
+      "step": 5783
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693962041043359e-05,
+      "loss": 1.0609,
+      "step": 5784
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6938498700256664e-05,
+      "loss": 1.0753,
+      "step": 5785
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693737682170278e-05,
+      "loss": 1.1824,
+      "step": 5786
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693625477479916e-05,
+      "loss": 1.2378,
+      "step": 5787
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693513255957304e-05,
+      "loss": 1.2677,
+      "step": 5788
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6934010176051642e-05,
+      "loss": 1.2241,
+      "step": 5789
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6932887624262214e-05,
+      "loss": 1.1943,
+      "step": 5790
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693176490423199e-05,
+      "loss": 1.2405,
+      "step": 5791
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6930642015988216e-05,
+      "loss": 1.1818,
+      "step": 5792
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6929518959558144e-05,
+      "loss": 1.2448,
+      "step": 5793
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6928395734969026e-05,
+      "loss": 0.9973,
+      "step": 5794
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692727234224812e-05,
+      "loss": 1.229,
+      "step": 5795
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6926148781422685e-05,
+      "loss": 1.1909,
+      "step": 5796
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6925025052519993e-05,
+      "loss": 1.1893,
+      "step": 5797
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6923901155567303e-05,
+      "loss": 1.296,
+      "step": 5798
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6922777090591895e-05,
+      "loss": 1.2148,
+      "step": 5799
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692165285762105e-05,
+      "loss": 1.1681,
+      "step": 5800
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6920528456682044e-05,
+      "loss": 1.0538,
+      "step": 5801
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6919403887802165e-05,
+      "loss": 1.2049,
+      "step": 5802
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69182791510087e-05,
+      "loss": 1.2498,
+      "step": 5803
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691715424632895e-05,
+      "loss": 1.1209,
+      "step": 5804
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6916029173790207e-05,
+      "loss": 1.2896,
+      "step": 5805
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6914903933419775e-05,
+      "loss": 1.2396,
+      "step": 5806
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691377852524496e-05,
+      "loss": 1.1271,
+      "step": 5807
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6912652949293077e-05,
+      "loss": 1.1615,
+      "step": 5808
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691152720559143e-05,
+      "loss": 1.1171,
+      "step": 5809
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691040129416735e-05,
+      "loss": 1.2159,
+      "step": 5810
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690927521504815e-05,
+      "loss": 1.1461,
+      "step": 5811
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6908148968261156e-05,
+      "loss": 1.2919,
+      "step": 5812
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690702255383371e-05,
+      "loss": 1.0795,
+      "step": 5813
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6905895971793135e-05,
+      "loss": 1.1095,
+      "step": 5814
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6904769222166777e-05,
+      "loss": 1.0126,
+      "step": 5815
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690364230498197e-05,
+      "loss": 1.2153,
+      "step": 5816
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6902515220266077e-05,
+      "loss": 0.9699,
+      "step": 5817
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6901387968046435e-05,
+      "loss": 1.075,
+      "step": 5818
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6900260548350403e-05,
+      "loss": 1.3017,
+      "step": 5819
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6899132961205338e-05,
+      "loss": 1.1744,
+      "step": 5820
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6898005206638607e-05,
+      "loss": 1.1162,
+      "step": 5821
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689687728467758e-05,
+      "loss": 1.0244,
+      "step": 5822
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689574919534962e-05,
+      "loss": 0.991,
+      "step": 5823
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6894620938682106e-05,
+      "loss": 1.1011,
+      "step": 5824
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689349251470242e-05,
+      "loss": 1.2816,
+      "step": 5825
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689236392343794e-05,
+      "loss": 1.2966,
+      "step": 5826
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6891235164916066e-05,
+      "loss": 1.2299,
+      "step": 5827
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6890106239164173e-05,
+      "loss": 1.196,
+      "step": 5828
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6888977146209668e-05,
+      "loss": 1.0734,
+      "step": 5829
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6887847886079946e-05,
+      "loss": 1.0623,
+      "step": 5830
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6886718458802414e-05,
+      "loss": 1.2033,
+      "step": 5831
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688558886440448e-05,
+      "loss": 1.0926,
+      "step": 5832
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688445910291355e-05,
+      "loss": 1.0669,
+      "step": 5833
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6883329174357044e-05,
+      "loss": 1.2046,
+      "step": 5834
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6882199078762385e-05,
+      "loss": 0.9929,
+      "step": 5835
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6881068816156993e-05,
+      "loss": 1.1096,
+      "step": 5836
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6879938386568295e-05,
+      "loss": 1.1808,
+      "step": 5837
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6878807790023728e-05,
+      "loss": 1.1607,
+      "step": 5838
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6877677026550725e-05,
+      "loss": 1.111,
+      "step": 5839
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6876546096176728e-05,
+      "loss": 1.2319,
+      "step": 5840
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687541499892918e-05,
+      "loss": 1.0785,
+      "step": 5841
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6874283734835528e-05,
+      "loss": 1.1313,
+      "step": 5842
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6873152303923226e-05,
+      "loss": 1.1525,
+      "step": 5843
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6872020706219736e-05,
+      "loss": 1.2323,
+      "step": 5844
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6870888941752507e-05,
+      "loss": 1.211,
+      "step": 5845
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6869757010549014e-05,
+      "loss": 1.0224,
+      "step": 5846
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686862491263672e-05,
+      "loss": 1.1425,
+      "step": 5847
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68674926480431e-05,
+      "loss": 1.2993,
+      "step": 5848
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686636021679563e-05,
+      "loss": 1.3139,
+      "step": 5849
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686522761892179e-05,
+      "loss": 1.2783,
+      "step": 5850
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6864094854449063e-05,
+      "loss": 1.0423,
+      "step": 5851
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6862961923404945e-05,
+      "loss": 1.0799,
+      "step": 5852
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6861828825816925e-05,
+      "loss": 1.062,
+      "step": 5853
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6860695561712493e-05,
+      "loss": 1.109,
+      "step": 5854
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6859562131119162e-05,
+      "loss": 1.0774,
+      "step": 5855
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6858428534064432e-05,
+      "loss": 1.0127,
+      "step": 5856
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685729477057581e-05,
+      "loss": 1.0669,
+      "step": 5857
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6856160840680807e-05,
+      "loss": 1.1478,
+      "step": 5858
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6855026744406945e-05,
+      "loss": 1.2707,
+      "step": 5859
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685389248178175e-05,
+      "loss": 1.0376,
+      "step": 5860
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6852758052832736e-05,
+      "loss": 1.1781,
+      "step": 5861
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6851623457587435e-05,
+      "loss": 1.1672,
+      "step": 5862
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6850488696073385e-05,
+      "loss": 1.2209,
+      "step": 5863
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684935376831812e-05,
+      "loss": 1.2188,
+      "step": 5864
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6848218674349182e-05,
+      "loss": 1.1448,
+      "step": 5865
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684708341419412e-05,
+      "loss": 1.2355,
+      "step": 5866
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684594798788048e-05,
+      "loss": 1.092,
+      "step": 5867
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684481239543581e-05,
+      "loss": 1.2723,
+      "step": 5868
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684367663688768e-05,
+      "loss": 1.309,
+      "step": 5869
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684254071226364e-05,
+      "loss": 1.0495,
+      "step": 5870
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684140462159126e-05,
+      "loss": 1.1191,
+      "step": 5871
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6840268364898112e-05,
+      "loss": 1.0246,
+      "step": 5872
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6839131942211767e-05,
+      "loss": 1.0467,
+      "step": 5873
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6837995353559802e-05,
+      "loss": 1.1658,
+      "step": 5874
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6836858598969797e-05,
+      "loss": 1.1514,
+      "step": 5875
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6835721678469345e-05,
+      "loss": 1.1922,
+      "step": 5876
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6834584592086028e-05,
+      "loss": 1.2033,
+      "step": 5877
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6833447339847442e-05,
+      "loss": 1.3375,
+      "step": 5878
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683230992178119e-05,
+      "loss": 1.1859,
+      "step": 5879
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6831172337914866e-05,
+      "loss": 1.2486,
+      "step": 5880
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6830034588276077e-05,
+      "loss": 1.2282,
+      "step": 5881
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682889667289244e-05,
+      "loss": 1.1912,
+      "step": 5882
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6827758591791563e-05,
+      "loss": 1.2698,
+      "step": 5883
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682662034500106e-05,
+      "loss": 1.1529,
+      "step": 5884
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6825481932548565e-05,
+      "loss": 1.0666,
+      "step": 5885
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682434335446169e-05,
+      "loss": 1.2352,
+      "step": 5886
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6823204610768075e-05,
+      "loss": 1.3577,
+      "step": 5887
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6822065701495347e-05,
+      "loss": 1.1997,
+      "step": 5888
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682092662667115e-05,
+      "loss": 1.1261,
+      "step": 5889
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6819787386323124e-05,
+      "loss": 1.033,
+      "step": 5890
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6818647980478913e-05,
+      "loss": 1.1297,
+      "step": 5891
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681750840916617e-05,
+      "loss": 1.3133,
+      "step": 5892
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6816368672412548e-05,
+      "loss": 1.1427,
+      "step": 5893
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68152287702457e-05,
+      "loss": 1.1126,
+      "step": 5894
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6814088702693294e-05,
+      "loss": 1.1964,
+      "step": 5895
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6812948469782996e-05,
+      "loss": 1.2085,
+      "step": 5896
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6811808071542475e-05,
+      "loss": 1.1588,
+      "step": 5897
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6810667507999406e-05,
+      "loss": 1.2782,
+      "step": 5898
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.680952677918146e-05,
+      "loss": 1.2101,
+      "step": 5899
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6808385885116333e-05,
+      "loss": 1.1812,
+      "step": 5900
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68072448258317e-05,
+      "loss": 1.1731,
+      "step": 5901
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6806103601355256e-05,
+      "loss": 1.1732,
+      "step": 5902
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.680496221171469e-05,
+      "loss": 1.2092,
+      "step": 5903
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6803820656937708e-05,
+      "loss": 1.2488,
+      "step": 5904
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6802678937052005e-05,
+      "loss": 1.2882,
+      "step": 5905
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6801537052085292e-05,
+      "loss": 1.1429,
+      "step": 5906
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.680039500206528e-05,
+      "loss": 1.1915,
+      "step": 5907
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6799252787019676e-05,
+      "loss": 1.1295,
+      "step": 5908
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6798110406976202e-05,
+      "loss": 1.1376,
+      "step": 5909
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6796967861962582e-05,
+      "loss": 1.0534,
+      "step": 5910
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6795825152006545e-05,
+      "loss": 1.1915,
+      "step": 5911
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6794682277135815e-05,
+      "loss": 1.1912,
+      "step": 5912
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679353923737813e-05,
+      "loss": 1.2832,
+      "step": 5913
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679239603276122e-05,
+      "loss": 1.2033,
+      "step": 5914
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679125266331284e-05,
+      "loss": 1.0865,
+      "step": 5915
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679010912906073e-05,
+      "loss": 1.2441,
+      "step": 5916
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6788965430032636e-05,
+      "loss": 1.2756,
+      "step": 5917
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678782156625632e-05,
+      "loss": 1.0419,
+      "step": 5918
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678667753775954e-05,
+      "loss": 1.335,
+      "step": 5919
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6785533344570053e-05,
+      "loss": 1.0316,
+      "step": 5920
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6784388986715625e-05,
+      "loss": 1.2236,
+      "step": 5921
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6783244464224033e-05,
+      "loss": 1.0041,
+      "step": 5922
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678209977712304e-05,
+      "loss": 1.0928,
+      "step": 5923
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6780954925440434e-05,
+      "loss": 1.1154,
+      "step": 5924
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6779809909203996e-05,
+      "loss": 1.2932,
+      "step": 5925
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6778664728441508e-05,
+      "loss": 1.2344,
+      "step": 5926
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6777519383180763e-05,
+      "loss": 1.1405,
+      "step": 5927
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6776373873449554e-05,
+      "loss": 1.1482,
+      "step": 5928
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.677522819927568e-05,
+      "loss": 1.261,
+      "step": 5929
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6774082360686944e-05,
+      "loss": 1.1136,
+      "step": 5930
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.677293635771115e-05,
+      "loss": 1.1643,
+      "step": 5931
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.677179019037611e-05,
+      "loss": 1.2768,
+      "step": 5932
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6770643858709635e-05,
+      "loss": 1.1733,
+      "step": 5933
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6769497362739543e-05,
+      "loss": 1.0626,
+      "step": 5934
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6768350702493663e-05,
+      "loss": 1.1892,
+      "step": 5935
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6767203877999817e-05,
+      "loss": 1.2142,
+      "step": 5936
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6766056889285832e-05,
+      "loss": 1.1318,
+      "step": 5937
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6764909736379545e-05,
+      "loss": 1.2997,
+      "step": 5938
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6763762419308793e-05,
+      "loss": 1.1897,
+      "step": 5939
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676261493810142e-05,
+      "loss": 1.3661,
+      "step": 5940
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6761467292785268e-05,
+      "loss": 1.1399,
+      "step": 5941
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6760319483388193e-05,
+      "loss": 1.2172,
+      "step": 5942
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6759171509938044e-05,
+      "loss": 1.2768,
+      "step": 5943
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675802337246268e-05,
+      "loss": 1.2328,
+      "step": 5944
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6756875070989967e-05,
+      "loss": 1.2029,
+      "step": 5945
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675572660554776e-05,
+      "loss": 1.0733,
+      "step": 5946
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6754577976163944e-05,
+      "loss": 1.2144,
+      "step": 5947
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675342918286638e-05,
+      "loss": 1.25,
+      "step": 5948
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6752280225682954e-05,
+      "loss": 1.2564,
+      "step": 5949
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6751131104641543e-05,
+      "loss": 1.1091,
+      "step": 5950
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6749981819770035e-05,
+      "loss": 1.0809,
+      "step": 5951
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674883237109632e-05,
+      "loss": 1.1309,
+      "step": 5952
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6747682758648288e-05,
+      "loss": 1.2967,
+      "step": 5953
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674653298245384e-05,
+      "loss": 1.2893,
+      "step": 5954
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674538304254088e-05,
+      "loss": 1.2384,
+      "step": 5955
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674423293893731e-05,
+      "loss": 1.2672,
+      "step": 5956
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6743082671671042e-05,
+      "loss": 1.0303,
+      "step": 5957
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6741932240769983e-05,
+      "loss": 1.1651,
+      "step": 5958
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674078164626206e-05,
+      "loss": 1.1207,
+      "step": 5959
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6739630888175188e-05,
+      "loss": 1.0464,
+      "step": 5960
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6738479966537297e-05,
+      "loss": 1.2866,
+      "step": 5961
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6737328881376315e-05,
+      "loss": 1.0872,
+      "step": 5962
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6736177632720175e-05,
+      "loss": 1.1957,
+      "step": 5963
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673502622059681e-05,
+      "loss": 1.1109,
+      "step": 5964
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673387464503417e-05,
+      "loss": 1.1419,
+      "step": 5965
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6732722906060197e-05,
+      "loss": 0.9756,
+      "step": 5966
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6731571003702837e-05,
+      "loss": 1.2906,
+      "step": 5967
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6730418937990046e-05,
+      "loss": 1.221,
+      "step": 5968
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6729266708949778e-05,
+      "loss": 1.2414,
+      "step": 5969
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.672811431661e-05,
+      "loss": 1.1562,
+      "step": 5970
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6726961760998673e-05,
+      "loss": 1.2508,
+      "step": 5971
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.672580904214377e-05,
+      "loss": 1.1998,
+      "step": 5972
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.672465616007326e-05,
+      "loss": 1.1895,
+      "step": 5973
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.672350311481512e-05,
+      "loss": 1.3001,
+      "step": 5974
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6722349906397334e-05,
+      "loss": 1.2145,
+      "step": 5975
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6721196534847886e-05,
+      "loss": 1.2146,
+      "step": 5976
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6720043000194764e-05,
+      "loss": 1.1602,
+      "step": 5977
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671888930246596e-05,
+      "loss": 1.2088,
+      "step": 5978
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671773544168947e-05,
+      "loss": 1.0149,
+      "step": 5979
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.67165814178933e-05,
+      "loss": 1.2018,
+      "step": 5980
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671542723110545e-05,
+      "loss": 1.0826,
+      "step": 5981
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6714272881353933e-05,
+      "loss": 1.1335,
+      "step": 5982
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6713118368666757e-05,
+      "loss": 1.1867,
+      "step": 5983
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6711963693071943e-05,
+      "loss": 1.333,
+      "step": 5984
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671080885459751e-05,
+      "loss": 1.124,
+      "step": 5985
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670965385327148e-05,
+      "loss": 1.0997,
+      "step": 5986
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6708498689121885e-05,
+      "loss": 1.1862,
+      "step": 5987
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6707343362176753e-05,
+      "loss": 1.2115,
+      "step": 5988
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6706187872464123e-05,
+      "loss": 1.2342,
+      "step": 5989
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670503222001204e-05,
+      "loss": 1.1468,
+      "step": 5990
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670387640484854e-05,
+      "loss": 1.264,
+      "step": 5991
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6702720427001675e-05,
+      "loss": 1.0984,
+      "step": 5992
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.67015642864995e-05,
+      "loss": 1.2544,
+      "step": 5993
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6700407983370065e-05,
+      "loss": 1.2583,
+      "step": 5994
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6699251517641435e-05,
+      "loss": 1.2915,
+      "step": 5995
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6698094889341673e-05,
+      "loss": 1.07,
+      "step": 5996
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6696938098498846e-05,
+      "loss": 1.1724,
+      "step": 5997
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6695781145141025e-05,
+      "loss": 1.1171,
+      "step": 5998
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6694624029296288e-05,
+      "loss": 1.2114,
+      "step": 5999
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669346675099272e-05,
+      "loss": 1.1918,
+      "step": 6000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669230931025839e-05,
+      "loss": 1.2421,
+      "step": 6001
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6691151707121397e-05,
+      "loss": 1.0928,
+      "step": 6002
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668999394160983e-05,
+      "loss": 1.1702,
+      "step": 6003
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6688836013751784e-05,
+      "loss": 1.2656,
+      "step": 6004
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668767792357536e-05,
+      "loss": 0.9742,
+      "step": 6005
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6686519671108662e-05,
+      "loss": 1.4004,
+      "step": 6006
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6685361256379795e-05,
+      "loss": 1.0091,
+      "step": 6007
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668420267941687e-05,
+      "loss": 1.1211,
+      "step": 6008
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6683043940248005e-05,
+      "loss": 1.2442,
+      "step": 6009
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668188503890132e-05,
+      "loss": 1.1078,
+      "step": 6010
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6680725975404934e-05,
+      "loss": 1.0478,
+      "step": 6011
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6679566749786974e-05,
+      "loss": 1.1931,
+      "step": 6012
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6678407362075572e-05,
+      "loss": 1.1378,
+      "step": 6013
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6677247812298866e-05,
+      "loss": 1.0413,
+      "step": 6014
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667608810048499e-05,
+      "loss": 1.2094,
+      "step": 6015
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667492822666209e-05,
+      "loss": 1.1915,
+      "step": 6016
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6673768190858317e-05,
+      "loss": 1.2696,
+      "step": 6017
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6672607993101813e-05,
+      "loss": 1.2508,
+      "step": 6018
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6671447633420738e-05,
+      "loss": 1.1601,
+      "step": 6019
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6670287111843247e-05,
+      "loss": 1.2472,
+      "step": 6020
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6669126428397505e-05,
+      "loss": 1.2477,
+      "step": 6021
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666796558311168e-05,
+      "loss": 1.1656,
+      "step": 6022
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6666804576013936e-05,
+      "loss": 1.1817,
+      "step": 6023
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6665643407132454e-05,
+      "loss": 1.2685,
+      "step": 6024
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666448207649541e-05,
+      "loss": 1.278,
+      "step": 6025
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6663320584130988e-05,
+      "loss": 1.1562,
+      "step": 6026
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6662158930067363e-05,
+      "loss": 1.3453,
+      "step": 6027
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666099711433274e-05,
+      "loss": 1.2484,
+      "step": 6028
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6659835136955305e-05,
+      "loss": 1.1882,
+      "step": 6029
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6658672997963258e-05,
+      "loss": 1.0649,
+      "step": 6030
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66575106973848e-05,
+      "loss": 1.2057,
+      "step": 6031
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6656348235248134e-05,
+      "loss": 1.1344,
+      "step": 6032
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6655185611581473e-05,
+      "loss": 1.1438,
+      "step": 6033
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665402282641303e-05,
+      "loss": 1.1488,
+      "step": 6034
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6652859879771024e-05,
+      "loss": 1.1822,
+      "step": 6035
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665169677168367e-05,
+      "loss": 1.2864,
+      "step": 6036
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66505335021792e-05,
+      "loss": 1.1199,
+      "step": 6037
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6649370071285837e-05,
+      "loss": 1.1636,
+      "step": 6038
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664820647903182e-05,
+      "loss": 1.1263,
+      "step": 6039
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6647042725445386e-05,
+      "loss": 1.1053,
+      "step": 6040
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6645878810554768e-05,
+      "loss": 1.2102,
+      "step": 6041
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664471473438822e-05,
+      "loss": 1.1792,
+      "step": 6042
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6643550496973984e-05,
+      "loss": 1.1134,
+      "step": 6043
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664238609834032e-05,
+      "loss": 1.1188,
+      "step": 6044
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6641221538515474e-05,
+      "loss": 1.244,
+      "step": 6045
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6640056817527713e-05,
+      "loss": 1.2966,
+      "step": 6046
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6638891935405307e-05,
+      "loss": 1.1075,
+      "step": 6047
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663772689217651e-05,
+      "loss": 1.1799,
+      "step": 6048
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6636561687869608e-05,
+      "loss": 1.0764,
+      "step": 6049
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663539632251287e-05,
+      "loss": 1.2066,
+      "step": 6050
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6634230796134578e-05,
+      "loss": 1.077,
+      "step": 6051
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663306510876301e-05,
+      "loss": 1.1862,
+      "step": 6052
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6631899260426463e-05,
+      "loss": 1.2826,
+      "step": 6053
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6630733251153224e-05,
+      "loss": 1.2531,
+      "step": 6054
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662956708097159e-05,
+      "loss": 1.2045,
+      "step": 6055
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662840074990986e-05,
+      "loss": 1.1722,
+      "step": 6056
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6627234257996338e-05,
+      "loss": 1.2855,
+      "step": 6057
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662606760525933e-05,
+      "loss": 1.1224,
+      "step": 6058
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6624900791727145e-05,
+      "loss": 1.224,
+      "step": 6059
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6623733817428104e-05,
+      "loss": 0.9978,
+      "step": 6060
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662256668239052e-05,
+      "loss": 1.1844,
+      "step": 6061
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662139938664272e-05,
+      "loss": 0.9502,
+      "step": 6062
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662023193021303e-05,
+      "loss": 1.1106,
+      "step": 6063
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661906431312978e-05,
+      "loss": 1.2617,
+      "step": 6064
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661789653542131e-05,
+      "loss": 1.241,
+      "step": 6065
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6616728597115947e-05,
+      "loss": 1.0583,
+      "step": 6066
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6615560498242043e-05,
+      "loss": 1.0928,
+      "step": 6067
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6614392238827942e-05,
+      "loss": 1.0614,
+      "step": 6068
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6613223818901992e-05,
+      "loss": 1.1202,
+      "step": 6069
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661205523849255e-05,
+      "loss": 0.9794,
+      "step": 6070
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6610886497627973e-05,
+      "loss": 1.2491,
+      "step": 6071
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660971759633662e-05,
+      "loss": 1.1588,
+      "step": 6072
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660854853464686e-05,
+      "loss": 1.0606,
+      "step": 6073
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6607379312587066e-05,
+      "loss": 1.2826,
+      "step": 6074
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6606209930185604e-05,
+      "loss": 1.1422,
+      "step": 6075
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6605040387470855e-05,
+      "loss": 1.1171,
+      "step": 6076
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66038706844712e-05,
+      "loss": 1.222,
+      "step": 6077
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6602700821215027e-05,
+      "loss": 1.0759,
+      "step": 6078
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6601530797730726e-05,
+      "loss": 1.3046,
+      "step": 6079
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660036061404668e-05,
+      "loss": 1.2191,
+      "step": 6080
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6599190270191297e-05,
+      "loss": 0.7939,
+      "step": 6081
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6598019766192975e-05,
+      "loss": 1.2335,
+      "step": 6082
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6596849102080112e-05,
+      "loss": 1.1653,
+      "step": 6083
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6595678277881126e-05,
+      "loss": 1.2579,
+      "step": 6084
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6594507293624426e-05,
+      "loss": 1.1631,
+      "step": 6085
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6593336149338425e-05,
+      "loss": 1.3219,
+      "step": 6086
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6592164845051546e-05,
+      "loss": 1.2196,
+      "step": 6087
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6590993380792215e-05,
+      "loss": 1.159,
+      "step": 6088
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658982175658886e-05,
+      "loss": 1.175,
+      "step": 6089
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6588649972469907e-05,
+      "loss": 1.136,
+      "step": 6090
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6587478028463797e-05,
+      "loss": 1.2519,
+      "step": 6091
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6586305924598967e-05,
+      "loss": 1.0767,
+      "step": 6092
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6585133660903862e-05,
+      "loss": 1.2938,
+      "step": 6093
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6583961237406932e-05,
+      "loss": 1.2817,
+      "step": 6094
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6582788654136623e-05,
+      "loss": 0.9017,
+      "step": 6095
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6581615911121393e-05,
+      "loss": 1.2654,
+      "step": 6096
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.65804430083897e-05,
+      "loss": 1.0051,
+      "step": 6097
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657926994597001e-05,
+      "loss": 1.1782,
+      "step": 6098
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6578096723890788e-05,
+      "loss": 1.1921,
+      "step": 6099
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.65769233421805e-05,
+      "loss": 1.2075,
+      "step": 6100
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6575749800867627e-05,
+      "loss": 1.1134,
+      "step": 6101
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6574576099980645e-05,
+      "loss": 1.0398,
+      "step": 6102
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6573402239548034e-05,
+      "loss": 1.3507,
+      "step": 6103
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6572228219598283e-05,
+      "loss": 1.189,
+      "step": 6104
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6571054040159884e-05,
+      "loss": 1.1524,
+      "step": 6105
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6569879701261323e-05,
+      "loss": 1.1878,
+      "step": 6106
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656870520293111e-05,
+      "loss": 1.196,
+      "step": 6107
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6567530545197733e-05,
+      "loss": 1.2051,
+      "step": 6108
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656635572808971e-05,
+      "loss": 1.2319,
+      "step": 6109
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656518075163554e-05,
+      "loss": 1.188,
+      "step": 6110
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6564005615863744e-05,
+      "loss": 1.2525,
+      "step": 6111
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656283032080283e-05,
+      "loss": 1.0976,
+      "step": 6112
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656165486648133e-05,
+      "loss": 1.3198,
+      "step": 6113
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6560479252927763e-05,
+      "loss": 1.2256,
+      "step": 6114
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655930348017066e-05,
+      "loss": 0.9655,
+      "step": 6115
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655812754823855e-05,
+      "loss": 1.2716,
+      "step": 6116
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655695145715997e-05,
+      "loss": 1.1853,
+      "step": 6117
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655577520696346e-05,
+      "loss": 1.0798,
+      "step": 6118
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6554598797677568e-05,
+      "loss": 1.0949,
+      "step": 6119
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6553422229330837e-05,
+      "loss": 1.139,
+      "step": 6120
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6552245501951823e-05,
+      "loss": 1.0572,
+      "step": 6121
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6551068615569082e-05,
+      "loss": 1.0288,
+      "step": 6122
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6549891570211173e-05,
+      "loss": 1.164,
+      "step": 6123
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6548714365906654e-05,
+      "loss": 1.1268,
+      "step": 6124
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6547537002684097e-05,
+      "loss": 1.1345,
+      "step": 6125
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6546359480572073e-05,
+      "loss": 1.3231,
+      "step": 6126
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6545181799599153e-05,
+      "loss": 1.0973,
+      "step": 6127
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6544003959793925e-05,
+      "loss": 1.22,
+      "step": 6128
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6542825961184967e-05,
+      "loss": 1.1067,
+      "step": 6129
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6541647803800863e-05,
+      "loss": 1.1,
+      "step": 6130
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6540469487670204e-05,
+      "loss": 1.2101,
+      "step": 6131
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6539291012821583e-05,
+      "loss": 1.2091,
+      "step": 6132
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6538112379283603e-05,
+      "loss": 1.1213,
+      "step": 6133
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6536933587084866e-05,
+      "loss": 1.3009,
+      "step": 6134
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.653575463625397e-05,
+      "loss": 1.3182,
+      "step": 6135
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6534575526819535e-05,
+      "loss": 1.1311,
+      "step": 6136
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6533396258810167e-05,
+      "loss": 1.1969,
+      "step": 6137
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6532216832254487e-05,
+      "loss": 1.1292,
+      "step": 6138
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6531037247181113e-05,
+      "loss": 1.1708,
+      "step": 6139
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652985750361867e-05,
+      "loss": 1.2959,
+      "step": 6140
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6528677601595796e-05,
+      "loss": 1.08,
+      "step": 6141
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6527497541141112e-05,
+      "loss": 1.1255,
+      "step": 6142
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652631732228326e-05,
+      "loss": 1.008,
+      "step": 6143
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652513694505088e-05,
+      "loss": 1.1168,
+      "step": 6144
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6523956409472616e-05,
+      "loss": 1.1748,
+      "step": 6145
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6522775715577115e-05,
+      "loss": 1.1518,
+      "step": 6146
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6521594863393028e-05,
+      "loss": 1.1496,
+      "step": 6147
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652041385294902e-05,
+      "loss": 1.0649,
+      "step": 6148
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6519232684273737e-05,
+      "loss": 1.2313,
+      "step": 6149
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6518051357395855e-05,
+      "loss": 1.0689,
+      "step": 6150
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6516869872344033e-05,
+      "loss": 1.0883,
+      "step": 6151
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6515688229146945e-05,
+      "loss": 1.1347,
+      "step": 6152
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6514506427833264e-05,
+      "loss": 1.1472,
+      "step": 6153
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6513324468431672e-05,
+      "loss": 1.2119,
+      "step": 6154
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6512142350970852e-05,
+      "loss": 1.2316,
+      "step": 6155
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6510960075479484e-05,
+      "loss": 1.1671,
+      "step": 6156
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6509777641986264e-05,
+      "loss": 1.2201,
+      "step": 6157
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6508595050519887e-05,
+      "loss": 1.1871,
+      "step": 6158
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6507412301109046e-05,
+      "loss": 1.3153,
+      "step": 6159
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.650622939378245e-05,
+      "loss": 1.0978,
+      "step": 6160
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.65050463285688e-05,
+      "loss": 1.2182,
+      "step": 6161
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6503863105496803e-05,
+      "loss": 1.1027,
+      "step": 6162
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6502679724595185e-05,
+      "loss": 1.2058,
+      "step": 6163
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6501496185892643e-05,
+      "loss": 1.0953,
+      "step": 6164
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6500312489417912e-05,
+      "loss": 1.2106,
+      "step": 6165
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6499128635199717e-05,
+      "loss": 1.156,
+      "step": 6166
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649794462326678e-05,
+      "loss": 1.131,
+      "step": 6167
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6496760453647836e-05,
+      "loss": 1.0921,
+      "step": 6168
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6495576126371626e-05,
+      "loss": 0.8976,
+      "step": 6169
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649439164146688e-05,
+      "loss": 1.1647,
+      "step": 6170
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6493206998962353e-05,
+      "loss": 1.054,
+      "step": 6171
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649202219888679e-05,
+      "loss": 1.1626,
+      "step": 6172
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649083724126894e-05,
+      "loss": 1.2553,
+      "step": 6173
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6489652126137552e-05,
+      "loss": 1.1534,
+      "step": 6174
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64884668535214e-05,
+      "loss": 1.0695,
+      "step": 6175
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648728142344923e-05,
+      "loss": 1.1754,
+      "step": 6176
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648609583594983e-05,
+      "loss": 1.1235,
+      "step": 6177
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6484910091051952e-05,
+      "loss": 1.1768,
+      "step": 6178
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648372418878438e-05,
+      "loss": 1.2767,
+      "step": 6179
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648253812917589e-05,
+      "loss": 1.0764,
+      "step": 6180
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648135191225526e-05,
+      "loss": 1.2639,
+      "step": 6181
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6480165538051282e-05,
+      "loss": 0.933,
+      "step": 6182
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6478979006592747e-05,
+      "loss": 1.1016,
+      "step": 6183
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6477792317908445e-05,
+      "loss": 1.1172,
+      "step": 6184
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6476605472027173e-05,
+      "loss": 1.2301,
+      "step": 6185
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647541846897773e-05,
+      "loss": 1.1697,
+      "step": 6186
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6474231308788928e-05,
+      "loss": 1.2303,
+      "step": 6187
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647304399148957e-05,
+      "loss": 1.156,
+      "step": 6188
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647185651710847e-05,
+      "loss": 1.1887,
+      "step": 6189
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647066888567445e-05,
+      "loss": 1.2372,
+      "step": 6190
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646948109721632e-05,
+      "loss": 1.0586,
+      "step": 6191
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6468293151762914e-05,
+      "loss": 1.0618,
+      "step": 6192
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6467105049343055e-05,
+      "loss": 1.1414,
+      "step": 6193
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6465916789985574e-05,
+      "loss": 1.2489,
+      "step": 6194
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6464728373719307e-05,
+      "loss": 1.2944,
+      "step": 6195
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6463539800573098e-05,
+      "loss": 1.25,
+      "step": 6196
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6462351070575787e-05,
+      "loss": 1.2541,
+      "step": 6197
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646116218375622e-05,
+      "loss": 1.2315,
+      "step": 6198
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6459973140143245e-05,
+      "loss": 1.1562,
+      "step": 6199
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.645878393976572e-05,
+      "loss": 1.1085,
+      "step": 6200
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6457594582652507e-05,
+      "loss": 1.1075,
+      "step": 6201
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6456405068832463e-05,
+      "loss": 1.1894,
+      "step": 6202
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6455215398334457e-05,
+      "loss": 1.0229,
+      "step": 6203
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6454025571187354e-05,
+      "loss": 1.1564,
+      "step": 6204
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6452835587420034e-05,
+      "loss": 1.1887,
+      "step": 6205
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.645164544706137e-05,
+      "loss": 1.1488,
+      "step": 6206
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6450455150140247e-05,
+      "loss": 1.2783,
+      "step": 6207
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6449264696685544e-05,
+      "loss": 1.1384,
+      "step": 6208
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6448074086726157e-05,
+      "loss": 1.357,
+      "step": 6209
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.644688332029097e-05,
+      "loss": 1.1674,
+      "step": 6210
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.644569239740889e-05,
+      "loss": 1.1964,
+      "step": 6211
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.644450131810881e-05,
+      "loss": 1.3264,
+      "step": 6212
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6443310082419636e-05,
+      "loss": 1.1753,
+      "step": 6213
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6442118690370272e-05,
+      "loss": 1.2139,
+      "step": 6214
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6440927141989635e-05,
+      "loss": 1.3078,
+      "step": 6215
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6439735437306638e-05,
+      "loss": 1.1639,
+      "step": 6216
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6438543576350202e-05,
+      "loss": 1.1523,
+      "step": 6217
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6437351559149246e-05,
+      "loss": 1.2238,
+      "step": 6218
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64361593857327e-05,
+      "loss": 1.3566,
+      "step": 6219
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6434967056129496e-05,
+      "loss": 1.1817,
+      "step": 6220
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6433774570368565e-05,
+      "loss": 1.1459,
+      "step": 6221
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6432581928478845e-05,
+      "loss": 1.0752,
+      "step": 6222
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643138913048928e-05,
+      "loss": 1.1969,
+      "step": 6223
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6430196176428818e-05,
+      "loss": 1.3539,
+      "step": 6224
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64290030663264e-05,
+      "loss": 1.1299,
+      "step": 6225
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642780980021099e-05,
+      "loss": 1.2188,
+      "step": 6226
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6426616378111533e-05,
+      "loss": 1.1812,
+      "step": 6227
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6425422800057002e-05,
+      "loss": 1.0613,
+      "step": 6228
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6424229066076353e-05,
+      "loss": 1.244,
+      "step": 6229
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6423035176198558e-05,
+      "loss": 1.1807,
+      "step": 6230
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642184113045259e-05,
+      "loss": 1.1743,
+      "step": 6231
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6420646928867422e-05,
+      "loss": 1.2131,
+      "step": 6232
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6419452571472035e-05,
+      "loss": 1.1676,
+      "step": 6233
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6418258058295413e-05,
+      "loss": 1.1388,
+      "step": 6234
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6417063389366546e-05,
+      "loss": 1.3047,
+      "step": 6235
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6415868564714422e-05,
+      "loss": 1.1883,
+      "step": 6236
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6414673584368034e-05,
+      "loss": 0.931,
+      "step": 6237
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6413478448356387e-05,
+      "loss": 1.2862,
+      "step": 6238
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6412283156708477e-05,
+      "loss": 1.2202,
+      "step": 6239
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6411087709453313e-05,
+      "loss": 1.1622,
+      "step": 6240
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6409892106619905e-05,
+      "loss": 1.2488,
+      "step": 6241
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6408696348237264e-05,
+      "loss": 1.0924,
+      "step": 6242
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640750043433441e-05,
+      "loss": 1.2035,
+      "step": 6243
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6406304364940368e-05,
+      "loss": 1.1518,
+      "step": 6244
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6405108140084157e-05,
+      "loss": 1.3016,
+      "step": 6245
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640391175979481e-05,
+      "loss": 1.1681,
+      "step": 6246
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6402715224101356e-05,
+      "loss": 1.1245,
+      "step": 6247
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640151853303283e-05,
+      "loss": 1.3388,
+      "step": 6248
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640032168661828e-05,
+      "loss": 1.0157,
+      "step": 6249
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6399124684886744e-05,
+      "loss": 1.2263,
+      "step": 6250
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6397927527867272e-05,
+      "loss": 1.126,
+      "step": 6251
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6396730215588913e-05,
+      "loss": 1.0925,
+      "step": 6252
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6395532748080725e-05,
+      "loss": 1.1436,
+      "step": 6253
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6394335125371768e-05,
+      "loss": 1.1677,
+      "step": 6254
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6393137347491098e-05,
+      "loss": 1.0688,
+      "step": 6255
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6391939414467788e-05,
+      "loss": 1.1732,
+      "step": 6256
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639074132633091e-05,
+      "loss": 1.1557,
+      "step": 6257
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6389543083109528e-05,
+      "loss": 1.1284,
+      "step": 6258
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.638834468483273e-05,
+      "loss": 1.2272,
+      "step": 6259
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6387146131529594e-05,
+      "loss": 0.8815,
+      "step": 6260
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63859474232292e-05,
+      "loss": 1.2055,
+      "step": 6261
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6384748559960648e-05,
+      "loss": 1.1491,
+      "step": 6262
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6383549541753024e-05,
+      "loss": 1.2487,
+      "step": 6263
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6382350368635428e-05,
+      "loss": 1.231,
+      "step": 6264
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6381151040636955e-05,
+      "loss": 1.2259,
+      "step": 6265
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6379951557786715e-05,
+      "loss": 1.1376,
+      "step": 6266
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6378751920113814e-05,
+      "loss": 1.2283,
+      "step": 6267
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637755212764736e-05,
+      "loss": 1.2931,
+      "step": 6268
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6376352180416476e-05,
+      "loss": 1.0379,
+      "step": 6269
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637515207845027e-05,
+      "loss": 1.1689,
+      "step": 6270
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637395182177788e-05,
+      "loss": 0.9432,
+      "step": 6271
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6372751410428418e-05,
+      "loss": 1.129,
+      "step": 6272
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6371550844431022e-05,
+      "loss": 1.1143,
+      "step": 6273
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6370350123814826e-05,
+      "loss": 1.1278,
+      "step": 6274
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6369149248608967e-05,
+      "loss": 1.2352,
+      "step": 6275
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6367948218842588e-05,
+      "loss": 1.1533,
+      "step": 6276
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6366747034544828e-05,
+      "loss": 1.1562,
+      "step": 6277
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6365545695744847e-05,
+      "loss": 0.9747,
+      "step": 6278
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6364344202471793e-05,
+      "loss": 1.1933,
+      "step": 6279
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6363142554754817e-05,
+      "loss": 1.1725,
+      "step": 6280
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.636194075262309e-05,
+      "loss": 1.2201,
+      "step": 6281
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6360738796105767e-05,
+      "loss": 1.2198,
+      "step": 6282
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6359536685232022e-05,
+      "loss": 1.2618,
+      "step": 6283
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6358334420031023e-05,
+      "loss": 1.2642,
+      "step": 6284
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6357132000531944e-05,
+      "loss": 1.0789,
+      "step": 6285
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6355929426763974e-05,
+      "loss": 1.186,
+      "step": 6286
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6354726698756285e-05,
+      "loss": 1.1943,
+      "step": 6287
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6353523816538065e-05,
+      "loss": 1.204,
+      "step": 6288
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635232078013851e-05,
+      "loss": 1.216,
+      "step": 6289
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635111758958681e-05,
+      "loss": 1.1559,
+      "step": 6290
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634991424491217e-05,
+      "loss": 1.1472,
+      "step": 6291
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634871074614378e-05,
+      "loss": 1.2392,
+      "step": 6292
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6347507093310852e-05,
+      "loss": 1.0241,
+      "step": 6293
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634630328644259e-05,
+      "loss": 1.2824,
+      "step": 6294
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6345099325568217e-05,
+      "loss": 1.2465,
+      "step": 6295
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634389521071694e-05,
+      "loss": 1.0994,
+      "step": 6296
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6342690941917986e-05,
+      "loss": 1.1784,
+      "step": 6297
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6341486519200572e-05,
+      "loss": 1.063,
+      "step": 6298
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6340281942593932e-05,
+      "loss": 1.1761,
+      "step": 6299
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6339077212127294e-05,
+      "loss": 1.0159,
+      "step": 6300
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6337872327829897e-05,
+      "loss": 1.2709,
+      "step": 6301
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633666728973098e-05,
+      "loss": 1.1972,
+      "step": 6302
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633546209785978e-05,
+      "loss": 1.0632,
+      "step": 6303
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6334256752245546e-05,
+      "loss": 1.0897,
+      "step": 6304
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633305125291753e-05,
+      "loss": 1.0041,
+      "step": 6305
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6331845599904984e-05,
+      "loss": 1.1392,
+      "step": 6306
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633063979323717e-05,
+      "loss": 1.206,
+      "step": 6307
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6329433832943344e-05,
+      "loss": 1.1627,
+      "step": 6308
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632822771905277e-05,
+      "loss": 1.0952,
+      "step": 6309
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6327021451594724e-05,
+      "loss": 1.1753,
+      "step": 6310
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6325815030598472e-05,
+      "loss": 1.1255,
+      "step": 6311
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632460845609329e-05,
+      "loss": 1.1212,
+      "step": 6312
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6323401728108467e-05,
+      "loss": 1.2104,
+      "step": 6313
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6322194846673278e-05,
+      "loss": 1.2257,
+      "step": 6314
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6320987811817013e-05,
+      "loss": 1.2174,
+      "step": 6315
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6319780623568964e-05,
+      "loss": 1.1385,
+      "step": 6316
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6318573281958423e-05,
+      "loss": 1.161,
+      "step": 6317
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631736578701469e-05,
+      "loss": 1.0797,
+      "step": 6318
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631615813876707e-05,
+      "loss": 1.2646,
+      "step": 6319
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6314950337244866e-05,
+      "loss": 1.2461,
+      "step": 6320
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631374238247739e-05,
+      "loss": 1.0911,
+      "step": 6321
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6312534274493955e-05,
+      "loss": 1.1869,
+      "step": 6322
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6311326013323878e-05,
+      "loss": 1.1926,
+      "step": 6323
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6310117598996477e-05,
+      "loss": 1.2093,
+      "step": 6324
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630890903154108e-05,
+      "loss": 1.1809,
+      "step": 6325
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6307700310987018e-05,
+      "loss": 1.0617,
+      "step": 6326
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6306491437363615e-05,
+      "loss": 1.1103,
+      "step": 6327
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6305282410700217e-05,
+      "loss": 1.241,
+      "step": 6328
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6304073231026154e-05,
+      "loss": 1.0096,
+      "step": 6329
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6302863898370772e-05,
+      "loss": 1.2179,
+      "step": 6330
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6301654412763425e-05,
+      "loss": 1.1406,
+      "step": 6331
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630044477423345e-05,
+      "loss": 1.1575,
+      "step": 6332
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6299234982810216e-05,
+      "loss": 1.1087,
+      "step": 6333
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6298025038523073e-05,
+      "loss": 1.2679,
+      "step": 6334
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6296814941401384e-05,
+      "loss": 1.2212,
+      "step": 6335
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6295604691474514e-05,
+      "loss": 1.169,
+      "step": 6336
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6294394288771832e-05,
+      "loss": 1.2111,
+      "step": 6337
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629318373332271e-05,
+      "loss": 1.2384,
+      "step": 6338
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629197302515653e-05,
+      "loss": 1.1187,
+      "step": 6339
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6290762164302666e-05,
+      "loss": 1.1337,
+      "step": 6340
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.628955115079051e-05,
+      "loss": 1.113,
+      "step": 6341
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628833998464944e-05,
+      "loss": 1.1327,
+      "step": 6342
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628712866590885e-05,
+      "loss": 1.1543,
+      "step": 6343
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628591719459814e-05,
+      "loss": 1.2041,
+      "step": 6344
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6284705570746708e-05,
+      "loss": 1.2172,
+      "step": 6345
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6283493794383953e-05,
+      "loss": 1.0732,
+      "step": 6346
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628228186553928e-05,
+      "loss": 1.2539,
+      "step": 6347
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.62810697842421e-05,
+      "loss": 1.1951,
+      "step": 6348
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6279857550521833e-05,
+      "loss": 1.2675,
+      "step": 6349
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6278645164407892e-05,
+      "loss": 1.08,
+      "step": 6350
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6277432625929696e-05,
+      "loss": 1.1724,
+      "step": 6351
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627621993511667e-05,
+      "loss": 1.0708,
+      "step": 6352
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6275007091998246e-05,
+      "loss": 1.1671,
+      "step": 6353
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6273794096603854e-05,
+      "loss": 1.2434,
+      "step": 6354
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6272580948962926e-05,
+      "loss": 1.2217,
+      "step": 6355
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6271367649104906e-05,
+      "loss": 1.1654,
+      "step": 6356
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627015419705924e-05,
+      "loss": 1.2434,
+      "step": 6357
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.626894059285537e-05,
+      "loss": 1.1959,
+      "step": 6358
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.626772683652275e-05,
+      "loss": 1.0433,
+      "step": 6359
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6266512928090833e-05,
+      "loss": 1.1542,
+      "step": 6360
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6265298867589074e-05,
+      "loss": 1.1788,
+      "step": 6361
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6264084655046938e-05,
+      "loss": 1.2113,
+      "step": 6362
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6262870290493887e-05,
+      "loss": 1.2754,
+      "step": 6363
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6261655773959395e-05,
+      "loss": 1.2224,
+      "step": 6364
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6260441105472932e-05,
+      "loss": 1.183,
+      "step": 6365
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6259226285063974e-05,
+      "loss": 1.2785,
+      "step": 6366
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6258011312762002e-05,
+      "loss": 1.1232,
+      "step": 6367
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.62567961885965e-05,
+      "loss": 1.2306,
+      "step": 6368
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6255580912596952e-05,
+      "loss": 1.18,
+      "step": 6369
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.625436548479286e-05,
+      "loss": 1.1136,
+      "step": 6370
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6253149905213702e-05,
+      "loss": 1.2082,
+      "step": 6371
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6251934173888992e-05,
+      "loss": 1.2829,
+      "step": 6372
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6250718290848223e-05,
+      "loss": 0.9413,
+      "step": 6373
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6249502256120906e-05,
+      "loss": 1.1325,
+      "step": 6374
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6248286069736544e-05,
+      "loss": 1.0718,
+      "step": 6375
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624706973172466e-05,
+      "loss": 1.063,
+      "step": 6376
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624585324211476e-05,
+      "loss": 1.1936,
+      "step": 6377
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6244636600936376e-05,
+      "loss": 1.1621,
+      "step": 6378
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6243419808219024e-05,
+      "loss": 1.168,
+      "step": 6379
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6242202863992235e-05,
+      "loss": 1.2632,
+      "step": 6380
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624098576828554e-05,
+      "loss": 1.0853,
+      "step": 6381
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623976852112847e-05,
+      "loss": 1.2616,
+      "step": 6382
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6238551122550573e-05,
+      "loss": 1.1863,
+      "step": 6383
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623733357258139e-05,
+      "loss": 0.8833,
+      "step": 6384
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623611587125046e-05,
+      "loss": 1.1692,
+      "step": 6385
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6234898018587336e-05,
+      "loss": 1.2229,
+      "step": 6386
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6233680014621576e-05,
+      "loss": 1.0625,
+      "step": 6387
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6232461859382736e-05,
+      "loss": 1.2174,
+      "step": 6388
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6231243552900372e-05,
+      "loss": 1.2011,
+      "step": 6389
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6230025095204058e-05,
+      "loss": 1.1109,
+      "step": 6390
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6228806486323353e-05,
+      "loss": 1.3202,
+      "step": 6391
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6227587726287833e-05,
+      "loss": 1.0732,
+      "step": 6392
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6226368815127073e-05,
+      "loss": 1.0574,
+      "step": 6393
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622514975287065e-05,
+      "loss": 1.1994,
+      "step": 6394
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6223930539548157e-05,
+      "loss": 1.013,
+      "step": 6395
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622271117518917e-05,
+      "loss": 1.1087,
+      "step": 6396
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6221491659823283e-05,
+      "loss": 1.2223,
+      "step": 6397
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622027199348009e-05,
+      "loss": 1.1549,
+      "step": 6398
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6219052176189193e-05,
+      "loss": 1.2688,
+      "step": 6399
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6217832207980185e-05,
+      "loss": 1.2221,
+      "step": 6400
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6216612088882677e-05,
+      "loss": 1.1816,
+      "step": 6401
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6215391818926272e-05,
+      "loss": 1.1756,
+      "step": 6402
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6214171398140592e-05,
+      "loss": 1.3382,
+      "step": 6403
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6212950826555243e-05,
+      "loss": 1.1065,
+      "step": 6404
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6211730104199853e-05,
+      "loss": 1.1979,
+      "step": 6405
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6210509231104037e-05,
+      "loss": 1.0739,
+      "step": 6406
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620928820729743e-05,
+      "loss": 1.0593,
+      "step": 6407
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6208067032809656e-05,
+      "loss": 1.1384,
+      "step": 6408
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6206845707670354e-05,
+      "loss": 1.2153,
+      "step": 6409
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6205624231909163e-05,
+      "loss": 1.2196,
+      "step": 6410
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6204402605555718e-05,
+      "loss": 1.1094,
+      "step": 6411
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620318082863967e-05,
+      "loss": 1.1817,
+      "step": 6412
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620195890119067e-05,
+      "loss": 1.1816,
+      "step": 6413
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620073682323836e-05,
+      "loss": 0.9769,
+      "step": 6414
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6199514594812405e-05,
+      "loss": 1.1901,
+      "step": 6415
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6198292215942463e-05,
+      "loss": 1.1927,
+      "step": 6416
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61970696866582e-05,
+      "loss": 1.1601,
+      "step": 6417
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619584700698928e-05,
+      "loss": 1.1807,
+      "step": 6418
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619462417696537e-05,
+      "loss": 1.1764,
+      "step": 6419
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6193401196616153e-05,
+      "loss": 1.3438,
+      "step": 6420
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6192178065971304e-05,
+      "loss": 1.1014,
+      "step": 6421
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6190954785060502e-05,
+      "loss": 1.1703,
+      "step": 6422
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6189731353913435e-05,
+      "loss": 1.2182,
+      "step": 6423
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.618850777255979e-05,
+      "loss": 1.1435,
+      "step": 6424
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6187284041029262e-05,
+      "loss": 1.1857,
+      "step": 6425
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6186060159351547e-05,
+      "loss": 1.1285,
+      "step": 6426
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6184836127556345e-05,
+      "loss": 1.1136,
+      "step": 6427
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6183611945673358e-05,
+      "loss": 1.3251,
+      "step": 6428
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6182387613732295e-05,
+      "loss": 1.0981,
+      "step": 6429
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6181163131762865e-05,
+      "loss": 1.2085,
+      "step": 6430
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6179938499794784e-05,
+      "loss": 1.3037,
+      "step": 6431
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6178713717857774e-05,
+      "loss": 1.2377,
+      "step": 6432
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617748878598155e-05,
+      "loss": 1.0773,
+      "step": 6433
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6176263704195842e-05,
+      "loss": 1.2449,
+      "step": 6434
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6175038472530375e-05,
+      "loss": 1.2109,
+      "step": 6435
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6173813091014886e-05,
+      "loss": 1.2253,
+      "step": 6436
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6172587559679106e-05,
+      "loss": 1.0958,
+      "step": 6437
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617136187855278e-05,
+      "loss": 1.1856,
+      "step": 6438
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6170136047665656e-05,
+      "loss": 1.2335,
+      "step": 6439
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616891006704747e-05,
+      "loss": 0.9118,
+      "step": 6440
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616768393672798e-05,
+      "loss": 1.236,
+      "step": 6441
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6166457656736944e-05,
+      "loss": 1.3073,
+      "step": 6442
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6165231227104113e-05,
+      "loss": 1.1511,
+      "step": 6443
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616400464785925e-05,
+      "loss": 1.23,
+      "step": 6444
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616277791903212e-05,
+      "loss": 1.14,
+      "step": 6445
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61615510406525e-05,
+      "loss": 1.2227,
+      "step": 6446
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6160324012750154e-05,
+      "loss": 1.1491,
+      "step": 6447
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6159096835354858e-05,
+      "loss": 1.3134,
+      "step": 6448
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6157869508496398e-05,
+      "loss": 1.325,
+      "step": 6449
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6156642032204554e-05,
+      "loss": 1.1271,
+      "step": 6450
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6155414406509112e-05,
+      "loss": 0.9734,
+      "step": 6451
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615418663143987e-05,
+      "loss": 1.1956,
+      "step": 6452
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6152958707026615e-05,
+      "loss": 1.3461,
+      "step": 6453
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6151730633299144e-05,
+      "loss": 1.0871,
+      "step": 6454
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6150502410287265e-05,
+      "loss": 1.0611,
+      "step": 6455
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.614927403802078e-05,
+      "loss": 1.14,
+      "step": 6456
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61480455165295e-05,
+      "loss": 1.1453,
+      "step": 6457
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6146816845843235e-05,
+      "loss": 1.1922,
+      "step": 6458
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61455880259918e-05,
+      "loss": 1.1926,
+      "step": 6459
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6144359057005018e-05,
+      "loss": 1.1673,
+      "step": 6460
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6143129938912712e-05,
+      "loss": 1.1283,
+      "step": 6461
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6141900671744712e-05,
+      "loss": 1.2628,
+      "step": 6462
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6140671255530843e-05,
+      "loss": 1.1882,
+      "step": 6463
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.613944169030094e-05,
+      "loss": 1.0312,
+      "step": 6464
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6138211976084845e-05,
+      "loss": 1.1197,
+      "step": 6465
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6136982112912394e-05,
+      "loss": 1.1442,
+      "step": 6466
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.613575210081344e-05,
+      "loss": 1.057,
+      "step": 6467
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6134521939817824e-05,
+      "loss": 1.1882,
+      "step": 6468
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61332916299554e-05,
+      "loss": 1.2966,
+      "step": 6469
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.613206117125603e-05,
+      "loss": 1.0692,
+      "step": 6470
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6130830563749565e-05,
+      "loss": 1.1245,
+      "step": 6471
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6129599807465873e-05,
+      "loss": 1.1979,
+      "step": 6472
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6128368902434823e-05,
+      "loss": 1.1001,
+      "step": 6473
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612713784868628e-05,
+      "loss": 1.1971,
+      "step": 6474
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6125906646250124e-05,
+      "loss": 1.1693,
+      "step": 6475
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6124675295156227e-05,
+      "loss": 1.0977,
+      "step": 6476
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612344379543447e-05,
+      "loss": 1.1392,
+      "step": 6477
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6122212147114745e-05,
+      "loss": 1.2624,
+      "step": 6478
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6120980350226933e-05,
+      "loss": 1.2141,
+      "step": 6479
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6119748404800927e-05,
+      "loss": 1.213,
+      "step": 6480
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.611851631086663e-05,
+      "loss": 1.1249,
+      "step": 6481
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6117284068453932e-05,
+      "loss": 0.8997,
+      "step": 6482
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6116051677592743e-05,
+      "loss": 1.1896,
+      "step": 6483
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6114819138312964e-05,
+      "loss": 1.1557,
+      "step": 6484
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.611358645064451e-05,
+      "loss": 1.1734,
+      "step": 6485
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.611235361461729e-05,
+      "loss": 1.185,
+      "step": 6486
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6111120630261224e-05,
+      "loss": 1.2315,
+      "step": 6487
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6109887497606233e-05,
+      "loss": 1.0911,
+      "step": 6488
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6108654216682244e-05,
+      "loss": 1.0576,
+      "step": 6489
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6107420787519178e-05,
+      "loss": 1.2163,
+      "step": 6490
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610618721014697e-05,
+      "loss": 1.2225,
+      "step": 6491
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610495348459556e-05,
+      "loss": 1.0488,
+      "step": 6492
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6103719610894878e-05,
+      "loss": 1.1815,
+      "step": 6493
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6102485589074877e-05,
+      "loss": 1.083,
+      "step": 6494
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6101251419165493e-05,
+      "loss": 1.1267,
+      "step": 6495
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6100017101196685e-05,
+      "loss": 1.042,
+      "step": 6496
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6098782635198396e-05,
+      "loss": 1.1938,
+      "step": 6497
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6097548021200592e-05,
+      "loss": 1.2162,
+      "step": 6498
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6096313259233227e-05,
+      "loss": 1.2376,
+      "step": 6499
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6095078349326268e-05,
+      "loss": 1.2179,
+      "step": 6500
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6093843291509685e-05,
+      "loss": 1.0232,
+      "step": 6501
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6092608085813448e-05,
+      "loss": 1.2524,
+      "step": 6502
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6091372732267525e-05,
+      "loss": 1.1525,
+      "step": 6503
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6090137230901905e-05,
+      "loss": 1.08,
+      "step": 6504
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6088901581746562e-05,
+      "loss": 1.0492,
+      "step": 6505
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6087665784831482e-05,
+      "loss": 1.2354,
+      "step": 6506
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6086429840186663e-05,
+      "loss": 1.227,
+      "step": 6507
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6085193747842086e-05,
+      "loss": 1.1989,
+      "step": 6508
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6083957507827758e-05,
+      "loss": 1.244,
+      "step": 6509
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608272112017367e-05,
+      "loss": 1.1515,
+      "step": 6510
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6081484584909828e-05,
+      "loss": 1.2523,
+      "step": 6511
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608024790206624e-05,
+      "loss": 1.0884,
+      "step": 6512
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6079011071672923e-05,
+      "loss": 1.1943,
+      "step": 6513
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6077774093759875e-05,
+      "loss": 1.2613,
+      "step": 6514
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607653696835713e-05,
+      "loss": 1.2185,
+      "step": 6515
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6075299695494703e-05,
+      "loss": 1.2814,
+      "step": 6516
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607406227520262e-05,
+      "loss": 1.1018,
+      "step": 6517
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6072824707510905e-05,
+      "loss": 1.349,
+      "step": 6518
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6071586992449597e-05,
+      "loss": 1.107,
+      "step": 6519
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6070349130048724e-05,
+      "loss": 1.2002,
+      "step": 6520
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6069111120338335e-05,
+      "loss": 1.2316,
+      "step": 6521
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6067872963348466e-05,
+      "loss": 1.0891,
+      "step": 6522
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6066634659109167e-05,
+      "loss": 1.1649,
+      "step": 6523
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6065396207650482e-05,
+      "loss": 1.2082,
+      "step": 6524
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6064157609002473e-05,
+      "loss": 1.0771,
+      "step": 6525
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606291886319519e-05,
+      "loss": 1.1031,
+      "step": 6526
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.60616799702587e-05,
+      "loss": 1.1584,
+      "step": 6527
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606044093022306e-05,
+      "loss": 1.1511,
+      "step": 6528
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6059201743118342e-05,
+      "loss": 1.1995,
+      "step": 6529
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.605796240897462e-05,
+      "loss": 1.3751,
+      "step": 6530
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6056722927821966e-05,
+      "loss": 1.136,
+      "step": 6531
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6055483299690458e-05,
+      "loss": 1.1968,
+      "step": 6532
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6054243524610175e-05,
+      "loss": 1.1983,
+      "step": 6533
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.605300360261121e-05,
+      "loss": 1.1054,
+      "step": 6534
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.605176353372365e-05,
+      "loss": 1.1579,
+      "step": 6535
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6050523317977586e-05,
+      "loss": 1.23,
+      "step": 6536
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6049282955403115e-05,
+      "loss": 1.1588,
+      "step": 6537
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6048042446030336e-05,
+      "loss": 1.1312,
+      "step": 6538
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6046801789889353e-05,
+      "loss": 0.9127,
+      "step": 6539
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6045560987010276e-05,
+      "loss": 1.1506,
+      "step": 6540
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.604432003742321e-05,
+      "loss": 1.1274,
+      "step": 6541
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6043078941158273e-05,
+      "loss": 1.2662,
+      "step": 6542
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6041837698245583e-05,
+      "loss": 1.2942,
+      "step": 6543
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.604059630871526e-05,
+      "loss": 1.2472,
+      "step": 6544
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6039354772597432e-05,
+      "loss": 1.3075,
+      "step": 6545
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6038113089922224e-05,
+      "loss": 1.1854,
+      "step": 6546
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6036871260719766e-05,
+      "loss": 1.202,
+      "step": 6547
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.60356292850202e-05,
+      "loss": 1.0738,
+      "step": 6548
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6034387162853656e-05,
+      "loss": 1.1218,
+      "step": 6549
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.603314489425029e-05,
+      "loss": 1.0575,
+      "step": 6550
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.603190247924023e-05,
+      "loss": 1.3158,
+      "step": 6551
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6030659917853643e-05,
+      "loss": 0.9928,
+      "step": 6552
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6029417210120674e-05,
+      "loss": 1.1738,
+      "step": 6553
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602817435607148e-05,
+      "loss": 1.2519,
+      "step": 6554
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6026931355736225e-05,
+      "loss": 1.061,
+      "step": 6555
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6025688209145068e-05,
+      "loss": 1.0981,
+      "step": 6556
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602444491632818e-05,
+      "loss": 1.2606,
+      "step": 6557
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602320147731573e-05,
+      "loss": 1.2034,
+      "step": 6558
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6021957892137895e-05,
+      "loss": 1.1098,
+      "step": 6559
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6020714160824854e-05,
+      "loss": 1.1782,
+      "step": 6560
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6019470283406785e-05,
+      "loss": 1.271,
+      "step": 6561
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6018226259913872e-05,
+      "loss": 1.3575,
+      "step": 6562
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6016982090376308e-05,
+      "loss": 1.119,
+      "step": 6563
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6015737774824287e-05,
+      "loss": 1.0033,
+      "step": 6564
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6014493313288e-05,
+      "loss": 1.2978,
+      "step": 6565
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601324870579765e-05,
+      "loss": 1.1691,
+      "step": 6566
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6012003952383436e-05,
+      "loss": 1.0764,
+      "step": 6567
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6010759053075563e-05,
+      "loss": 1.0092,
+      "step": 6568
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6009514007904252e-05,
+      "loss": 1.1904,
+      "step": 6569
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6008268816899706e-05,
+      "loss": 1.2752,
+      "step": 6570
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6007023480092146e-05,
+      "loss": 1.1546,
+      "step": 6571
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600577799751179e-05,
+      "loss": 1.2178,
+      "step": 6572
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6004532369188868e-05,
+      "loss": 1.1438,
+      "step": 6573
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6003286595153604e-05,
+      "loss": 1.0814,
+      "step": 6574
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6002040675436225e-05,
+      "loss": 1.0361,
+      "step": 6575
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6000794610066974e-05,
+      "loss": 1.2326,
+      "step": 6576
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5999548399076083e-05,
+      "loss": 1.2333,
+      "step": 6577
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59983020424938e-05,
+      "loss": 1.1468,
+      "step": 6578
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5997055540350364e-05,
+      "loss": 1.1517,
+      "step": 6579
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5995808892676028e-05,
+      "loss": 1.1443,
+      "step": 6580
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5994562099501042e-05,
+      "loss": 1.1912,
+      "step": 6581
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5993315160855662e-05,
+      "loss": 1.1972,
+      "step": 6582
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599206807677015e-05,
+      "loss": 1.149,
+      "step": 6583
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599082084727477e-05,
+      "loss": 1.2517,
+      "step": 6584
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5989573472399785e-05,
+      "loss": 1.1147,
+      "step": 6585
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5988325952175468e-05,
+      "loss": 1.2128,
+      "step": 6586
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5987078286632088e-05,
+      "loss": 1.1909,
+      "step": 6587
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5985830475799925e-05,
+      "loss": 1.1752,
+      "step": 6588
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5984582519709266e-05,
+      "loss": 1.0906,
+      "step": 6589
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5983334418390382e-05,
+      "loss": 1.0458,
+      "step": 6590
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5982086171873575e-05,
+      "loss": 1.2694,
+      "step": 6591
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5980837780189124e-05,
+      "loss": 1.2886,
+      "step": 6592
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597958924336733e-05,
+      "loss": 1.2561,
+      "step": 6593
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597834056143849e-05,
+      "loss": 1.1426,
+      "step": 6594
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597709173443291e-05,
+      "loss": 1.1657,
+      "step": 6595
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597584276238089e-05,
+      "loss": 1.2099,
+      "step": 6596
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5974593645312737e-05,
+      "loss": 1.0871,
+      "step": 6597
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597334438325877e-05,
+      "loss": 1.2544,
+      "step": 6598
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59720949762493e-05,
+      "loss": 1.1975,
+      "step": 6599
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597084542431465e-05,
+      "loss": 1.2423,
+      "step": 6600
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596959572748514e-05,
+      "loss": 1.1967,
+      "step": 6601
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59683458857911e-05,
+      "loss": 1.1853,
+      "step": 6602
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5967095899262853e-05,
+      "loss": 1.167,
+      "step": 6603
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5965845767930738e-05,
+      "loss": 1.2562,
+      "step": 6604
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5964595491825093e-05,
+      "loss": 1.0721,
+      "step": 6605
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596334507097626e-05,
+      "loss": 1.1931,
+      "step": 6606
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5962094505414573e-05,
+      "loss": 1.1596,
+      "step": 6607
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596084379517039e-05,
+      "loss": 1.2097,
+      "step": 6608
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5959592940274057e-05,
+      "loss": 1.1193,
+      "step": 6609
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595834194075593e-05,
+      "loss": 1.1373,
+      "step": 6610
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5957090796646372e-05,
+      "loss": 1.2092,
+      "step": 6611
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5955839507975732e-05,
+      "loss": 1.215,
+      "step": 6612
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5954588074774387e-05,
+      "loss": 1.3321,
+      "step": 6613
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59533364970727e-05,
+      "loss": 1.2635,
+      "step": 6614
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5952084774901043e-05,
+      "loss": 1.0813,
+      "step": 6615
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5950832908289795e-05,
+      "loss": 1.1362,
+      "step": 6616
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5949580897269332e-05,
+      "loss": 1.2083,
+      "step": 6617
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594832874187004e-05,
+      "loss": 1.0415,
+      "step": 6618
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59470764421223e-05,
+      "loss": 1.1122,
+      "step": 6619
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5945823998056505e-05,
+      "loss": 1.1557,
+      "step": 6620
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594457140970305e-05,
+      "loss": 1.184,
+      "step": 6621
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5943318677092327e-05,
+      "loss": 1.1849,
+      "step": 6622
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5942065800254738e-05,
+      "loss": 1.0781,
+      "step": 6623
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5940812779220687e-05,
+      "loss": 1.1195,
+      "step": 6624
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5939559614020584e-05,
+      "loss": 1.2646,
+      "step": 6625
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5938306304684835e-05,
+      "loss": 1.1791,
+      "step": 6626
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5937052851243855e-05,
+      "loss": 1.1643,
+      "step": 6627
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.593579925372806e-05,
+      "loss": 1.1296,
+      "step": 6628
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5934545512167878e-05,
+      "loss": 1.1236,
+      "step": 6629
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.593329162659373e-05,
+      "loss": 1.1254,
+      "step": 6630
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5932037597036042e-05,
+      "loss": 1.3054,
+      "step": 6631
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5930783423525246e-05,
+      "loss": 1.2362,
+      "step": 6632
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5929529106091778e-05,
+      "loss": 1.2262,
+      "step": 6633
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5928274644766077e-05,
+      "loss": 1.2462,
+      "step": 6634
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5927020039578587e-05,
+      "loss": 1.092,
+      "step": 6635
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592576529055975e-05,
+      "loss": 1.24,
+      "step": 6636
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5924510397740015e-05,
+      "loss": 1.0905,
+      "step": 6637
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5923255361149835e-05,
+      "loss": 1.0452,
+      "step": 6638
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592200018081967e-05,
+      "loss": 1.153,
+      "step": 6639
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5920744856779974e-05,
+      "loss": 1.2959,
+      "step": 6640
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5919489389061215e-05,
+      "loss": 1.2167,
+      "step": 6641
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5918233777693858e-05,
+      "loss": 1.0503,
+      "step": 6642
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591697802270837e-05,
+      "loss": 1.185,
+      "step": 6643
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5915722124135227e-05,
+      "loss": 1.3069,
+      "step": 6644
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5914466082004904e-05,
+      "loss": 1.0909,
+      "step": 6645
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5913209896347886e-05,
+      "loss": 1.1889,
+      "step": 6646
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591195356719465e-05,
+      "loss": 1.0251,
+      "step": 6647
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5910697094575693e-05,
+      "loss": 1.1149,
+      "step": 6648
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5909440478521495e-05,
+      "loss": 1.2604,
+      "step": 6649
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5908183719062557e-05,
+      "loss": 1.0438,
+      "step": 6650
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5906926816229374e-05,
+      "loss": 1.1275,
+      "step": 6651
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5905669770052453e-05,
+      "loss": 1.1993,
+      "step": 6652
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5904412580562294e-05,
+      "loss": 1.1498,
+      "step": 6653
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5903155247789403e-05,
+      "loss": 1.1312,
+      "step": 6654
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5901897771764298e-05,
+      "loss": 1.1166,
+      "step": 6655
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.590064015251749e-05,
+      "loss": 1.0674,
+      "step": 6656
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5899382390079495e-05,
+      "loss": 1.23,
+      "step": 6657
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5898124484480845e-05,
+      "loss": 1.2057,
+      "step": 6658
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5896866435752055e-05,
+      "loss": 1.093,
+      "step": 6659
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5895608243923662e-05,
+      "loss": 1.109,
+      "step": 6660
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589434990902619e-05,
+      "loss": 1.1277,
+      "step": 6661
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5893091431090186e-05,
+      "loss": 0.9772,
+      "step": 6662
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589183281014618e-05,
+      "loss": 0.9908,
+      "step": 6663
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589057404622472e-05,
+      "loss": 1.1496,
+      "step": 6664
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5889315139356352e-05,
+      "loss": 1.0761,
+      "step": 6665
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5888056089571628e-05,
+      "loss": 1.0997,
+      "step": 6666
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5886796896901095e-05,
+      "loss": 1.0644,
+      "step": 6667
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5885537561375314e-05,
+      "loss": 1.2007,
+      "step": 6668
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5884278083024848e-05,
+      "loss": 1.1601,
+      "step": 6669
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5883018461880255e-05,
+      "loss": 1.2349,
+      "step": 6670
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5881758697972106e-05,
+      "loss": 1.2524,
+      "step": 6671
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5880498791330972e-05,
+      "loss": 1.0671,
+      "step": 6672
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5879238741987424e-05,
+      "loss": 1.2219,
+      "step": 6673
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5877978549972044e-05,
+      "loss": 1.1229,
+      "step": 6674
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587671821531541e-05,
+      "loss": 1.172,
+      "step": 6675
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587545773804811e-05,
+      "loss": 1.1203,
+      "step": 6676
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587419711820073e-05,
+      "loss": 1.2193,
+      "step": 6677
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5872936355803857e-05,
+      "loss": 1.207,
+      "step": 6678
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5871675450888093e-05,
+      "loss": 1.1365,
+      "step": 6679
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5870414403484034e-05,
+      "loss": 1.294,
+      "step": 6680
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586915321362228e-05,
+      "loss": 1.08,
+      "step": 6681
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586789188133344e-05,
+      "loss": 1.1729,
+      "step": 6682
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586663040664812e-05,
+      "loss": 0.9362,
+      "step": 6683
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5865368789596932e-05,
+      "loss": 1.0084,
+      "step": 6684
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.58641070302105e-05,
+      "loss": 1.2582,
+      "step": 6685
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5862845128519428e-05,
+      "loss": 1.2727,
+      "step": 6686
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586158308455435e-05,
+      "loss": 1.2227,
+      "step": 6687
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5860320898345887e-05,
+      "loss": 1.0627,
+      "step": 6688
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585905856992467e-05,
+      "loss": 1.0444,
+      "step": 6689
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5857796099321333e-05,
+      "loss": 1.1385,
+      "step": 6690
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5856533486566513e-05,
+      "loss": 1.2323,
+      "step": 6691
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585527073169085e-05,
+      "loss": 1.1103,
+      "step": 6692
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5854007834724986e-05,
+      "loss": 1.3014,
+      "step": 6693
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5852744795699564e-05,
+      "loss": 1.0321,
+      "step": 6694
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585148161464524e-05,
+      "loss": 1.1503,
+      "step": 6695
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5850218291592666e-05,
+      "loss": 1.1958,
+      "step": 6696
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5848954826572498e-05,
+      "loss": 1.1311,
+      "step": 6697
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.58476912196154e-05,
+      "loss": 0.9247,
+      "step": 6698
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5846427470752033e-05,
+      "loss": 1.0133,
+      "step": 6699
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5845163580013062e-05,
+      "loss": 1.0129,
+      "step": 6700
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5843899547429164e-05,
+      "loss": 1.1379,
+      "step": 6701
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584263537303101e-05,
+      "loss": 1.3604,
+      "step": 6702
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5841371056849278e-05,
+      "loss": 1.0385,
+      "step": 6703
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584010659891465e-05,
+      "loss": 1.2083,
+      "step": 6704
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.583884199925781e-05,
+      "loss": 1.2301,
+      "step": 6705
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5837577257909443e-05,
+      "loss": 1.2563,
+      "step": 6706
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5836312374900244e-05,
+      "loss": 1.268,
+      "step": 6707
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.583504735026091e-05,
+      "loss": 0.9973,
+      "step": 6708
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5833782184022134e-05,
+      "loss": 1.2278,
+      "step": 6709
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5832516876214627e-05,
+      "loss": 1.1659,
+      "step": 6710
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5831251426869082e-05,
+      "loss": 1.2001,
+      "step": 6711
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5829985836016218e-05,
+      "loss": 1.095,
+      "step": 6712
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.582872010368674e-05,
+      "loss": 1.0519,
+      "step": 6713
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.582745422991137e-05,
+      "loss": 1.1185,
+      "step": 6714
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.582618821472082e-05,
+      "loss": 1.2823,
+      "step": 6715
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.582492205814582e-05,
+      "loss": 1.0426,
+      "step": 6716
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.582365576021709e-05,
+      "loss": 1.1895,
+      "step": 6717
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5822389320965364e-05,
+      "loss": 1.2385,
+      "step": 6718
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5821122740421366e-05,
+      "loss": 1.1458,
+      "step": 6719
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5819856018615843e-05,
+      "loss": 1.1926,
+      "step": 6720
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581858915557953e-05,
+      "loss": 1.3001,
+      "step": 6721
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581732215134317e-05,
+      "loss": 1.1793,
+      "step": 6722
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581605500593751e-05,
+      "loss": 1.1835,
+      "step": 6723
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5814787719393296e-05,
+      "loss": 1.1671,
+      "step": 6724
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581352029174129e-05,
+      "loss": 1.2587,
+      "step": 6725
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581225272301224e-05,
+      "loss": 0.9749,
+      "step": 6726
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581098501323691e-05,
+      "loss": 1.3344,
+      "step": 6727
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5809717162446064e-05,
+      "loss": 1.1379,
+      "step": 6728
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5808449170670464e-05,
+      "loss": 0.9983,
+      "step": 6729
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580718103794089e-05,
+      "loss": 1.0551,
+      "step": 6730
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5805912764288108e-05,
+      "loss": 1.0939,
+      "step": 6731
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.58046443497429e-05,
+      "loss": 1.0898,
+      "step": 6732
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5803375794336043e-05,
+      "loss": 1.2573,
+      "step": 6733
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580210709809832e-05,
+      "loss": 1.3474,
+      "step": 6734
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5800838261060528e-05,
+      "loss": 1.207,
+      "step": 6735
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5799569283253444e-05,
+      "loss": 1.1039,
+      "step": 6736
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5798300164707875e-05,
+      "loss": 1.223,
+      "step": 6737
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579703090545461e-05,
+      "loss": 1.27,
+      "step": 6738
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579576150552445e-05,
+      "loss": 1.1638,
+      "step": 6739
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579449196494821e-05,
+      "loss": 1.0497,
+      "step": 6740
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5793222283756688e-05,
+      "loss": 1.1241,
+      "step": 6741
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5791952461980696e-05,
+      "loss": 1.0945,
+      "step": 6742
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5790682499651056e-05,
+      "loss": 1.1904,
+      "step": 6743
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578941239679858e-05,
+      "loss": 1.2127,
+      "step": 6744
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5788142153454086e-05,
+      "loss": 1.2352,
+      "step": 6745
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578687176964841e-05,
+      "loss": 1.21,
+      "step": 6746
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5785601245412373e-05,
+      "loss": 1.0521,
+      "step": 6747
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578433058077681e-05,
+      "loss": 1.2441,
+      "step": 6748
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5783059775772553e-05,
+      "loss": 0.9587,
+      "step": 6749
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5781788830430443e-05,
+      "loss": 1.19,
+      "step": 6750
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5780517744781324e-05,
+      "loss": 1.1964,
+      "step": 6751
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5779246518856038e-05,
+      "loss": 1.3425,
+      "step": 6752
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5777975152685433e-05,
+      "loss": 1.3167,
+      "step": 6753
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5776703646300365e-05,
+      "loss": 1.2098,
+      "step": 6754
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5775431999731686e-05,
+      "loss": 1.0829,
+      "step": 6755
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.577416021301026e-05,
+      "loss": 1.2625,
+      "step": 6756
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5772888286166947e-05,
+      "loss": 1.1354,
+      "step": 6757
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5771616219232612e-05,
+      "loss": 1.1699,
+      "step": 6758
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5770344012238123e-05,
+      "loss": 1.032,
+      "step": 6759
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5769071665214356e-05,
+      "loss": 1.2336,
+      "step": 6760
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576779917819219e-05,
+      "loss": 1.0796,
+      "step": 6761
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5766526551202496e-05,
+      "loss": 1.171,
+      "step": 6762
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5765253784276162e-05,
+      "loss": 1.199,
+      "step": 6763
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5763980877444075e-05,
+      "loss": 1.2056,
+      "step": 6764
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5762707830737124e-05,
+      "loss": 1.1527,
+      "step": 6765
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.57614346441862e-05,
+      "loss": 1.1922,
+      "step": 6766
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.57601613178222e-05,
+      "loss": 1.1835,
+      "step": 6767
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5758887851676026e-05,
+      "loss": 1.1887,
+      "step": 6768
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5757614245778578e-05,
+      "loss": 1.1169,
+      "step": 6769
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5756340500160768e-05,
+      "loss": 1.0364,
+      "step": 6770
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.57550666148535e-05,
+      "loss": 1.1727,
+      "step": 6771
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575379258988769e-05,
+      "loss": 1.1901,
+      "step": 6772
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5752518425294258e-05,
+      "loss": 1.2927,
+      "step": 6773
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5751244121104122e-05,
+      "loss": 1.2089,
+      "step": 6774
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.57499696773482e-05,
+      "loss": 1.1204,
+      "step": 6775
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5748695094057432e-05,
+      "loss": 1.2372,
+      "step": 6776
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5747420371262732e-05,
+      "loss": 1.1045,
+      "step": 6777
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574614550899505e-05,
+      "loss": 1.1465,
+      "step": 6778
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5744870507285308e-05,
+      "loss": 0.9777,
+      "step": 6779
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5743595366164458e-05,
+      "loss": 0.9475,
+      "step": 6780
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574232008566344e-05,
+      "loss": 1.1416,
+      "step": 6781
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.57410446658132e-05,
+      "loss": 1.1167,
+      "step": 6782
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573976910664469e-05,
+      "loss": 1.1652,
+      "step": 6783
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5738493408188863e-05,
+      "loss": 1.1019,
+      "step": 6784
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573721757047668e-05,
+      "loss": 1.1636,
+      "step": 6785
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573594159353909e-05,
+      "loss": 1.0376,
+      "step": 6786
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5734665477407076e-05,
+      "loss": 1.0039,
+      "step": 6787
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5733389222111592e-05,
+      "loss": 1.3462,
+      "step": 6788
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5732112827683615e-05,
+      "loss": 1.1952,
+      "step": 6789
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5730836294154116e-05,
+      "loss": 1.1083,
+      "step": 6790
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5729559621554075e-05,
+      "loss": 1.1259,
+      "step": 6791
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.572828280991447e-05,
+      "loss": 1.1371,
+      "step": 6792
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.572700585926629e-05,
+      "loss": 1.1788,
+      "step": 6793
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5725728769640517e-05,
+      "loss": 1.0206,
+      "step": 6794
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5724451541068147e-05,
+      "loss": 1.1653,
+      "step": 6795
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5723174173580176e-05,
+      "loss": 1.1798,
+      "step": 6796
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5721896667207593e-05,
+      "loss": 1.0974,
+      "step": 6797
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.572061902198141e-05,
+      "loss": 1.2848,
+      "step": 6798
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5719341237932628e-05,
+      "loss": 1.1689,
+      "step": 6799
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571806331509225e-05,
+      "loss": 1.1674,
+      "step": 6800
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5716785253491297e-05,
+      "loss": 1.0577,
+      "step": 6801
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5715507053160775e-05,
+      "loss": 1.2055,
+      "step": 6802
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5714228714131702e-05,
+      "loss": 1.063,
+      "step": 6803
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571295023643511e-05,
+      "loss": 1.15,
+      "step": 6804
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5711671620102014e-05,
+      "loss": 1.1442,
+      "step": 6805
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5710392865163443e-05,
+      "loss": 1.0186,
+      "step": 6806
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5709113971650434e-05,
+      "loss": 1.19,
+      "step": 6807
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5707834939594016e-05,
+      "loss": 1.3187,
+      "step": 6808
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570655576902523e-05,
+      "loss": 1.1335,
+      "step": 6809
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5705276459975124e-05,
+      "loss": 1.3409,
+      "step": 6810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570399701247473e-05,
+      "loss": 1.2438,
+      "step": 6811
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5702717426555104e-05,
+      "loss": 1.1346,
+      "step": 6812
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5701437702247298e-05,
+      "loss": 1.1909,
+      "step": 6813
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570015783958237e-05,
+      "loss": 1.086,
+      "step": 6814
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569887783859137e-05,
+      "loss": 1.0429,
+      "step": 6815
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569759769930537e-05,
+      "loss": 1.2968,
+      "step": 6816
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569631742175542e-05,
+      "loss": 1.1572,
+      "step": 6817
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5695037005972608e-05,
+      "loss": 1.0738,
+      "step": 6818
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5693756451987994e-05,
+      "loss": 1.0907,
+      "step": 6819
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5692475759832655e-05,
+      "loss": 1.0589,
+      "step": 6820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569119492953767e-05,
+      "loss": 1.1973,
+      "step": 6821
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568991396113412e-05,
+      "loss": 1.2443,
+      "step": 6822
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5688632854653095e-05,
+      "loss": 1.2712,
+      "step": 6823
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5687351610125676e-05,
+      "loss": 1.1877,
+      "step": 6824
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5686070227582964e-05,
+      "loss": 1.0696,
+      "step": 6825
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5684788707056046e-05,
+      "loss": 1.1776,
+      "step": 6826
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5683507048576027e-05,
+      "loss": 1.1873,
+      "step": 6827
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5682225252174003e-05,
+      "loss": 1.0926,
+      "step": 6828
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568094331788109e-05,
+      "loss": 1.1684,
+      "step": 6829
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5679661245728383e-05,
+      "loss": 1.1456,
+      "step": 6830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5678379035747e-05,
+      "loss": 1.0446,
+      "step": 6831
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5677096687968063e-05,
+      "loss": 1.1681,
+      "step": 6832
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567581420242268e-05,
+      "loss": 1.0533,
+      "step": 6833
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567453157914198e-05,
+      "loss": 1.1355,
+      "step": 6834
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5673248818157088e-05,
+      "loss": 1.1887,
+      "step": 6835
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5671965919499133e-05,
+      "loss": 1.197,
+      "step": 6836
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5670682883199245e-05,
+      "loss": 1.2364,
+      "step": 6837
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5669399709288558e-05,
+      "loss": 1.1516,
+      "step": 6838
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5668116397798213e-05,
+      "loss": 1.1075,
+      "step": 6839
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5666832948759352e-05,
+      "loss": 1.2964,
+      "step": 6840
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5665549362203125e-05,
+      "loss": 1.2072,
+      "step": 6841
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5664265638160673e-05,
+      "loss": 1.1679,
+      "step": 6842
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5662981776663155e-05,
+      "loss": 1.2583,
+      "step": 6843
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566169777774172e-05,
+      "loss": 1.1258,
+      "step": 6844
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5660413641427533e-05,
+      "loss": 1.2824,
+      "step": 6845
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5659129367751754e-05,
+      "loss": 1.1516,
+      "step": 6846
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5657844956745545e-05,
+      "loss": 1.1435,
+      "step": 6847
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565656040844008e-05,
+      "loss": 1.285,
+      "step": 6848
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565527572286653e-05,
+      "loss": 1.2493,
+      "step": 6849
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565399090005607e-05,
+      "loss": 1.1775,
+      "step": 6850
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5652705940039876e-05,
+      "loss": 1.2559,
+      "step": 6851
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5651420842849135e-05,
+      "loss": 1.2303,
+      "step": 6852
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565013560851503e-05,
+      "loss": 1.1713,
+      "step": 6853
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5648850237068745e-05,
+      "loss": 1.2652,
+      "step": 6854
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5647564728541485e-05,
+      "loss": 0.9438,
+      "step": 6855
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5646279082964437e-05,
+      "loss": 1.0438,
+      "step": 6856
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5644993300368797e-05,
+      "loss": 1.1421,
+      "step": 6857
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5643707380785772e-05,
+      "loss": 1.2144,
+      "step": 6858
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564242132424657e-05,
+      "loss": 1.2247,
+      "step": 6859
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5641135130782393e-05,
+      "loss": 1.0956,
+      "step": 6860
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5639848800424458e-05,
+      "loss": 1.081,
+      "step": 6861
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5638562333203977e-05,
+      "loss": 1.2882,
+      "step": 6862
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563727572915217e-05,
+      "loss": 1.1122,
+      "step": 6863
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5635988988300263e-05,
+      "loss": 1.2804,
+      "step": 6864
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5634702110679474e-05,
+      "loss": 1.2101,
+      "step": 6865
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5633415096321042e-05,
+      "loss": 1.2034,
+      "step": 6866
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5632127945256185e-05,
+      "loss": 1.2094,
+      "step": 6867
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5630840657516155e-05,
+      "loss": 1.22,
+      "step": 6868
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5629553233132173e-05,
+      "loss": 1.0922,
+      "step": 6869
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5628265672135496e-05,
+      "loss": 1.1146,
+      "step": 6870
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562697797455736e-05,
+      "loss": 1.0872,
+      "step": 6871
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562569014042902e-05,
+      "loss": 1.1343,
+      "step": 6872
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562440216978172e-05,
+      "loss": 1.1104,
+      "step": 6873
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5623114062646724e-05,
+      "loss": 1.1181,
+      "step": 6874
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5621825819055288e-05,
+      "loss": 1.1441,
+      "step": 6875
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562053743903867e-05,
+      "loss": 1.31,
+      "step": 6876
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561924892262814e-05,
+      "loss": 1.2262,
+      "step": 6877
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561796026985496e-05,
+      "loss": 1.0065,
+      "step": 6878
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561667148075041e-05,
+      "loss": 1.1572,
+      "step": 6879
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5615382555345758e-05,
+      "loss": 1.1778,
+      "step": 6880
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561409349367229e-05,
+      "loss": 1.1655,
+      "step": 6881
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5612804295761277e-05,
+      "loss": 1.118,
+      "step": 6882
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5611514961644016e-05,
+      "loss": 1.3415,
+      "step": 6883
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561022549135179e-05,
+      "loss": 1.0945,
+      "step": 6884
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5608935884915887e-05,
+      "loss": 1.1297,
+      "step": 6885
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5607646142367606e-05,
+      "loss": 1.0689,
+      "step": 6886
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5606356263738246e-05,
+      "loss": 1.0564,
+      "step": 6887
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5605066249059106e-05,
+      "loss": 1.1608,
+      "step": 6888
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5603776098361493e-05,
+      "loss": 1.0794,
+      "step": 6889
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5602485811676712e-05,
+      "loss": 1.1176,
+      "step": 6890
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560119538903608e-05,
+      "loss": 1.1298,
+      "step": 6891
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5599904830470906e-05,
+      "loss": 1.4022,
+      "step": 6892
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5598614136012514e-05,
+      "loss": 1.22,
+      "step": 6893
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559732330569222e-05,
+      "loss": 1.1989,
+      "step": 6894
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5596032339541345e-05,
+      "loss": 1.2137,
+      "step": 6895
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5594741237591228e-05,
+      "loss": 1.1602,
+      "step": 6896
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5593449999873195e-05,
+      "loss": 1.2057,
+      "step": 6897
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5592158626418576e-05,
+      "loss": 1.0701,
+      "step": 6898
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5590867117258718e-05,
+      "loss": 1.087,
+      "step": 6899
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5589575472424954e-05,
+      "loss": 1.1132,
+      "step": 6900
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5588283691948633e-05,
+      "loss": 1.0306,
+      "step": 6901
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5586991775861103e-05,
+      "loss": 1.2646,
+      "step": 6902
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.558569972419371e-05,
+      "loss": 1.0756,
+      "step": 6903
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5584407536977816e-05,
+      "loss": 1.2474,
+      "step": 6904
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.558311521424477e-05,
+      "loss": 1.185,
+      "step": 6905
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.558182275602594e-05,
+      "loss": 1.1854,
+      "step": 6906
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5580530162352685e-05,
+      "loss": 1.1298,
+      "step": 6907
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557923743325638e-05,
+      "loss": 1.1801,
+      "step": 6908
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5577944568768383e-05,
+      "loss": 1.1519,
+      "step": 6909
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557665156892008e-05,
+      "loss": 1.2318,
+      "step": 6910
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557535843374284e-05,
+      "loss": 1.1225,
+      "step": 6911
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557406516326805e-05,
+      "loss": 1.1765,
+      "step": 6912
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557277175752709e-05,
+      "loss": 1.0971,
+      "step": 6913
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557147821655135e-05,
+      "loss": 1.0946,
+      "step": 6914
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557018454037222e-05,
+      "loss": 1.0714,
+      "step": 6915
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556889072902109e-05,
+      "loss": 1.1615,
+      "step": 6916
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556759678252936e-05,
+      "loss": 1.3376,
+      "step": 6917
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556630270092843e-05,
+      "loss": 1.0745,
+      "step": 6918
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.55650084842497e-05,
+      "loss": 1.1065,
+      "step": 6919
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5563714132524586e-05,
+      "loss": 1.1389,
+      "step": 6920
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556241964578449e-05,
+      "loss": 1.1379,
+      "step": 6921
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556112502406083e-05,
+      "loss": 1.2037,
+      "step": 6922
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5559830267385017e-05,
+      "loss": 1.1655,
+      "step": 6923
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5558535375788474e-05,
+      "loss": 1.2225,
+      "step": 6924
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5557240349302625e-05,
+      "loss": 1.1547,
+      "step": 6925
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5555945187958896e-05,
+      "loss": 1.154,
+      "step": 6926
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5554649891788717e-05,
+      "loss": 1.1012,
+      "step": 6927
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555335446082352e-05,
+      "loss": 1.0267,
+      "step": 6928
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5552058895094747e-05,
+      "loss": 1.2843,
+      "step": 6929
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555076319463383e-05,
+      "loss": 1.0309,
+      "step": 6930
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554946735947221e-05,
+      "loss": 1.0769,
+      "step": 6931
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554817138964134e-05,
+      "loss": 1.3865,
+      "step": 6932
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554687528517267e-05,
+      "loss": 1.1728,
+      "step": 6933
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5545579046097645e-05,
+      "loss": 1.0783,
+      "step": 6934
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554428267244773e-05,
+      "loss": 1.1339,
+      "step": 6935
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5542986164254378e-05,
+      "loss": 1.1657,
+      "step": 6936
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5541689521549052e-05,
+      "loss": 1.0454,
+      "step": 6937
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554039274436322e-05,
+      "loss": 1.111,
+      "step": 6938
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5539095832728344e-05,
+      "loss": 1.1881,
+      "step": 6939
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.553779878667591e-05,
+      "loss": 1.2447,
+      "step": 6940
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5536501606237383e-05,
+      "loss": 1.0244,
+      "step": 6941
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5535204291444245e-05,
+      "loss": 1.2048,
+      "step": 6942
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5533906842327978e-05,
+      "loss": 1.1655,
+      "step": 6943
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5532609258920065e-05,
+      "loss": 1.2422,
+      "step": 6944
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5531311541251995e-05,
+      "loss": 1.229,
+      "step": 6945
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5530013689355265e-05,
+      "loss": 1.2326,
+      "step": 6946
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552871570326136e-05,
+      "loss": 1.0453,
+      "step": 6947
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552741758300179e-05,
+      "loss": 1.2069,
+      "step": 6948
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5526119328608044e-05,
+      "loss": 1.1009,
+      "step": 6949
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552482094011164e-05,
+      "loss": 1.0511,
+      "step": 6950
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5523522417544076e-05,
+      "loss": 1.1158,
+      "step": 6951
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552222376093687e-05,
+      "loss": 1.2609,
+      "step": 6952
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552092497032153e-05,
+      "loss": 1.1057,
+      "step": 6953
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.551962604572958e-05,
+      "loss": 1.055,
+      "step": 6954
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5518326987192537e-05,
+      "loss": 1.233,
+      "step": 6955
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.551702779474193e-05,
+      "loss": 1.0169,
+      "step": 6956
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5515728468409282e-05,
+      "loss": 1.1704,
+      "step": 6957
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5514429008226124e-05,
+      "loss": 1.2022,
+      "step": 6958
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5513129414223992e-05,
+      "loss": 1.0804,
+      "step": 6959
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5511829686434422e-05,
+      "loss": 0.9743,
+      "step": 6960
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5510529824888957e-05,
+      "loss": 1.1119,
+      "step": 6961
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5509229829619136e-05,
+      "loss": 1.1188,
+      "step": 6962
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5507929700656513e-05,
+      "loss": 1.2014,
+      "step": 6963
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.550662943803263e-05,
+      "loss": 1.2862,
+      "step": 6964
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5505329041779048e-05,
+      "loss": 1.1739,
+      "step": 6965
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.550402851192732e-05,
+      "loss": 1.2098,
+      "step": 6966
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5502727848509008e-05,
+      "loss": 1.0577,
+      "step": 6967
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.550142705155567e-05,
+      "loss": 1.2019,
+      "step": 6968
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5500126121098883e-05,
+      "loss": 1.1873,
+      "step": 6969
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5498825057170207e-05,
+      "loss": 1.1454,
+      "step": 6970
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5497523859801215e-05,
+      "loss": 1.1491,
+      "step": 6971
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549622252902349e-05,
+      "loss": 1.0182,
+      "step": 6972
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549492106486861e-05,
+      "loss": 0.9897,
+      "step": 6973
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549361946736815e-05,
+      "loss": 1.3104,
+      "step": 6974
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549231773655371e-05,
+      "loss": 1.1332,
+      "step": 6975
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5491015872456864e-05,
+      "loss": 1.1702,
+      "step": 6976
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.548971387510921e-05,
+      "loss": 1.1587,
+      "step": 6977
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.548841174454235e-05,
+      "loss": 1.2019,
+      "step": 6978
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5487109480787875e-05,
+      "loss": 1.1406,
+      "step": 6979
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5485807083877393e-05,
+      "loss": 1.1216,
+      "step": 6980
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5484504553842503e-05,
+      "loss": 1.0382,
+      "step": 6981
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5483201890714816e-05,
+      "loss": 1.213,
+      "step": 6982
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5481899094525945e-05,
+      "loss": 1.0532,
+      "step": 6983
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.548059616530751e-05,
+      "loss": 1.1417,
+      "step": 6984
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5479293103091116e-05,
+      "loss": 1.2342,
+      "step": 6985
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54779899079084e-05,
+      "loss": 1.155,
+      "step": 6986
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5476686579790975e-05,
+      "loss": 1.1486,
+      "step": 6987
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5475383118770473e-05,
+      "loss": 1.1671,
+      "step": 6988
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5474079524878525e-05,
+      "loss": 1.1435,
+      "step": 6989
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5472775798146766e-05,
+      "loss": 1.0676,
+      "step": 6990
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5471471938606834e-05,
+      "loss": 1.2853,
+      "step": 6991
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.547016794629037e-05,
+      "loss": 1.1476,
+      "step": 6992
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5468863821229016e-05,
+      "loss": 1.0397,
+      "step": 6993
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546755956345442e-05,
+      "loss": 1.172,
+      "step": 6994
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546625517299823e-05,
+      "loss": 1.2499,
+      "step": 6995
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546495064989211e-05,
+      "loss": 1.1471,
+      "step": 6996
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5463645994167705e-05,
+      "loss": 1.1833,
+      "step": 6997
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546234120585668e-05,
+      "loss": 1.1237,
+      "step": 6998
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5461036284990698e-05,
+      "loss": 1.3381,
+      "step": 6999
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5459731231601425e-05,
+      "loss": 1.1514,
+      "step": 7000
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5458426045720533e-05,
+      "loss": 1.3403,
+      "step": 7001
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5457120727379694e-05,
+      "loss": 1.1729,
+      "step": 7002
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5455815276610586e-05,
+      "loss": 1.1745,
+      "step": 7003
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5454509693444878e-05,
+      "loss": 1.2083,
+      "step": 7004
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5453203977914267e-05,
+      "loss": 1.1702,
+      "step": 7005
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545189813005043e-05,
+      "loss": 1.1687,
+      "step": 7006
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545059214988506e-05,
+      "loss": 1.1873,
+      "step": 7007
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544928603744985e-05,
+      "loss": 1.1526,
+      "step": 7008
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5447979792776493e-05,
+      "loss": 1.0574,
+      "step": 7009
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5446673415896686e-05,
+      "loss": 1.1427,
+      "step": 7010
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5445366906842137e-05,
+      "loss": 1.067,
+      "step": 7011
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5444060265644546e-05,
+      "loss": 1.2099,
+      "step": 7012
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544275349233562e-05,
+      "loss": 1.1174,
+      "step": 7013
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5441446586947076e-05,
+      "loss": 1.1173,
+      "step": 7014
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5440139549510624e-05,
+      "loss": 1.1958,
+      "step": 7015
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5438832380057987e-05,
+      "loss": 1.2737,
+      "step": 7016
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.543752507862088e-05,
+      "loss": 1.131,
+      "step": 7017
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5436217645231037e-05,
+      "loss": 1.1021,
+      "step": 7018
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5434910079920172e-05,
+      "loss": 1.2648,
+      "step": 7019
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5433602382720022e-05,
+      "loss": 1.287,
+      "step": 7020
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5432294553662325e-05,
+      "loss": 1.1459,
+      "step": 7021
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5430986592778817e-05,
+      "loss": 1.0149,
+      "step": 7022
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5429678500101234e-05,
+      "loss": 1.2324,
+      "step": 7023
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542837027566132e-05,
+      "loss": 1.2138,
+      "step": 7024
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5427061919490826e-05,
+      "loss": 1.0905,
+      "step": 7025
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54257534316215e-05,
+      "loss": 1.1883,
+      "step": 7026
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5424444812085094e-05,
+      "loss": 1.1865,
+      "step": 7027
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5423136060913365e-05,
+      "loss": 1.2403,
+      "step": 7028
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5421827178138072e-05,
+      "loss": 1.0765,
+      "step": 7029
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542051816379098e-05,
+      "loss": 1.0594,
+      "step": 7030
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5419209017903855e-05,
+      "loss": 1.1328,
+      "step": 7031
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541789974050846e-05,
+      "loss": 1.1094,
+      "step": 7032
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5416590331636574e-05,
+      "loss": 1.0272,
+      "step": 7033
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541528079131997e-05,
+      "loss": 1.3184,
+      "step": 7034
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541397111959043e-05,
+      "loss": 1.2621,
+      "step": 7035
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5412661316479728e-05,
+      "loss": 1.05,
+      "step": 7036
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5411351382019657e-05,
+      "loss": 1.2197,
+      "step": 7037
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5410041316242003e-05,
+      "loss": 1.1948,
+      "step": 7038
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5408731119178554e-05,
+      "loss": 1.1468,
+      "step": 7039
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5407420790861106e-05,
+      "loss": 1.2469,
+      "step": 7040
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5406110331321463e-05,
+      "loss": 1.2705,
+      "step": 7041
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5404799740591415e-05,
+      "loss": 1.1218,
+      "step": 7042
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5403489018702775e-05,
+      "loss": 1.16,
+      "step": 7043
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5402178165687352e-05,
+      "loss": 1.1402,
+      "step": 7044
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5400867181576945e-05,
+      "loss": 0.9432,
+      "step": 7045
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5399556066403382e-05,
+      "loss": 0.9593,
+      "step": 7046
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5398244820198467e-05,
+      "loss": 1.133,
+      "step": 7047
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5396933442994025e-05,
+      "loss": 0.8468,
+      "step": 7048
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5395621934821884e-05,
+      "loss": 1.2073,
+      "step": 7049
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5394310295713862e-05,
+      "loss": 1.286,
+      "step": 7050
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.53929985257018e-05,
+      "loss": 1.2514,
+      "step": 7051
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5391686624817518e-05,
+      "loss": 1.0783,
+      "step": 7052
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5390374593092862e-05,
+      "loss": 1.1799,
+      "step": 7053
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538906243055967e-05,
+      "loss": 1.236,
+      "step": 7054
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5387750137249775e-05,
+      "loss": 1.1353,
+      "step": 7055
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538643771319503e-05,
+      "loss": 1.0644,
+      "step": 7056
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5385125158427285e-05,
+      "loss": 1.0501,
+      "step": 7057
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5383812472978386e-05,
+      "loss": 1.2949,
+      "step": 7058
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5382499656880198e-05,
+      "loss": 1.2141,
+      "step": 7059
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5381186710164565e-05,
+      "loss": 1.2311,
+      "step": 7060
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537987363286336e-05,
+      "loss": 1.0931,
+      "step": 7061
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5378560425008444e-05,
+      "loss": 1.2226,
+      "step": 7062
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537724708663168e-05,
+      "loss": 1.1781,
+      "step": 7063
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537593361776495e-05,
+      "loss": 1.1496,
+      "step": 7064
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537462001844012e-05,
+      "loss": 1.0998,
+      "step": 7065
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5373306288689063e-05,
+      "loss": 1.1559,
+      "step": 7066
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5371992428543668e-05,
+      "loss": 1.3266,
+      "step": 7067
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5370678438035813e-05,
+      "loss": 1.0325,
+      "step": 7068
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536936431719739e-05,
+      "loss": 1.1624,
+      "step": 7069
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5368050066060285e-05,
+      "loss": 1.1314,
+      "step": 7070
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536673568465639e-05,
+      "loss": 1.0871,
+      "step": 7071
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5365421173017604e-05,
+      "loss": 1.0245,
+      "step": 7072
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536410653117583e-05,
+      "loss": 1.1685,
+      "step": 7073
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536279175916296e-05,
+      "loss": 1.2354,
+      "step": 7074
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5361476857010907e-05,
+      "loss": 1.0835,
+      "step": 7075
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5360161824751575e-05,
+      "loss": 1.1377,
+      "step": 7076
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5358846662416884e-05,
+      "loss": 0.9357,
+      "step": 7077
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.535753137003874e-05,
+      "loss": 1.1563,
+      "step": 7078
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5356215947649072e-05,
+      "loss": 1.1544,
+      "step": 7079
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.535490039527979e-05,
+      "loss": 1.142,
+      "step": 7080
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5353584712962825e-05,
+      "loss": 1.1163,
+      "step": 7081
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5352268900730104e-05,
+      "loss": 1.1721,
+      "step": 7082
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5350952958613558e-05,
+      "loss": 1.1738,
+      "step": 7083
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5349636886645115e-05,
+      "loss": 1.1702,
+      "step": 7084
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.534832068485672e-05,
+      "loss": 1.1974,
+      "step": 7085
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5347004353280313e-05,
+      "loss": 1.1721,
+      "step": 7086
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5345687891947837e-05,
+      "loss": 1.2384,
+      "step": 7087
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.534437130089123e-05,
+      "loss": 1.3245,
+      "step": 7088
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5343054580142457e-05,
+      "loss": 1.0626,
+      "step": 7089
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5341737729733453e-05,
+      "loss": 1.0415,
+      "step": 7090
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5340420749696194e-05,
+      "loss": 1.0951,
+      "step": 7091
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5339103640062624e-05,
+      "loss": 1.2019,
+      "step": 7092
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5337786400864708e-05,
+      "loss": 1.1045,
+      "step": 7093
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533646903213442e-05,
+      "loss": 1.2451,
+      "step": 7094
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533515153390372e-05,
+      "loss": 1.2519,
+      "step": 7095
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533383390620458e-05,
+      "loss": 1.2254,
+      "step": 7096
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533251614906898e-05,
+      "loss": 1.114,
+      "step": 7097
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5331198262528895e-05,
+      "loss": 0.9621,
+      "step": 7098
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5329880246616308e-05,
+      "loss": 1.1631,
+      "step": 7099
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5328562101363196e-05,
+      "loss": 1.1852,
+      "step": 7100
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5327243826801557e-05,
+      "loss": 1.2069,
+      "step": 7101
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5325925422963378e-05,
+      "loss": 1.2338,
+      "step": 7102
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532460688988065e-05,
+      "loss": 1.2274,
+      "step": 7103
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5323288227585373e-05,
+      "loss": 1.3108,
+      "step": 7104
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5321969436109545e-05,
+      "loss": 1.2505,
+      "step": 7105
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532065051548517e-05,
+      "loss": 1.2714,
+      "step": 7106
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5319331465744257e-05,
+      "loss": 1.0713,
+      "step": 7107
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531801228691881e-05,
+      "loss": 1.1481,
+      "step": 7108
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5316692979040847e-05,
+      "loss": 1.159,
+      "step": 7109
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531537354214238e-05,
+      "loss": 1.2135,
+      "step": 7110
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5314053976255427e-05,
+      "loss": 1.1541,
+      "step": 7111
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5312734281412016e-05,
+      "loss": 1.1563,
+      "step": 7112
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5311414457644165e-05,
+      "loss": 0.9976,
+      "step": 7113
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5310094504983907e-05,
+      "loss": 1.1533,
+      "step": 7114
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.530877442346327e-05,
+      "loss": 1.2155,
+      "step": 7115
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5307454213114292e-05,
+      "loss": 1.0592,
+      "step": 7116
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5306133873969008e-05,
+      "loss": 1.3281,
+      "step": 7117
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5304813406059462e-05,
+      "loss": 1.1773,
+      "step": 7118
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5303492809417693e-05,
+      "loss": 1.1738,
+      "step": 7119
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5302172084075752e-05,
+      "loss": 1.1359,
+      "step": 7120
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.530085123006569e-05,
+      "loss": 1.1617,
+      "step": 7121
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529953024741955e-05,
+      "loss": 1.0753,
+      "step": 7122
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5298209136169403e-05,
+      "loss": 1.2655,
+      "step": 7123
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5296887896347297e-05,
+      "loss": 1.2478,
+      "step": 7124
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5295566527985306e-05,
+      "loss": 1.2187,
+      "step": 7125
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5294245031115484e-05,
+      "loss": 1.2224,
+      "step": 7126
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5292923405769907e-05,
+      "loss": 1.2544,
+      "step": 7127
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5291601651980644e-05,
+      "loss": 1.0855,
+      "step": 7128
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529027976977977e-05,
+      "loss": 1.0098,
+      "step": 7129
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5288957759199366e-05,
+      "loss": 1.2784,
+      "step": 7130
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5287635620271512e-05,
+      "loss": 1.1164,
+      "step": 7131
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528631335302829e-05,
+      "loss": 1.1454,
+      "step": 7132
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5284990957501788e-05,
+      "loss": 1.0916,
+      "step": 7133
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5283668433724098e-05,
+      "loss": 1.1832,
+      "step": 7134
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5282345781727318e-05,
+      "loss": 1.2591,
+      "step": 7135
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528102300154354e-05,
+      "loss": 1.1322,
+      "step": 7136
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.527970009320486e-05,
+      "loss": 1.0997,
+      "step": 7137
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.527837705674339e-05,
+      "loss": 1.0738,
+      "step": 7138
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5277053892191233e-05,
+      "loss": 1.1582,
+      "step": 7139
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5275730599580497e-05,
+      "loss": 1.2868,
+      "step": 7140
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.527440717894329e-05,
+      "loss": 1.1606,
+      "step": 7141
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5273083630311736e-05,
+      "loss": 1.2827,
+      "step": 7142
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.527175995371795e-05,
+      "loss": 1.2172,
+      "step": 7143
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5270436149194053e-05,
+      "loss": 1.1487,
+      "step": 7144
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526911221677217e-05,
+      "loss": 1.2456,
+      "step": 7145
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5267788156484428e-05,
+      "loss": 1.1107,
+      "step": 7146
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526646396836296e-05,
+      "loss": 1.1138,
+      "step": 7147
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.52651396524399e-05,
+      "loss": 1.1818,
+      "step": 7148
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526381520874739e-05,
+      "loss": 1.2698,
+      "step": 7149
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5262490637317555e-05,
+      "loss": 1.2255,
+      "step": 7150
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5261165938182555e-05,
+      "loss": 1.171,
+      "step": 7151
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.525984111137453e-05,
+      "loss": 1.0895,
+      "step": 7152
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5258516156925627e-05,
+      "loss": 1.2665,
+      "step": 7153
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5257191074868004e-05,
+      "loss": 1.1356,
+      "step": 7154
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5255865865233817e-05,
+      "loss": 1.0695,
+      "step": 7155
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5254540528055219e-05,
+      "loss": 1.1569,
+      "step": 7156
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5253215063364378e-05,
+      "loss": 1.2115,
+      "step": 7157
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5251889471193453e-05,
+      "loss": 0.9811,
+      "step": 7158
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.525056375157462e-05,
+      "loss": 1.0982,
+      "step": 7159
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5249237904540041e-05,
+      "loss": 1.141,
+      "step": 7160
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5247911930121899e-05,
+      "loss": 1.1495,
+      "step": 7161
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5246585828352367e-05,
+      "loss": 1.1111,
+      "step": 7162
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5245259599263627e-05,
+      "loss": 1.2256,
+      "step": 7163
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5243933242887863e-05,
+      "loss": 1.0392,
+      "step": 7164
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.524260675925726e-05,
+      "loss": 1.081,
+      "step": 7165
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5241280148404009e-05,
+      "loss": 1.3375,
+      "step": 7166
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5239953410360302e-05,
+      "loss": 1.2181,
+      "step": 7167
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.523862654515834e-05,
+      "loss": 1.1386,
+      "step": 7168
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5237299552830313e-05,
+      "loss": 1.158,
+      "step": 7169
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.523597243340843e-05,
+      "loss": 1.2757,
+      "step": 7170
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5234645186924892e-05,
+      "loss": 1.2184,
+      "step": 7171
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5233317813411916e-05,
+      "loss": 1.1813,
+      "step": 7172
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5231990312901701e-05,
+      "loss": 1.2365,
+      "step": 7173
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5230662685426472e-05,
+      "loss": 1.2353,
+      "step": 7174
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.522933493101844e-05,
+      "loss": 1.0502,
+      "step": 7175
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5228007049709829e-05,
+      "loss": 1.0686,
+      "step": 7176
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5226679041532863e-05,
+      "loss": 1.2687,
+      "step": 7177
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5225350906519766e-05,
+      "loss": 1.0025,
+      "step": 7178
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5224022644702769e-05,
+      "loss": 1.222,
+      "step": 7179
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.522269425611411e-05,
+      "loss": 0.9187,
+      "step": 7180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5221365740786016e-05,
+      "loss": 1.081,
+      "step": 7181
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5220037098750733e-05,
+      "loss": 1.1027,
+      "step": 7182
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5218708330040502e-05,
+      "loss": 1.1108,
+      "step": 7183
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5217379434687566e-05,
+      "loss": 1.3226,
+      "step": 7184
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5216050412724178e-05,
+      "loss": 1.2432,
+      "step": 7185
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5214721264182584e-05,
+      "loss": 1.1433,
+      "step": 7186
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5213391989095036e-05,
+      "loss": 1.1658,
+      "step": 7187
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5212062587493801e-05,
+      "loss": 1.2343,
+      "step": 7188
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5210733059411135e-05,
+      "loss": 1.2073,
+      "step": 7189
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5209403404879305e-05,
+      "loss": 1.1137,
+      "step": 7190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.520807362393057e-05,
+      "loss": 1.041,
+      "step": 7191
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5206743716597208e-05,
+      "loss": 1.0703,
+      "step": 7192
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5205413682911485e-05,
+      "loss": 1.0451,
+      "step": 7193
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5204083522905684e-05,
+      "loss": 1.3286,
+      "step": 7194
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5202753236612078e-05,
+      "loss": 1.0996,
+      "step": 7195
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5201422824062952e-05,
+      "loss": 1.1442,
+      "step": 7196
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5200092285290594e-05,
+      "loss": 1.1666,
+      "step": 7197
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5198761620327287e-05,
+      "loss": 1.2949,
+      "step": 7198
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5197430829205321e-05,
+      "loss": 1.1391,
+      "step": 7199
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5196099911956998e-05,
+      "loss": 1.0491,
+      "step": 7200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.519476886861461e-05,
+      "loss": 1.2436,
+      "step": 7201
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.519343769921046e-05,
+      "loss": 1.0793,
+      "step": 7202
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.519210640377685e-05,
+      "loss": 1.0372,
+      "step": 7203
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5190774982346085e-05,
+      "loss": 1.1306,
+      "step": 7204
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.518944343495048e-05,
+      "loss": 1.1987,
+      "step": 7205
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5188111761622342e-05,
+      "loss": 1.059,
+      "step": 7206
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5186779962393989e-05,
+      "loss": 1.3342,
+      "step": 7207
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5185448037297742e-05,
+      "loss": 1.2271,
+      "step": 7208
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.518411598636592e-05,
+      "loss": 1.2265,
+      "step": 7209
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5182783809630847e-05,
+      "loss": 1.1616,
+      "step": 7210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5181451507124856e-05,
+      "loss": 1.1964,
+      "step": 7211
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5180119078880276e-05,
+      "loss": 1.263,
+      "step": 7212
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5178786524929436e-05,
+      "loss": 1.3254,
+      "step": 7213
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5177453845304681e-05,
+      "loss": 1.0724,
+      "step": 7214
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5176121040038342e-05,
+      "loss": 1.1209,
+      "step": 7215
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5174788109162775e-05,
+      "loss": 1.1096,
+      "step": 7216
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5173455052710315e-05,
+      "loss": 1.2078,
+      "step": 7217
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5172121870713316e-05,
+      "loss": 1.1846,
+      "step": 7218
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.517078856320413e-05,
+      "loss": 1.2589,
+      "step": 7219
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5169455130215111e-05,
+      "loss": 1.065,
+      "step": 7220
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5168121571778622e-05,
+      "loss": 1.2511,
+      "step": 7221
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5166787887927017e-05,
+      "loss": 1.1058,
+      "step": 7222
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.516545407869267e-05,
+      "loss": 1.0536,
+      "step": 7223
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5164120144107939e-05,
+      "loss": 1.0761,
+      "step": 7224
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5162786084205204e-05,
+      "loss": 1.1496,
+      "step": 7225
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5161451899016828e-05,
+      "loss": 1.141,
+      "step": 7226
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5160117588575198e-05,
+      "loss": 1.1356,
+      "step": 7227
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5158783152912688e-05,
+      "loss": 1.1349,
+      "step": 7228
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5157448592061683e-05,
+      "loss": 1.1067,
+      "step": 7229
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5156113906054568e-05,
+      "loss": 1.1419,
+      "step": 7230
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5154779094923733e-05,
+      "loss": 1.1044,
+      "step": 7231
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5153444158701567e-05,
+      "loss": 1.1964,
+      "step": 7232
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.515210909742047e-05,
+      "loss": 1.1785,
+      "step": 7233
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5150773911112833e-05,
+      "loss": 1.1896,
+      "step": 7234
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5149438599811062e-05,
+      "loss": 0.8548,
+      "step": 7235
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5148103163547562e-05,
+      "loss": 1.07,
+      "step": 7236
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5146767602354738e-05,
+      "loss": 1.1822,
+      "step": 7237
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5145431916265e-05,
+      "loss": 1.169,
+      "step": 7238
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5144096105310761e-05,
+      "loss": 1.1993,
+      "step": 7239
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5142760169524438e-05,
+      "loss": 1.0791,
+      "step": 7240
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.514142410893845e-05,
+      "loss": 1.1171,
+      "step": 7241
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5140087923585219e-05,
+      "loss": 1.1826,
+      "step": 7242
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.513875161349717e-05,
+      "loss": 1.1871,
+      "step": 7243
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5137415178706733e-05,
+      "loss": 1.2663,
+      "step": 7244
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5136078619246339e-05,
+      "loss": 1.1619,
+      "step": 7245
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.513474193514842e-05,
+      "loss": 1.2065,
+      "step": 7246
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5133405126445417e-05,
+      "loss": 1.2142,
+      "step": 7247
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5132068193169765e-05,
+      "loss": 1.0819,
+      "step": 7248
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5130731135353914e-05,
+      "loss": 1.1135,
+      "step": 7249
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5129393953030307e-05,
+      "loss": 1.2768,
+      "step": 7250
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5128056646231393e-05,
+      "loss": 1.1125,
+      "step": 7251
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5126719214989622e-05,
+      "loss": 1.1411,
+      "step": 7252
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5125381659337458e-05,
+      "loss": 1.2797,
+      "step": 7253
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5124043979307353e-05,
+      "loss": 1.1891,
+      "step": 7254
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.512270617493177e-05,
+      "loss": 1.0856,
+      "step": 7255
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.512136824624317e-05,
+      "loss": 1.1558,
+      "step": 7256
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.512003019327403e-05,
+      "loss": 1.0232,
+      "step": 7257
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.511869201605681e-05,
+      "loss": 1.2105,
+      "step": 7258
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.511735371462399e-05,
+      "loss": 0.8651,
+      "step": 7259
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5116015289008043e-05,
+      "loss": 1.2834,
+      "step": 7260
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5114676739241454e-05,
+      "loss": 1.2247,
+      "step": 7261
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.51133380653567e-05,
+      "loss": 1.2594,
+      "step": 7262
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5111999267386268e-05,
+      "loss": 1.0238,
+      "step": 7263
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5110660345362651e-05,
+      "loss": 1.2987,
+      "step": 7264
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5109321299318333e-05,
+      "loss": 1.1245,
+      "step": 7265
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5107982129285816e-05,
+      "loss": 1.1736,
+      "step": 7266
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5106642835297596e-05,
+      "loss": 1.1748,
+      "step": 7267
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5105303417386168e-05,
+      "loss": 0.968,
+      "step": 7268
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5103963875584046e-05,
+      "loss": 1.223,
+      "step": 7269
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5102624209923723e-05,
+      "loss": 1.1933,
+      "step": 7270
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5101284420437722e-05,
+      "loss": 1.2382,
+      "step": 7271
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.509994450715855e-05,
+      "loss": 1.2495,
+      "step": 7272
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5098604470118724e-05,
+      "loss": 1.1124,
+      "step": 7273
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5097264309350761e-05,
+      "loss": 1.086,
+      "step": 7274
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5095924024887185e-05,
+      "loss": 1.0417,
+      "step": 7275
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5094583616760518e-05,
+      "loss": 1.1122,
+      "step": 7276
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.509324308500329e-05,
+      "loss": 1.0623,
+      "step": 7277
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5091902429648029e-05,
+      "loss": 1.2473,
+      "step": 7278
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5090561650727274e-05,
+      "loss": 1.1635,
+      "step": 7279
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5089220748273557e-05,
+      "loss": 1.2019,
+      "step": 7280
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.508787972231942e-05,
+      "loss": 1.2191,
+      "step": 7281
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5086538572897404e-05,
+      "loss": 1.1562,
+      "step": 7282
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5085197300040055e-05,
+      "loss": 1.0983,
+      "step": 7283
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5083855903779925e-05,
+      "loss": 1.1645,
+      "step": 7284
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.508251438414956e-05,
+      "loss": 1.174,
+      "step": 7285
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5081172741181522e-05,
+      "loss": 1.1325,
+      "step": 7286
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5079830974908362e-05,
+      "loss": 1.1089,
+      "step": 7287
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5078489085362643e-05,
+      "loss": 1.3409,
+      "step": 7288
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5077147072576932e-05,
+      "loss": 0.992,
+      "step": 7289
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5075804936583791e-05,
+      "loss": 1.2419,
+      "step": 7290
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5074462677415793e-05,
+      "loss": 1.1864,
+      "step": 7291
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5073120295105507e-05,
+      "loss": 1.1539,
+      "step": 7292
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5071777789685514e-05,
+      "loss": 0.95,
+      "step": 7293
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5070435161188386e-05,
+      "loss": 1.1314,
+      "step": 7294
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5069092409646713e-05,
+      "loss": 1.0563,
+      "step": 7295
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5067749535093068e-05,
+      "loss": 1.2439,
+      "step": 7296
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5066406537560053e-05,
+      "loss": 1.1569,
+      "step": 7297
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5065063417080246e-05,
+      "loss": 1.3059,
+      "step": 7298
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5063720173686246e-05,
+      "loss": 1.0454,
+      "step": 7299
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.506237680741065e-05,
+      "loss": 1.093,
+      "step": 7300
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.506103331828606e-05,
+      "loss": 1.2905,
+      "step": 7301
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5059689706345072e-05,
+      "loss": 1.1631,
+      "step": 7302
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5058345971620294e-05,
+      "loss": 1.2864,
+      "step": 7303
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5057002114144336e-05,
+      "loss": 1.2392,
+      "step": 7304
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5055658133949811e-05,
+      "loss": 1.2469,
+      "step": 7305
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.505431403106933e-05,
+      "loss": 0.9542,
+      "step": 7306
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5052969805535509e-05,
+      "loss": 1.2136,
+      "step": 7307
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5051625457380975e-05,
+      "loss": 1.1609,
+      "step": 7308
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5050280986638345e-05,
+      "loss": 1.1351,
+      "step": 7309
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.504893639334025e-05,
+      "loss": 1.0982,
+      "step": 7310
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5047591677519315e-05,
+      "loss": 1.0334,
+      "step": 7311
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5046246839208176e-05,
+      "loss": 1.1567,
+      "step": 7312
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5044901878439467e-05,
+      "loss": 1.0376,
+      "step": 7313
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5043556795245825e-05,
+      "loss": 1.1277,
+      "step": 7314
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5042211589659892e-05,
+      "loss": 1.204,
+      "step": 7315
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5040866261714311e-05,
+      "loss": 1.0308,
+      "step": 7316
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5039520811441733e-05,
+      "loss": 1.205,
+      "step": 7317
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5038175238874803e-05,
+      "loss": 1.3597,
+      "step": 7318
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5036829544046178e-05,
+      "loss": 1.2844,
+      "step": 7319
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5035483726988513e-05,
+      "loss": 1.1712,
+      "step": 7320
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5034137787734468e-05,
+      "loss": 1.2182,
+      "step": 7321
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.50327917263167e-05,
+      "loss": 1.0748,
+      "step": 7322
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5031445542767884e-05,
+      "loss": 1.0844,
+      "step": 7323
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5030099237120674e-05,
+      "loss": 1.1215,
+      "step": 7324
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5028752809407752e-05,
+      "loss": 1.2046,
+      "step": 7325
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5027406259661791e-05,
+      "loss": 1.1127,
+      "step": 7326
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5026059587915463e-05,
+      "loss": 0.9676,
+      "step": 7327
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5024712794201449e-05,
+      "loss": 1.0962,
+      "step": 7328
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5023365878552434e-05,
+      "loss": 1.1239,
+      "step": 7329
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.50220188410011e-05,
+      "loss": 1.2778,
+      "step": 7330
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.502067168158014e-05,
+      "loss": 1.1672,
+      "step": 7331
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5019324400322244e-05,
+      "loss": 1.2214,
+      "step": 7332
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5017976997260104e-05,
+      "loss": 1.1876,
+      "step": 7333
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5016629472426423e-05,
+      "loss": 1.1118,
+      "step": 7334
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5015281825853896e-05,
+      "loss": 1.1839,
+      "step": 7335
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5013934057575228e-05,
+      "loss": 1.2285,
+      "step": 7336
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5012586167623124e-05,
+      "loss": 1.2153,
+      "step": 7337
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.50112381560303e-05,
+      "loss": 1.1833,
+      "step": 7338
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.500989002282946e-05,
+      "loss": 1.1639,
+      "step": 7339
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5008541768053324e-05,
+      "loss": 1.1834,
+      "step": 7340
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5007193391734607e-05,
+      "loss": 1.2676,
+      "step": 7341
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5005844893906036e-05,
+      "loss": 1.2453,
+      "step": 7342
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5004496274600327e-05,
+      "loss": 1.1241,
+      "step": 7343
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5003147533850215e-05,
+      "loss": 1.0855,
+      "step": 7344
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5001798671688424e-05,
+      "loss": 1.2454,
+      "step": 7345
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.500044968814769e-05,
+      "loss": 0.9956,
+      "step": 7346
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4999100583260743e-05,
+      "loss": 1.0251,
+      "step": 7347
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4997751357060334e-05,
+      "loss": 1.2779,
+      "step": 7348
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4996402009579192e-05,
+      "loss": 1.1009,
+      "step": 7349
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.499505254085007e-05,
+      "loss": 1.0289,
+      "step": 7350
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4993702950905711e-05,
+      "loss": 1.1174,
+      "step": 7351
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.499235323977887e-05,
+      "loss": 1.0914,
+      "step": 7352
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4991003407502293e-05,
+      "loss": 1.1536,
+      "step": 7353
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4989653454108744e-05,
+      "loss": 1.3335,
+      "step": 7354
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.498830337963098e-05,
+      "loss": 1.3615,
+      "step": 7355
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4986953184101763e-05,
+      "loss": 1.1338,
+      "step": 7356
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4985602867553857e-05,
+      "loss": 1.094,
+      "step": 7357
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4984252430020033e-05,
+      "loss": 1.1343,
+      "step": 7358
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.498290187153306e-05,
+      "loss": 1.1458,
+      "step": 7359
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4981551192125714e-05,
+      "loss": 1.2501,
+      "step": 7360
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4980200391830771e-05,
+      "loss": 1.0474,
+      "step": 7361
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.497884947068101e-05,
+      "loss": 1.1642,
+      "step": 7362
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4977498428709214e-05,
+      "loss": 1.2075,
+      "step": 7363
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4976147265948168e-05,
+      "loss": 1.0864,
+      "step": 7364
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4974795982430665e-05,
+      "loss": 1.239,
+      "step": 7365
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4973444578189492e-05,
+      "loss": 1.1996,
+      "step": 7366
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4972093053257447e-05,
+      "loss": 1.3102,
+      "step": 7367
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4970741407667326e-05,
+      "loss": 1.2492,
+      "step": 7368
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4969389641451928e-05,
+      "loss": 1.1694,
+      "step": 7369
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4968037754644061e-05,
+      "loss": 1.1558,
+      "step": 7370
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4966685747276524e-05,
+      "loss": 1.1498,
+      "step": 7371
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4965333619382134e-05,
+      "loss": 1.0855,
+      "step": 7372
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4963981370993696e-05,
+      "loss": 1.1225,
+      "step": 7373
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.496262900214403e-05,
+      "loss": 1.0375,
+      "step": 7374
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4961276512865954e-05,
+      "loss": 1.1808,
+      "step": 7375
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4959923903192289e-05,
+      "loss": 1.1924,
+      "step": 7376
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4958571173155854e-05,
+      "loss": 1.0561,
+      "step": 7377
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4957218322789481e-05,
+      "loss": 1.225,
+      "step": 7378
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4955865352125995e-05,
+      "loss": 1.1688,
+      "step": 7379
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4954512261198235e-05,
+      "loss": 1.1319,
+      "step": 7380
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4953159050039033e-05,
+      "loss": 1.2335,
+      "step": 7381
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4951805718681227e-05,
+      "loss": 1.1305,
+      "step": 7382
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4950452267157658e-05,
+      "loss": 1.03,
+      "step": 7383
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4949098695501172e-05,
+      "loss": 1.2482,
+      "step": 7384
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4947745003744616e-05,
+      "loss": 1.2059,
+      "step": 7385
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4946391191920839e-05,
+      "loss": 1.183,
+      "step": 7386
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4945037260062694e-05,
+      "loss": 1.0713,
+      "step": 7387
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4943683208203039e-05,
+      "loss": 1.2681,
+      "step": 7388
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4942329036374729e-05,
+      "loss": 1.2061,
+      "step": 7389
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.494097474461063e-05,
+      "loss": 1.0453,
+      "step": 7390
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4939620332943605e-05,
+      "loss": 1.2119,
+      "step": 7391
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.493826580140652e-05,
+      "loss": 1.2015,
+      "step": 7392
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.493691115003225e-05,
+      "loss": 1.1482,
+      "step": 7393
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4935556378853662e-05,
+      "loss": 1.1516,
+      "step": 7394
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4934201487903635e-05,
+      "loss": 1.1523,
+      "step": 7395
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.493284647721505e-05,
+      "loss": 1.0933,
+      "step": 7396
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4931491346820787e-05,
+      "loss": 1.1525,
+      "step": 7397
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4930136096753736e-05,
+      "loss": 1.2403,
+      "step": 7398
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4928780727046777e-05,
+      "loss": 1.1456,
+      "step": 7399
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.492742523773281e-05,
+      "loss": 1.2114,
+      "step": 7400
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4926069628844717e-05,
+      "loss": 0.9722,
+      "step": 7401
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4924713900415407e-05,
+      "loss": 1.3125,
+      "step": 7402
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4923358052477767e-05,
+      "loss": 1.1475,
+      "step": 7403
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4922002085064711e-05,
+      "loss": 1.1026,
+      "step": 7404
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4920645998209138e-05,
+      "loss": 0.9034,
+      "step": 7405
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4919289791943958e-05,
+      "loss": 0.9902,
+      "step": 7406
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4917933466302082e-05,
+      "loss": 1.1814,
+      "step": 7407
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4916577021316425e-05,
+      "loss": 1.0734,
+      "step": 7408
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4915220457019902e-05,
+      "loss": 1.1631,
+      "step": 7409
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4913863773445432e-05,
+      "loss": 1.0338,
+      "step": 7410
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.491250697062594e-05,
+      "loss": 1.1078,
+      "step": 7411
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.491115004859435e-05,
+      "loss": 1.0217,
+      "step": 7412
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4909793007383595e-05,
+      "loss": 1.2607,
+      "step": 7413
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4908435847026601e-05,
+      "loss": 1.0525,
+      "step": 7414
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4907078567556303e-05,
+      "loss": 1.1692,
+      "step": 7415
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.490572116900564e-05,
+      "loss": 1.121,
+      "step": 7416
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4904363651407554e-05,
+      "loss": 1.1685,
+      "step": 7417
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4903006014794983e-05,
+      "loss": 1.1176,
+      "step": 7418
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4901648259200875e-05,
+      "loss": 1.2738,
+      "step": 7419
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4900290384658181e-05,
+      "loss": 1.2619,
+      "step": 7420
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4898932391199847e-05,
+      "loss": 1.1457,
+      "step": 7421
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4897574278858833e-05,
+      "loss": 1.1419,
+      "step": 7422
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4896216047668097e-05,
+      "loss": 1.0882,
+      "step": 7423
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4894857697660592e-05,
+      "loss": 1.2559,
+      "step": 7424
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.489349922886929e-05,
+      "loss": 1.3286,
+      "step": 7425
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4892140641327151e-05,
+      "loss": 1.1368,
+      "step": 7426
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.489078193506715e-05,
+      "loss": 1.1896,
+      "step": 7427
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4889423110122248e-05,
+      "loss": 1.0527,
+      "step": 7428
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4888064166525431e-05,
+      "loss": 1.1425,
+      "step": 7429
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.488670510430967e-05,
+      "loss": 1.2439,
+      "step": 7430
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4885345923507948e-05,
+      "loss": 1.1496,
+      "step": 7431
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4883986624153246e-05,
+      "loss": 1.119,
+      "step": 7432
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4882627206278555e-05,
+      "loss": 1.3408,
+      "step": 7433
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.488126766991686e-05,
+      "loss": 1.1368,
+      "step": 7434
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4879908015101152e-05,
+      "loss": 1.0964,
+      "step": 7435
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4878548241864428e-05,
+      "loss": 1.0849,
+      "step": 7436
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4877188350239687e-05,
+      "loss": 1.1712,
+      "step": 7437
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4875828340259925e-05,
+      "loss": 1.2295,
+      "step": 7438
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4874468211958152e-05,
+      "loss": 1.2456,
+      "step": 7439
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.487310796536737e-05,
+      "loss": 1.1653,
+      "step": 7440
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4871747600520592e-05,
+      "loss": 1.195,
+      "step": 7441
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4870387117450822e-05,
+      "loss": 1.0777,
+      "step": 7442
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4869026516191082e-05,
+      "loss": 1.1717,
+      "step": 7443
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4867665796774389e-05,
+      "loss": 1.1843,
+      "step": 7444
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4866304959233762e-05,
+      "loss": 1.1029,
+      "step": 7445
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4864944003602222e-05,
+      "loss": 1.0644,
+      "step": 7446
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.48635829299128e-05,
+      "loss": 1.1302,
+      "step": 7447
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4862221738198523e-05,
+      "loss": 1.2722,
+      "step": 7448
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4860860428492427e-05,
+      "loss": 1.1552,
+      "step": 7449
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4859499000827539e-05,
+      "loss": 0.9862,
+      "step": 7450
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4858137455236906e-05,
+      "loss": 1.1482,
+      "step": 7451
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4856775791753561e-05,
+      "loss": 1.3242,
+      "step": 7452
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4855414010410552e-05,
+      "loss": 1.0579,
+      "step": 7453
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4854052111240924e-05,
+      "loss": 1.045,
+      "step": 7454
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4852690094277727e-05,
+      "loss": 1.2645,
+      "step": 7455
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4851327959554012e-05,
+      "loss": 1.1108,
+      "step": 7456
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4849965707102836e-05,
+      "loss": 1.1479,
+      "step": 7457
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4848603336957252e-05,
+      "loss": 1.1348,
+      "step": 7458
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4847240849150328e-05,
+      "loss": 1.2709,
+      "step": 7459
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4845878243715122e-05,
+      "loss": 1.0819,
+      "step": 7460
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4844515520684703e-05,
+      "loss": 1.2014,
+      "step": 7461
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.484315268009214e-05,
+      "loss": 1.2121,
+      "step": 7462
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4841789721970503e-05,
+      "loss": 1.3017,
+      "step": 7463
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.484042664635287e-05,
+      "loss": 1.2122,
+      "step": 7464
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4839063453272318e-05,
+      "loss": 1.2043,
+      "step": 7465
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4837700142761923e-05,
+      "loss": 1.1639,
+      "step": 7466
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4836336714854778e-05,
+      "loss": 1.1285,
+      "step": 7467
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4834973169583962e-05,
+      "loss": 0.9245,
+      "step": 7468
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4833609506982568e-05,
+      "loss": 1.2195,
+      "step": 7469
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4832245727083683e-05,
+      "loss": 1.0293,
+      "step": 7470
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4830881829920407e-05,
+      "loss": 1.2179,
+      "step": 7471
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4829517815525836e-05,
+      "loss": 1.1432,
+      "step": 7472
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4828153683933072e-05,
+      "loss": 1.1913,
+      "step": 7473
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4826789435175215e-05,
+      "loss": 1.1259,
+      "step": 7474
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4825425069285374e-05,
+      "loss": 1.1434,
+      "step": 7475
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4824060586296658e-05,
+      "loss": 1.0652,
+      "step": 7476
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4822695986242178e-05,
+      "loss": 1.0953,
+      "step": 7477
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.482133126915505e-05,
+      "loss": 1.0648,
+      "step": 7478
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4819966435068391e-05,
+      "loss": 1.2093,
+      "step": 7479
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4818601484015323e-05,
+      "loss": 1.2394,
+      "step": 7480
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4817236416028966e-05,
+      "loss": 1.1638,
+      "step": 7481
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.481587123114245e-05,
+      "loss": 1.2349,
+      "step": 7482
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.48145059293889e-05,
+      "loss": 1.1209,
+      "step": 7483
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4813140510801452e-05,
+      "loss": 1.1613,
+      "step": 7484
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4811774975413238e-05,
+      "loss": 1.2295,
+      "step": 7485
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4810409323257396e-05,
+      "loss": 1.2403,
+      "step": 7486
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4809043554367066e-05,
+      "loss": 1.0867,
+      "step": 7487
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4807677668775393e-05,
+      "loss": 1.0433,
+      "step": 7488
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4806311666515523e-05,
+      "loss": 1.1333,
+      "step": 7489
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4804945547620601e-05,
+      "loss": 1.2392,
+      "step": 7490
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4803579312123783e-05,
+      "loss": 1.1809,
+      "step": 7491
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4802212960058222e-05,
+      "loss": 1.1684,
+      "step": 7492
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4800846491457078e-05,
+      "loss": 1.1537,
+      "step": 7493
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4799479906353503e-05,
+      "loss": 1.1081,
+      "step": 7494
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4798113204780668e-05,
+      "loss": 1.0508,
+      "step": 7495
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4796746386771735e-05,
+      "loss": 1.2541,
+      "step": 7496
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4795379452359875e-05,
+      "loss": 1.0697,
+      "step": 7497
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.479401240157826e-05,
+      "loss": 1.0587,
+      "step": 7498
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.479264523446006e-05,
+      "loss": 1.1186,
+      "step": 7499
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4791277951038456e-05,
+      "loss": 1.0121,
+      "step": 7500
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4789910551346625e-05,
+      "loss": 1.2143,
+      "step": 7501
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4788543035417754e-05,
+      "loss": 1.1801,
+      "step": 7502
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4787175403285025e-05,
+      "loss": 1.1279,
+      "step": 7503
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4785807654981627e-05,
+      "loss": 1.0514,
+      "step": 7504
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4784439790540753e-05,
+      "loss": 1.1742,
+      "step": 7505
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4783071809995598e-05,
+      "loss": 1.0115,
+      "step": 7506
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4781703713379352e-05,
+      "loss": 1.1809,
+      "step": 7507
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4780335500725223e-05,
+      "loss": 1.1757,
+      "step": 7508
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4778967172066407e-05,
+      "loss": 1.2768,
+      "step": 7509
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4777598727436115e-05,
+      "loss": 1.163,
+      "step": 7510
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4776230166867549e-05,
+      "loss": 1.0572,
+      "step": 7511
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4774861490393926e-05,
+      "loss": 1.3582,
+      "step": 7512
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4773492698048454e-05,
+      "loss": 1.1785,
+      "step": 7513
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4772123789864356e-05,
+      "loss": 1.1465,
+      "step": 7514
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4770754765874846e-05,
+      "loss": 1.2203,
+      "step": 7515
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4769385626113151e-05,
+      "loss": 1.3192,
+      "step": 7516
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4768016370612488e-05,
+      "loss": 1.0563,
+      "step": 7517
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4766646999406097e-05,
+      "loss": 1.147,
+      "step": 7518
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4765277512527195e-05,
+      "loss": 1.2669,
+      "step": 7519
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.476390791000902e-05,
+      "loss": 1.1833,
+      "step": 7520
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4762538191884815e-05,
+      "loss": 1.3024,
+      "step": 7521
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4761168358187817e-05,
+      "loss": 1.0936,
+      "step": 7522
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4759798408951258e-05,
+      "loss": 1.1584,
+      "step": 7523
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4758428344208395e-05,
+      "loss": 1.1282,
+      "step": 7524
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4757058163992466e-05,
+      "loss": 1.0295,
+      "step": 7525
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.475568786833673e-05,
+      "loss": 1.161,
+      "step": 7526
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4754317457274432e-05,
+      "loss": 1.1467,
+      "step": 7527
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4752946930838834e-05,
+      "loss": 1.1378,
+      "step": 7528
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.475157628906319e-05,
+      "loss": 1.1553,
+      "step": 7529
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4750205531980764e-05,
+      "loss": 1.0696,
+      "step": 7530
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4748834659624817e-05,
+      "loss": 1.0626,
+      "step": 7531
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4747463672028621e-05,
+      "loss": 1.1705,
+      "step": 7532
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4746092569225444e-05,
+      "loss": 1.0149,
+      "step": 7533
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4744721351248558e-05,
+      "loss": 1.1185,
+      "step": 7534
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4743350018131237e-05,
+      "loss": 1.1327,
+      "step": 7535
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4741978569906762e-05,
+      "loss": 1.1523,
+      "step": 7536
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4740607006608412e-05,
+      "loss": 1.0358,
+      "step": 7537
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4739235328269474e-05,
+      "loss": 1.2181,
+      "step": 7538
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.473786353492323e-05,
+      "loss": 1.135,
+      "step": 7539
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.473649162660297e-05,
+      "loss": 0.9886,
+      "step": 7540
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4735119603341991e-05,
+      "loss": 1.1119,
+      "step": 7541
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4733747465173584e-05,
+      "loss": 1.0488,
+      "step": 7542
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4732375212131045e-05,
+      "loss": 1.2219,
+      "step": 7543
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.473100284424768e-05,
+      "loss": 1.1782,
+      "step": 7544
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4729630361556789e-05,
+      "loss": 1.1889,
+      "step": 7545
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4728257764091676e-05,
+      "loss": 1.1458,
+      "step": 7546
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4726885051885654e-05,
+      "loss": 1.1746,
+      "step": 7547
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4725512224972033e-05,
+      "loss": 1.1171,
+      "step": 7548
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4724139283384127e-05,
+      "loss": 1.1809,
+      "step": 7549
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4722766227155256e-05,
+      "loss": 1.067,
+      "step": 7550
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4721393056318736e-05,
+      "loss": 1.2709,
+      "step": 7551
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4720019770907892e-05,
+      "loss": 1.2345,
+      "step": 7552
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4718646370956049e-05,
+      "loss": 1.1253,
+      "step": 7553
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4717272856496537e-05,
+      "loss": 1.1213,
+      "step": 7554
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4715899227562682e-05,
+      "loss": 1.0625,
+      "step": 7555
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4714525484187825e-05,
+      "loss": 1.0762,
+      "step": 7556
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4713151626405297e-05,
+      "loss": 1.1281,
+      "step": 7557
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4711777654248442e-05,
+      "loss": 1.0758,
+      "step": 7558
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.47104035677506e-05,
+      "loss": 1.2563,
+      "step": 7559
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4709029366945115e-05,
+      "loss": 1.2148,
+      "step": 7560
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4707655051865337e-05,
+      "loss": 1.0899,
+      "step": 7561
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4706280622544615e-05,
+      "loss": 1.0683,
+      "step": 7562
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4704906079016304e-05,
+      "loss": 1.2169,
+      "step": 7563
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.470353142131376e-05,
+      "loss": 1.1337,
+      "step": 7564
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4702156649470336e-05,
+      "loss": 1.1141,
+      "step": 7565
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4700781763519403e-05,
+      "loss": 1.2039,
+      "step": 7566
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.469940676349432e-05,
+      "loss": 0.9908,
+      "step": 7567
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4698031649428457e-05,
+      "loss": 1.053,
+      "step": 7568
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.469665642135518e-05,
+      "loss": 1.2407,
+      "step": 7569
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4695281079307865e-05,
+      "loss": 1.201,
+      "step": 7570
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4693905623319886e-05,
+      "loss": 1.0502,
+      "step": 7571
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4692530053424624e-05,
+      "loss": 1.2652,
+      "step": 7572
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4691154369655454e-05,
+      "loss": 1.0819,
+      "step": 7573
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4689778572045767e-05,
+      "loss": 1.1324,
+      "step": 7574
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4688402660628945e-05,
+      "loss": 1.1461,
+      "step": 7575
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4687026635438377e-05,
+      "loss": 1.1764,
+      "step": 7576
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4685650496507458e-05,
+      "loss": 1.1638,
+      "step": 7577
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4684274243869583e-05,
+      "loss": 1.1023,
+      "step": 7578
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4682897877558144e-05,
+      "loss": 1.0496,
+      "step": 7579
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4681521397606551e-05,
+      "loss": 1.1085,
+      "step": 7580
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4680144804048198e-05,
+      "loss": 1.0757,
+      "step": 7581
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4678768096916493e-05,
+      "loss": 1.0793,
+      "step": 7582
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4677391276244847e-05,
+      "loss": 1.1728,
+      "step": 7583
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4676014342066673e-05,
+      "loss": 1.2006,
+      "step": 7584
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.467463729441538e-05,
+      "loss": 1.1096,
+      "step": 7585
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4673260133324388e-05,
+      "loss": 1.1432,
+      "step": 7586
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4671882858827117e-05,
+      "loss": 1.0946,
+      "step": 7587
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4670505470956988e-05,
+      "loss": 1.2483,
+      "step": 7588
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4669127969747427e-05,
+      "loss": 1.0627,
+      "step": 7589
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4667750355231863e-05,
+      "loss": 1.1305,
+      "step": 7590
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4666372627443724e-05,
+      "loss": 1.0397,
+      "step": 7591
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4664994786416442e-05,
+      "loss": 1.278,
+      "step": 7592
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.466361683218346e-05,
+      "loss": 1.1378,
+      "step": 7593
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4662238764778209e-05,
+      "loss": 0.9747,
+      "step": 7594
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4660860584234138e-05,
+      "loss": 1.1414,
+      "step": 7595
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4659482290584683e-05,
+      "loss": 1.1673,
+      "step": 7596
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4658103883863301e-05,
+      "loss": 1.1507,
+      "step": 7597
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4656725364103434e-05,
+      "loss": 1.149,
+      "step": 7598
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4655346731338537e-05,
+      "loss": 1.2217,
+      "step": 7599
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4653967985602066e-05,
+      "loss": 1.2467,
+      "step": 7600
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4652589126927481e-05,
+      "loss": 1.2831,
+      "step": 7601
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4651210155348237e-05,
+      "loss": 1.1302,
+      "step": 7602
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4649831070897804e-05,
+      "loss": 1.1205,
+      "step": 7603
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.464845187360964e-05,
+      "loss": 1.1351,
+      "step": 7604
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4647072563517228e-05,
+      "loss": 1.0955,
+      "step": 7605
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4645693140654028e-05,
+      "loss": 1.2087,
+      "step": 7606
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4644313605053517e-05,
+      "loss": 1.2243,
+      "step": 7607
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4642933956749174e-05,
+      "loss": 1.1951,
+      "step": 7608
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.464155419577448e-05,
+      "loss": 1.1392,
+      "step": 7609
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4640174322162913e-05,
+      "loss": 1.1741,
+      "step": 7610
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4638794335947963e-05,
+      "loss": 1.1928,
+      "step": 7611
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4637414237163116e-05,
+      "loss": 1.2605,
+      "step": 7612
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4636034025841865e-05,
+      "loss": 0.9582,
+      "step": 7613
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.46346537020177e-05,
+      "loss": 1.1822,
+      "step": 7614
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4633273265724121e-05,
+      "loss": 1.2652,
+      "step": 7615
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4631892716994624e-05,
+      "loss": 1.127,
+      "step": 7616
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4630512055862718e-05,
+      "loss": 1.1686,
+      "step": 7617
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4629131282361897e-05,
+      "loss": 1.1028,
+      "step": 7618
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4627750396525675e-05,
+      "loss": 1.2655,
+      "step": 7619
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.462636939838756e-05,
+      "loss": 1.1783,
+      "step": 7620
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4624988287981068e-05,
+      "loss": 1.0494,
+      "step": 7621
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4623607065339709e-05,
+      "loss": 1.1587,
+      "step": 7622
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4622225730497006e-05,
+      "loss": 1.2003,
+      "step": 7623
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4620844283486477e-05,
+      "loss": 1.228,
+      "step": 7624
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4619462724341647e-05,
+      "loss": 1.1529,
+      "step": 7625
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.461808105309604e-05,
+      "loss": 1.1314,
+      "step": 7626
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.461669926978319e-05,
+      "loss": 1.213,
+      "step": 7627
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4615317374436624e-05,
+      "loss": 1.2458,
+      "step": 7628
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4613935367089881e-05,
+      "loss": 1.1569,
+      "step": 7629
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4612553247776493e-05,
+      "loss": 1.1634,
+      "step": 7630
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4611171016530006e-05,
+      "loss": 1.1097,
+      "step": 7631
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4609788673383955e-05,
+      "loss": 1.2648,
+      "step": 7632
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4608406218371894e-05,
+      "loss": 1.2197,
+      "step": 7633
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4607023651527364e-05,
+      "loss": 0.9674,
+      "step": 7634
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4605640972883919e-05,
+      "loss": 1.0026,
+      "step": 7635
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4604258182475112e-05,
+      "loss": 1.1405,
+      "step": 7636
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.46028752803345e-05,
+      "loss": 1.137,
+      "step": 7637
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4601492266495643e-05,
+      "loss": 1.1567,
+      "step": 7638
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4600109140992102e-05,
+      "loss": 1.1404,
+      "step": 7639
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4598725903857438e-05,
+      "loss": 1.1654,
+      "step": 7640
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4597342555125222e-05,
+      "loss": 1.2234,
+      "step": 7641
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4595959094829021e-05,
+      "loss": 1.2061,
+      "step": 7642
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.459457552300241e-05,
+      "loss": 1.2966,
+      "step": 7643
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4593191839678965e-05,
+      "loss": 1.1303,
+      "step": 7644
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4591808044892257e-05,
+      "loss": 1.1243,
+      "step": 7645
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4590424138675874e-05,
+      "loss": 1.1645,
+      "step": 7646
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.45890401210634e-05,
+      "loss": 1.2174,
+      "step": 7647
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4587655992088415e-05,
+      "loss": 1.1093,
+      "step": 7648
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4586271751784513e-05,
+      "loss": 1.048,
+      "step": 7649
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.458488740018528e-05,
+      "loss": 1.0645,
+      "step": 7650
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4583502937324319e-05,
+      "loss": 0.9637,
+      "step": 7651
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4582118363235215e-05,
+      "loss": 1.1262,
+      "step": 7652
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4580733677951579e-05,
+      "loss": 1.1919,
+      "step": 7653
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4579348881507008e-05,
+      "loss": 1.1606,
+      "step": 7654
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4577963973935105e-05,
+      "loss": 1.0629,
+      "step": 7655
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4576578955269479e-05,
+      "loss": 1.2064,
+      "step": 7656
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4575193825543743e-05,
+      "loss": 1.1154,
+      "step": 7657
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4573808584791507e-05,
+      "loss": 1.2868,
+      "step": 7658
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4572423233046386e-05,
+      "loss": 1.156,
+      "step": 7659
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4571037770342003e-05,
+      "loss": 1.2003,
+      "step": 7660
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4569652196711976e-05,
+      "loss": 1.2449,
+      "step": 7661
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4568266512189926e-05,
+      "loss": 1.2273,
+      "step": 7662
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4566880716809487e-05,
+      "loss": 1.1796,
+      "step": 7663
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.456549481060428e-05,
+      "loss": 1.0166,
+      "step": 7664
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4564108793607944e-05,
+      "loss": 1.0557,
+      "step": 7665
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4562722665854105e-05,
+      "loss": 1.3009,
+      "step": 7666
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4561336427376409e-05,
+      "loss": 1.2324,
+      "step": 7667
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4559950078208491e-05,
+      "loss": 1.1261,
+      "step": 7668
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4558563618383996e-05,
+      "loss": 1.1851,
+      "step": 7669
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4557177047936568e-05,
+      "loss": 1.0596,
+      "step": 7670
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4555790366899856e-05,
+      "loss": 1.0692,
+      "step": 7671
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.455440357530751e-05,
+      "loss": 1.1629,
+      "step": 7672
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.455301667319318e-05,
+      "loss": 1.086,
+      "step": 7673
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4551629660590529e-05,
+      "loss": 1.1806,
+      "step": 7674
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.455024253753321e-05,
+      "loss": 1.2288,
+      "step": 7675
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4548855304054888e-05,
+      "loss": 1.1665,
+      "step": 7676
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4547467960189223e-05,
+      "loss": 1.1314,
+      "step": 7677
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4546080505969888e-05,
+      "loss": 1.0404,
+      "step": 7678
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4544692941430546e-05,
+      "loss": 1.3072,
+      "step": 7679
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4543305266604873e-05,
+      "loss": 1.1363,
+      "step": 7680
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.454191748152654e-05,
+      "loss": 1.103,
+      "step": 7681
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4540529586229228e-05,
+      "loss": 1.1828,
+      "step": 7682
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4539141580746617e-05,
+      "loss": 1.106,
+      "step": 7683
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.453775346511239e-05,
+      "loss": 1.1428,
+      "step": 7684
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4536365239360231e-05,
+      "loss": 1.2394,
+      "step": 7685
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4534976903523828e-05,
+      "loss": 1.318,
+      "step": 7686
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4533588457636871e-05,
+      "loss": 1.1333,
+      "step": 7687
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4532199901733054e-05,
+      "loss": 1.0877,
+      "step": 7688
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4530811235846077e-05,
+      "loss": 1.1497,
+      "step": 7689
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4529422460009635e-05,
+      "loss": 1.1471,
+      "step": 7690
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4528033574257432e-05,
+      "loss": 1.1811,
+      "step": 7691
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.452664457862317e-05,
+      "loss": 1.1813,
+      "step": 7692
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4525255473140554e-05,
+      "loss": 1.1118,
+      "step": 7693
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4523866257843299e-05,
+      "loss": 1.2365,
+      "step": 7694
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4522476932765112e-05,
+      "loss": 1.2342,
+      "step": 7695
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4521087497939712e-05,
+      "loss": 1.0432,
+      "step": 7696
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4519697953400813e-05,
+      "loss": 1.2073,
+      "step": 7697
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4518308299182138e-05,
+      "loss": 1.1753,
+      "step": 7698
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4516918535317405e-05,
+      "loss": 1.159,
+      "step": 7699
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4515528661840347e-05,
+      "loss": 1.1862,
+      "step": 7700
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4514138678784687e-05,
+      "loss": 1.2708,
+      "step": 7701
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4512748586184156e-05,
+      "loss": 1.2559,
+      "step": 7702
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4511358384072489e-05,
+      "loss": 1.1153,
+      "step": 7703
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.450996807248342e-05,
+      "loss": 0.9488,
+      "step": 7704
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.450857765145069e-05,
+      "loss": 1.1593,
+      "step": 7705
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4507187121008042e-05,
+      "loss": 1.1595,
+      "step": 7706
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4505796481189214e-05,
+      "loss": 1.1162,
+      "step": 7707
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.450440573202796e-05,
+      "loss": 1.1257,
+      "step": 7708
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4503014873558023e-05,
+      "loss": 0.9695,
+      "step": 7709
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4501623905813161e-05,
+      "loss": 1.2138,
+      "step": 7710
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.450023282882712e-05,
+      "loss": 1.1763,
+      "step": 7711
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.449884164263367e-05,
+      "loss": 1.3329,
+      "step": 7712
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4497450347266559e-05,
+      "loss": 1.2041,
+      "step": 7713
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4496058942759556e-05,
+      "loss": 1.3729,
+      "step": 7714
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4494667429146426e-05,
+      "loss": 1.1906,
+      "step": 7715
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4493275806460934e-05,
+      "loss": 1.1765,
+      "step": 7716
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4491884074736853e-05,
+      "loss": 1.3791,
+      "step": 7717
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4490492234007955e-05,
+      "loss": 1.2515,
+      "step": 7718
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4489100284308018e-05,
+      "loss": 1.1075,
+      "step": 7719
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.448770822567082e-05,
+      "loss": 1.1911,
+      "step": 7720
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4486316058130139e-05,
+      "loss": 1.075,
+      "step": 7721
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4484923781719762e-05,
+      "loss": 1.0523,
+      "step": 7722
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4483531396473474e-05,
+      "loss": 1.1162,
+      "step": 7723
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4482138902425065e-05,
+      "loss": 1.1134,
+      "step": 7724
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4480746299608326e-05,
+      "loss": 1.2051,
+      "step": 7725
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4479353588057052e-05,
+      "loss": 1.1266,
+      "step": 7726
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.447796076780504e-05,
+      "loss": 1.1557,
+      "step": 7727
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4476567838886089e-05,
+      "loss": 1.152,
+      "step": 7728
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4475174801334e-05,
+      "loss": 1.1867,
+      "step": 7729
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4473781655182582e-05,
+      "loss": 0.9376,
+      "step": 7730
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4472388400465639e-05,
+      "loss": 0.994,
+      "step": 7731
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4470995037216983e-05,
+      "loss": 1.2282,
+      "step": 7732
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4469601565470424e-05,
+      "loss": 1.1373,
+      "step": 7733
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4468207985259784e-05,
+      "loss": 1.2435,
+      "step": 7734
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4466814296618875e-05,
+      "loss": 1.1939,
+      "step": 7735
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4465420499581516e-05,
+      "loss": 1.0842,
+      "step": 7736
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4464026594181537e-05,
+      "loss": 1.1713,
+      "step": 7737
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4462632580452763e-05,
+      "loss": 1.1497,
+      "step": 7738
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4461238458429017e-05,
+      "loss": 1.2451,
+      "step": 7739
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4459844228144134e-05,
+      "loss": 1.2206,
+      "step": 7740
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4458449889631946e-05,
+      "loss": 1.1096,
+      "step": 7741
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4457055442926296e-05,
+      "loss": 1.2587,
+      "step": 7742
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4455660888061011e-05,
+      "loss": 1.1916,
+      "step": 7743
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4454266225069946e-05,
+      "loss": 1.1288,
+      "step": 7744
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4452871453986936e-05,
+      "loss": 1.0039,
+      "step": 7745
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4451476574845833e-05,
+      "loss": 0.9981,
+      "step": 7746
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4450081587680485e-05,
+      "loss": 1.2847,
+      "step": 7747
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4448686492524744e-05,
+      "loss": 1.123,
+      "step": 7748
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4447291289412464e-05,
+      "loss": 1.2512,
+      "step": 7749
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4445895978377504e-05,
+      "loss": 1.1347,
+      "step": 7750
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4444500559453722e-05,
+      "loss": 1.1625,
+      "step": 7751
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4443105032674985e-05,
+      "loss": 1.1748,
+      "step": 7752
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4441709398075152e-05,
+      "loss": 1.2257,
+      "step": 7753
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4440313655688096e-05,
+      "loss": 1.0742,
+      "step": 7754
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4438917805547687e-05,
+      "loss": 1.1127,
+      "step": 7755
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4437521847687796e-05,
+      "loss": 1.1032,
+      "step": 7756
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4436125782142303e-05,
+      "loss": 1.1243,
+      "step": 7757
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.443472960894508e-05,
+      "loss": 1.1115,
+      "step": 7758
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4433333328130015e-05,
+      "loss": 1.1265,
+      "step": 7759
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4431936939730984e-05,
+      "loss": 0.9512,
+      "step": 7760
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.443054044378188e-05,
+      "loss": 1.1229,
+      "step": 7761
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4429143840316586e-05,
+      "loss": 1.0156,
+      "step": 7762
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4427747129369002e-05,
+      "loss": 1.179,
+      "step": 7763
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4426350310973011e-05,
+      "loss": 1.2405,
+      "step": 7764
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4424953385162522e-05,
+      "loss": 1.0658,
+      "step": 7765
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4423556351971422e-05,
+      "loss": 1.2168,
+      "step": 7766
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4422159211433622e-05,
+      "loss": 1.1351,
+      "step": 7767
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4420761963583018e-05,
+      "loss": 1.2355,
+      "step": 7768
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4419364608453528e-05,
+      "loss": 0.9837,
+      "step": 7769
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4417967146079053e-05,
+      "loss": 1.0101,
+      "step": 7770
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4416569576493506e-05,
+      "loss": 1.2167,
+      "step": 7771
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4415171899730808e-05,
+      "loss": 1.1644,
+      "step": 7772
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.441377411582487e-05,
+      "loss": 1.062,
+      "step": 7773
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4412376224809614e-05,
+      "loss": 1.1654,
+      "step": 7774
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4410978226718966e-05,
+      "loss": 1.1178,
+      "step": 7775
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4409580121586848e-05,
+      "loss": 1.2637,
+      "step": 7776
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.440818190944719e-05,
+      "loss": 1.1096,
+      "step": 7777
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4406783590333916e-05,
+      "loss": 1.2489,
+      "step": 7778
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.440538516428097e-05,
+      "loss": 1.0287,
+      "step": 7779
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4403986631322278e-05,
+      "loss": 1.0914,
+      "step": 7780
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4402587991491785e-05,
+      "loss": 1.0378,
+      "step": 7781
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4401189244823425e-05,
+      "loss": 1.2687,
+      "step": 7782
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.439979039135115e-05,
+      "loss": 1.1909,
+      "step": 7783
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4398391431108897e-05,
+      "loss": 1.0572,
+      "step": 7784
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4396992364130624e-05,
+      "loss": 1.1923,
+      "step": 7785
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4395593190450275e-05,
+      "loss": 1.0255,
+      "step": 7786
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4394193910101805e-05,
+      "loss": 1.1644,
+      "step": 7787
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4392794523119174e-05,
+      "loss": 1.2251,
+      "step": 7788
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.439139502953634e-05,
+      "loss": 1.238,
+      "step": 7789
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4389995429387258e-05,
+      "loss": 1.2151,
+      "step": 7790
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.43885957227059e-05,
+      "loss": 1.1462,
+      "step": 7791
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.438719590952623e-05,
+      "loss": 1.2566,
+      "step": 7792
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4385795989882221e-05,
+      "loss": 1.1637,
+      "step": 7793
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4384395963807835e-05,
+      "loss": 1.2146,
+      "step": 7794
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4382995831337058e-05,
+      "loss": 1.3844,
+      "step": 7795
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.438159559250386e-05,
+      "loss": 1.126,
+      "step": 7796
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4380195247342226e-05,
+      "loss": 1.107,
+      "step": 7797
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.437879479588613e-05,
+      "loss": 1.1321,
+      "step": 7798
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4377394238169567e-05,
+      "loss": 1.0364,
+      "step": 7799
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4375993574226515e-05,
+      "loss": 1.2839,
+      "step": 7800
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.437459280409097e-05,
+      "loss": 1.1363,
+      "step": 7801
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4373191927796924e-05,
+      "loss": 1.1147,
+      "step": 7802
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4371790945378368e-05,
+      "loss": 1.2328,
+      "step": 7803
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4370389856869304e-05,
+      "loss": 1.0905,
+      "step": 7804
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4368988662303733e-05,
+      "loss": 1.2392,
+      "step": 7805
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4367587361715654e-05,
+      "loss": 1.1618,
+      "step": 7806
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4366185955139077e-05,
+      "loss": 1.2093,
+      "step": 7807
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4364784442608001e-05,
+      "loss": 1.261,
+      "step": 7808
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4363382824156449e-05,
+      "loss": 1.1962,
+      "step": 7809
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4361981099818425e-05,
+      "loss": 1.1397,
+      "step": 7810
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.436057926962795e-05,
+      "loss": 0.9958,
+      "step": 7811
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4359177333619036e-05,
+      "loss": 1.1026,
+      "step": 7812
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4357775291825713e-05,
+      "loss": 1.2511,
+      "step": 7813
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4356373144281996e-05,
+      "loss": 1.2788,
+      "step": 7814
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4354970891021918e-05,
+      "loss": 1.0667,
+      "step": 7815
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.43535685320795e-05,
+      "loss": 1.077,
+      "step": 7816
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.435216606748878e-05,
+      "loss": 1.0387,
+      "step": 7817
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4350763497283789e-05,
+      "loss": 1.0095,
+      "step": 7818
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4349360821498562e-05,
+      "loss": 1.2269,
+      "step": 7819
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4347958040167138e-05,
+      "loss": 1.2517,
+      "step": 7820
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4346555153323561e-05,
+      "loss": 1.1362,
+      "step": 7821
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4345152161001871e-05,
+      "loss": 1.1148,
+      "step": 7822
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4343749063236119e-05,
+      "loss": 1.1546,
+      "step": 7823
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.434234586006035e-05,
+      "loss": 1.1078,
+      "step": 7824
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4340942551508619e-05,
+      "loss": 1.2075,
+      "step": 7825
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4339539137614975e-05,
+      "loss": 1.1415,
+      "step": 7826
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4338135618413484e-05,
+      "loss": 1.1634,
+      "step": 7827
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4336731993938196e-05,
+      "loss": 1.1214,
+      "step": 7828
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4335328264223176e-05,
+      "loss": 1.1345,
+      "step": 7829
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4333924429302489e-05,
+      "loss": 1.0265,
+      "step": 7830
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4332520489210203e-05,
+      "loss": 1.2083,
+      "step": 7831
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4331116443980384e-05,
+      "loss": 1.2569,
+      "step": 7832
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.432971229364711e-05,
+      "loss": 1.2408,
+      "step": 7833
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4328308038244447e-05,
+      "loss": 1.1875,
+      "step": 7834
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4326903677806481e-05,
+      "loss": 1.1837,
+      "step": 7835
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4325499212367287e-05,
+      "loss": 1.2356,
+      "step": 7836
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4324094641960949e-05,
+      "loss": 1.1171,
+      "step": 7837
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4322689966621545e-05,
+      "loss": 1.2295,
+      "step": 7838
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4321285186383173e-05,
+      "loss": 1.0736,
+      "step": 7839
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4319880301279918e-05,
+      "loss": 1.1317,
+      "step": 7840
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4318475311345869e-05,
+      "loss": 1.0541,
+      "step": 7841
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4317070216615127e-05,
+      "loss": 1.0165,
+      "step": 7842
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4315665017121787e-05,
+      "loss": 1.152,
+      "step": 7843
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.431425971289995e-05,
+      "loss": 1.2885,
+      "step": 7844
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4312854303983715e-05,
+      "loss": 1.03,
+      "step": 7845
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4311448790407192e-05,
+      "loss": 1.0043,
+      "step": 7846
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4310043172204484e-05,
+      "loss": 1.1774,
+      "step": 7847
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4308637449409705e-05,
+      "loss": 1.2808,
+      "step": 7848
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4307231622056965e-05,
+      "loss": 1.1887,
+      "step": 7849
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4305825690180383e-05,
+      "loss": 1.0419,
+      "step": 7850
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4304419653814075e-05,
+      "loss": 1.2728,
+      "step": 7851
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.430301351299216e-05,
+      "loss": 1.2352,
+      "step": 7852
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4301607267748759e-05,
+      "loss": 1.2588,
+      "step": 7853
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4300200918118001e-05,
+      "loss": 1.2302,
+      "step": 7854
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4298794464134015e-05,
+      "loss": 1.1518,
+      "step": 7855
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4297387905830928e-05,
+      "loss": 1.1396,
+      "step": 7856
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4295981243242876e-05,
+      "loss": 1.1929,
+      "step": 7857
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4294574476403994e-05,
+      "loss": 1.0472,
+      "step": 7858
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4293167605348417e-05,
+      "loss": 1.1812,
+      "step": 7859
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.429176063011029e-05,
+      "loss": 1.2003,
+      "step": 7860
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.429035355072375e-05,
+      "loss": 0.9483,
+      "step": 7861
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.428894636722295e-05,
+      "loss": 0.8857,
+      "step": 7862
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4287539079642037e-05,
+      "loss": 1.1457,
+      "step": 7863
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.428613168801516e-05,
+      "loss": 0.979,
+      "step": 7864
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4284724192376467e-05,
+      "loss": 1.3167,
+      "step": 7865
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.428331659276012e-05,
+      "loss": 1.0105,
+      "step": 7866
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4281908889200278e-05,
+      "loss": 1.1332,
+      "step": 7867
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4280501081731101e-05,
+      "loss": 1.3033,
+      "step": 7868
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4279093170386748e-05,
+      "loss": 1.2235,
+      "step": 7869
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4277685155201392e-05,
+      "loss": 1.1268,
+      "step": 7870
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4276277036209194e-05,
+      "loss": 1.1814,
+      "step": 7871
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4274868813444332e-05,
+      "loss": 1.1113,
+      "step": 7872
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4273460486940973e-05,
+      "loss": 1.2125,
+      "step": 7873
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4272052056733297e-05,
+      "loss": 1.2071,
+      "step": 7874
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.427064352285548e-05,
+      "loss": 1.231,
+      "step": 7875
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4269234885341705e-05,
+      "loss": 1.2009,
+      "step": 7876
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4267826144226154e-05,
+      "loss": 1.2527,
+      "step": 7877
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4266417299543016e-05,
+      "loss": 1.0802,
+      "step": 7878
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4265008351326475e-05,
+      "loss": 1.178,
+      "step": 7879
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4263599299610725e-05,
+      "loss": 1.2853,
+      "step": 7880
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4262190144429958e-05,
+      "loss": 1.0608,
+      "step": 7881
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.426078088581837e-05,
+      "loss": 1.0604,
+      "step": 7882
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4259371523810162e-05,
+      "loss": 1.2537,
+      "step": 7883
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4257962058439536e-05,
+      "loss": 1.0568,
+      "step": 7884
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.425655248974069e-05,
+      "loss": 1.0233,
+      "step": 7885
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4255142817747833e-05,
+      "loss": 1.1794,
+      "step": 7886
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4253733042495173e-05,
+      "loss": 1.1359,
+      "step": 7887
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4252323164016924e-05,
+      "loss": 1.2789,
+      "step": 7888
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4250913182347294e-05,
+      "loss": 1.0646,
+      "step": 7889
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4249503097520507e-05,
+      "loss": 1.3078,
+      "step": 7890
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4248092909570774e-05,
+      "loss": 1.1845,
+      "step": 7891
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4246682618532321e-05,
+      "loss": 0.9617,
+      "step": 7892
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4245272224439371e-05,
+      "loss": 1.2112,
+      "step": 7893
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4243861727326147e-05,
+      "loss": 1.1991,
+      "step": 7894
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4242451127226877e-05,
+      "loss": 1.1328,
+      "step": 7895
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.42410404241758e-05,
+      "loss": 1.1427,
+      "step": 7896
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.423962961820714e-05,
+      "loss": 1.0734,
+      "step": 7897
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4238218709355141e-05,
+      "loss": 1.0518,
+      "step": 7898
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4236807697654037e-05,
+      "loss": 1.2255,
+      "step": 7899
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.423539658313807e-05,
+      "loss": 1.1672,
+      "step": 7900
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4233985365841482e-05,
+      "loss": 1.1339,
+      "step": 7901
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4232574045798525e-05,
+      "loss": 0.9225,
+      "step": 7902
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4231162623043439e-05,
+      "loss": 1.0988,
+      "step": 7903
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4229751097610483e-05,
+      "loss": 1.253,
+      "step": 7904
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4228339469533902e-05,
+      "loss": 1.2208,
+      "step": 7905
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4226927738847964e-05,
+      "loss": 1.2154,
+      "step": 7906
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4225515905586918e-05,
+      "loss": 1.2952,
+      "step": 7907
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4224103969785026e-05,
+      "loss": 1.2423,
+      "step": 7908
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4222691931476552e-05,
+      "loss": 1.1163,
+      "step": 7909
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4221279790695768e-05,
+      "loss": 1.1377,
+      "step": 7910
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4219867547476936e-05,
+      "loss": 0.9154,
+      "step": 7911
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.421845520185433e-05,
+      "loss": 1.025,
+      "step": 7912
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.421704275386222e-05,
+      "loss": 1.1094,
+      "step": 7913
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4215630203534886e-05,
+      "loss": 1.1275,
+      "step": 7914
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4214217550906602e-05,
+      "loss": 1.0881,
+      "step": 7915
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4212804796011655e-05,
+      "loss": 1.1513,
+      "step": 7916
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4211391938884322e-05,
+      "loss": 1.2587,
+      "step": 7917
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4209978979558893e-05,
+      "loss": 1.0888,
+      "step": 7918
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4208565918069654e-05,
+      "loss": 1.1707,
+      "step": 7919
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4207152754450902e-05,
+      "loss": 1.1746,
+      "step": 7920
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.420573948873692e-05,
+      "loss": 1.1668,
+      "step": 7921
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4204326120962014e-05,
+      "loss": 1.2343,
+      "step": 7922
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4202912651160471e-05,
+      "loss": 1.072,
+      "step": 7923
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4201499079366604e-05,
+      "loss": 1.1348,
+      "step": 7924
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.420008540561471e-05,
+      "loss": 1.0321,
+      "step": 7925
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4198671629939091e-05,
+      "loss": 1.1498,
+      "step": 7926
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4197257752374062e-05,
+      "loss": 1.2178,
+      "step": 7927
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4195843772953932e-05,
+      "loss": 1.1249,
+      "step": 7928
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4194429691713014e-05,
+      "loss": 1.2587,
+      "step": 7929
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4193015508685618e-05,
+      "loss": 1.242,
+      "step": 7930
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4191601223906073e-05,
+      "loss": 1.1782,
+      "step": 7931
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4190186837408688e-05,
+      "loss": 1.2801,
+      "step": 7932
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4188772349227795e-05,
+      "loss": 1.1541,
+      "step": 7933
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4187357759397716e-05,
+      "loss": 1.0933,
+      "step": 7934
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4185943067952776e-05,
+      "loss": 1.2582,
+      "step": 7935
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4184528274927309e-05,
+      "loss": 0.9897,
+      "step": 7936
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4183113380355646e-05,
+      "loss": 1.3571,
+      "step": 7937
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4181698384272127e-05,
+      "loss": 1.1759,
+      "step": 7938
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4180283286711083e-05,
+      "loss": 1.1109,
+      "step": 7939
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4178868087706858e-05,
+      "loss": 1.1884,
+      "step": 7940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4177452787293795e-05,
+      "loss": 1.0866,
+      "step": 7941
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4176037385506238e-05,
+      "loss": 1.2399,
+      "step": 7942
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4174621882378535e-05,
+      "loss": 1.1806,
+      "step": 7943
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4173206277945035e-05,
+      "loss": 1.2599,
+      "step": 7944
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4171790572240092e-05,
+      "loss": 1.0809,
+      "step": 7945
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.417037476529806e-05,
+      "loss": 1.24,
+      "step": 7946
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4168958857153298e-05,
+      "loss": 1.1682,
+      "step": 7947
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4167542847840166e-05,
+      "loss": 1.3312,
+      "step": 7948
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4166126737393024e-05,
+      "loss": 1.1161,
+      "step": 7949
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4164710525846238e-05,
+      "loss": 1.1412,
+      "step": 7950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4163294213234177e-05,
+      "loss": 1.2239,
+      "step": 7951
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4161877799591209e-05,
+      "loss": 1.1595,
+      "step": 7952
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4160461284951705e-05,
+      "loss": 1.1169,
+      "step": 7953
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4159044669350041e-05,
+      "loss": 1.1694,
+      "step": 7954
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4157627952820598e-05,
+      "loss": 1.1272,
+      "step": 7955
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4156211135397748e-05,
+      "loss": 1.1464,
+      "step": 7956
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4154794217115878e-05,
+      "loss": 1.0652,
+      "step": 7957
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.415337719800937e-05,
+      "loss": 1.0139,
+      "step": 7958
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4151960078112614e-05,
+      "loss": 1.2821,
+      "step": 7959
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4150542857459995e-05,
+      "loss": 1.1382,
+      "step": 7960
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4149125536085909e-05,
+      "loss": 1.072,
+      "step": 7961
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4147708114024747e-05,
+      "loss": 1.1586,
+      "step": 7962
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.414629059131091e-05,
+      "loss": 1.3201,
+      "step": 7963
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4144872967978789e-05,
+      "loss": 1.2179,
+      "step": 7964
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4143455244062796e-05,
+      "loss": 1.2594,
+      "step": 7965
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4142037419597326e-05,
+      "loss": 1.08,
+      "step": 7966
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.414061949461679e-05,
+      "loss": 1.2314,
+      "step": 7967
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4139201469155596e-05,
+      "loss": 1.007,
+      "step": 7968
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4137783343248158e-05,
+      "loss": 1.2054,
+      "step": 7969
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4136365116928883e-05,
+      "loss": 1.1866,
+      "step": 7970
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4134946790232193e-05,
+      "loss": 1.161,
+      "step": 7971
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4133528363192506e-05,
+      "loss": 1.2201,
+      "step": 7972
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4132109835844241e-05,
+      "loss": 1.0675,
+      "step": 7973
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.413069120822182e-05,
+      "loss": 1.1343,
+      "step": 7974
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4129272480359674e-05,
+      "loss": 1.0567,
+      "step": 7975
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.412785365229223e-05,
+      "loss": 1.1291,
+      "step": 7976
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4126434724053915e-05,
+      "loss": 1.168,
+      "step": 7977
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4125015695679163e-05,
+      "loss": 1.1858,
+      "step": 7978
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4123596567202413e-05,
+      "loss": 0.9893,
+      "step": 7979
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4122177338658101e-05,
+      "loss": 1.1653,
+      "step": 7980
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.412075801008067e-05,
+      "loss": 1.1823,
+      "step": 7981
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4119338581504559e-05,
+      "loss": 1.154,
+      "step": 7982
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4117919052964216e-05,
+      "loss": 1.225,
+      "step": 7983
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4116499424494085e-05,
+      "loss": 1.1558,
+      "step": 7984
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4115079696128624e-05,
+      "loss": 1.0918,
+      "step": 7985
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4113659867902278e-05,
+      "loss": 1.0376,
+      "step": 7986
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4112239939849507e-05,
+      "loss": 1.1958,
+      "step": 7987
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4110819912004763e-05,
+      "loss": 1.2531,
+      "step": 7988
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4109399784402513e-05,
+      "loss": 0.9425,
+      "step": 7989
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4107979557077212e-05,
+      "loss": 1.1226,
+      "step": 7990
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.410655923006333e-05,
+      "loss": 1.2022,
+      "step": 7991
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4105138803395333e-05,
+      "loss": 1.1655,
+      "step": 7992
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.410371827710769e-05,
+      "loss": 1.3154,
+      "step": 7993
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.410229765123487e-05,
+      "loss": 1.0869,
+      "step": 7994
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4100876925811355e-05,
+      "loss": 1.2448,
+      "step": 7995
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4099456100871613e-05,
+      "loss": 1.2762,
+      "step": 7996
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4098035176450131e-05,
+      "loss": 0.9678,
+      "step": 7997
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4096614152581384e-05,
+      "loss": 1.1303,
+      "step": 7998
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4095193029299863e-05,
+      "loss": 1.1753,
+      "step": 7999
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4093771806640046e-05,
+      "loss": 1.1274,
+      "step": 8000
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4092350484636432e-05,
+      "loss": 1.1199,
+      "step": 8001
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4090929063323502e-05,
+      "loss": 1.1026,
+      "step": 8002
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4089507542735758e-05,
+      "loss": 1.0364,
+      "step": 8003
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4088085922907691e-05,
+      "loss": 1.1461,
+      "step": 8004
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4086664203873803e-05,
+      "loss": 1.1661,
+      "step": 8005
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4085242385668592e-05,
+      "loss": 1.1942,
+      "step": 8006
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4083820468326562e-05,
+      "loss": 1.1107,
+      "step": 8007
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4082398451882218e-05,
+      "loss": 1.1435,
+      "step": 8008
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4080976336370074e-05,
+      "loss": 1.068,
+      "step": 8009
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4079554121824633e-05,
+      "loss": 1.0778,
+      "step": 8010
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4078131808280409e-05,
+      "loss": 1.3311,
+      "step": 8011
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4076709395771921e-05,
+      "loss": 1.1876,
+      "step": 8012
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4075286884333687e-05,
+      "loss": 1.0916,
+      "step": 8013
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4073864274000224e-05,
+      "loss": 1.2208,
+      "step": 8014
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4072441564806055e-05,
+      "loss": 1.1615,
+      "step": 8015
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4071018756785709e-05,
+      "loss": 1.1333,
+      "step": 8016
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4069595849973708e-05,
+      "loss": 1.2853,
+      "step": 8017
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4068172844404584e-05,
+      "loss": 1.0939,
+      "step": 8018
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4066749740112869e-05,
+      "loss": 1.203,
+      "step": 8019
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4065326537133094e-05,
+      "loss": 1.1307,
+      "step": 8020
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4063903235499807e-05,
+      "loss": 1.0186,
+      "step": 8021
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4062479835247533e-05,
+      "loss": 1.0534,
+      "step": 8022
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4061056336410825e-05,
+      "loss": 1.0623,
+      "step": 8023
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4059632739024223e-05,
+      "loss": 1.198,
+      "step": 8024
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4058209043122273e-05,
+      "loss": 1.2277,
+      "step": 8025
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4056785248739524e-05,
+      "loss": 1.1392,
+      "step": 8026
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4055361355910528e-05,
+      "loss": 1.0389,
+      "step": 8027
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4053937364669836e-05,
+      "loss": 1.0382,
+      "step": 8028
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4052513275052008e-05,
+      "loss": 1.0582,
+      "step": 8029
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.40510890870916e-05,
+      "loss": 1.1666,
+      "step": 8030
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4049664800823175e-05,
+      "loss": 1.0603,
+      "step": 8031
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4048240416281296e-05,
+      "loss": 0.9755,
+      "step": 8032
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4046815933500524e-05,
+      "loss": 1.1712,
+      "step": 8033
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4045391352515432e-05,
+      "loss": 1.1136,
+      "step": 8034
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.404396667336059e-05,
+      "loss": 1.1542,
+      "step": 8035
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.404254189607057e-05,
+      "loss": 1.1064,
+      "step": 8036
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4041117020679943e-05,
+      "loss": 1.1167,
+      "step": 8037
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4039692047223297e-05,
+      "loss": 1.2374,
+      "step": 8038
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.40382669757352e-05,
+      "loss": 1.0523,
+      "step": 8039
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4036841806250242e-05,
+      "loss": 1.248,
+      "step": 8040
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4035416538803004e-05,
+      "loss": 1.1135,
+      "step": 8041
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4033991173428077e-05,
+      "loss": 1.2414,
+      "step": 8042
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4032565710160044e-05,
+      "loss": 1.1309,
+      "step": 8043
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4031140149033502e-05,
+      "loss": 1.084,
+      "step": 8044
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4029714490083043e-05,
+      "loss": 1.0689,
+      "step": 8045
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4028288733343267e-05,
+      "loss": 1.0721,
+      "step": 8046
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4026862878848767e-05,
+      "loss": 1.1778,
+      "step": 8047
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.402543692663415e-05,
+      "loss": 1.0576,
+      "step": 8048
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4024010876734016e-05,
+      "loss": 0.9992,
+      "step": 8049
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4022584729182973e-05,
+      "loss": 1.122,
+      "step": 8050
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4021158484015626e-05,
+      "loss": 1.0566,
+      "step": 8051
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4019732141266593e-05,
+      "loss": 1.0675,
+      "step": 8052
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4018305700970479e-05,
+      "loss": 1.097,
+      "step": 8053
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4016879163161906e-05,
+      "loss": 1.2173,
+      "step": 8054
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4015452527875487e-05,
+      "loss": 1.1106,
+      "step": 8055
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4014025795145845e-05,
+      "loss": 1.0261,
+      "step": 8056
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4012598965007598e-05,
+      "loss": 1.2995,
+      "step": 8057
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.401117203749538e-05,
+      "loss": 1.0884,
+      "step": 8058
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4009745012643809e-05,
+      "loss": 1.1216,
+      "step": 8059
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4008317890487522e-05,
+      "loss": 1.0425,
+      "step": 8060
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4006890671061143e-05,
+      "loss": 1.1173,
+      "step": 8061
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4005463354399316e-05,
+      "loss": 1.2184,
+      "step": 8062
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4004035940536671e-05,
+      "loss": 1.0684,
+      "step": 8063
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4002608429507852e-05,
+      "loss": 1.0207,
+      "step": 8064
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4001180821347492e-05,
+      "loss": 1.1277,
+      "step": 8065
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3999753116090244e-05,
+      "loss": 1.1179,
+      "step": 8066
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3998325313770751e-05,
+      "loss": 1.0894,
+      "step": 8067
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3996897414423663e-05,
+      "loss": 1.2225,
+      "step": 8068
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3995469418083625e-05,
+      "loss": 1.2264,
+      "step": 8069
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3994041324785298e-05,
+      "loss": 1.2094,
+      "step": 8070
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3992613134563331e-05,
+      "loss": 1.0491,
+      "step": 8071
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3991184847452388e-05,
+      "loss": 1.0331,
+      "step": 8072
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3989756463487126e-05,
+      "loss": 1.0951,
+      "step": 8073
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3988327982702208e-05,
+      "loss": 1.1234,
+      "step": 8074
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3986899405132297e-05,
+      "loss": 1.1065,
+      "step": 8075
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3985470730812065e-05,
+      "loss": 0.9871,
+      "step": 8076
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3984041959776177e-05,
+      "loss": 1.2094,
+      "step": 8077
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.398261309205931e-05,
+      "loss": 1.2525,
+      "step": 8078
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3981184127696133e-05,
+      "loss": 1.0681,
+      "step": 8079
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3979755066721327e-05,
+      "loss": 1.2082,
+      "step": 8080
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3978325909169568e-05,
+      "loss": 1.0913,
+      "step": 8081
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3976896655075543e-05,
+      "loss": 1.2291,
+      "step": 8082
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3975467304473926e-05,
+      "loss": 1.1363,
+      "step": 8083
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3974037857399414e-05,
+      "loss": 1.2415,
+      "step": 8084
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.397260831388669e-05,
+      "loss": 1.2158,
+      "step": 8085
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3971178673970442e-05,
+      "loss": 1.1276,
+      "step": 8086
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3969748937685368e-05,
+      "loss": 1.0376,
+      "step": 8087
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3968319105066162e-05,
+      "loss": 1.2107,
+      "step": 8088
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.396688917614752e-05,
+      "loss": 1.0835,
+      "step": 8089
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3965459150964146e-05,
+      "loss": 1.1667,
+      "step": 8090
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3964029029550741e-05,
+      "loss": 1.2256,
+      "step": 8091
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3962598811942007e-05,
+      "loss": 1.2514,
+      "step": 8092
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3961168498172655e-05,
+      "loss": 1.2158,
+      "step": 8093
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3959738088277393e-05,
+      "loss": 1.3022,
+      "step": 8094
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3958307582290932e-05,
+      "loss": 1.0252,
+      "step": 8095
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3956876980247988e-05,
+      "loss": 1.3174,
+      "step": 8096
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3955446282183274e-05,
+      "loss": 1.2653,
+      "step": 8097
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3954015488131511e-05,
+      "loss": 1.1436,
+      "step": 8098
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3952584598127425e-05,
+      "loss": 1.1552,
+      "step": 8099
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3951153612205732e-05,
+      "loss": 1.2934,
+      "step": 8100
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3949722530401162e-05,
+      "loss": 1.2523,
+      "step": 8101
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.394829135274844e-05,
+      "loss": 1.0979,
+      "step": 8102
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.39468600792823e-05,
+      "loss": 1.0212,
+      "step": 8103
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3945428710037472e-05,
+      "loss": 1.1214,
+      "step": 8104
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.394399724504869e-05,
+      "loss": 1.2875,
+      "step": 8105
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.39425656843507e-05,
+      "loss": 1.235,
+      "step": 8106
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.394113402797823e-05,
+      "loss": 1.15,
+      "step": 8107
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3939702275966031e-05,
+      "loss": 1.1045,
+      "step": 8108
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3938270428348842e-05,
+      "loss": 1.2461,
+      "step": 8109
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3936838485161413e-05,
+      "loss": 1.2254,
+      "step": 8110
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3935406446438491e-05,
+      "loss": 1.196,
+      "step": 8111
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3933974312214828e-05,
+      "loss": 1.2917,
+      "step": 8112
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3932542082525177e-05,
+      "loss": 0.8955,
+      "step": 8113
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3931109757404299e-05,
+      "loss": 1.162,
+      "step": 8114
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3929677336886945e-05,
+      "loss": 1.1355,
+      "step": 8115
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3928244821007879e-05,
+      "loss": 1.2187,
+      "step": 8116
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3926812209801863e-05,
+      "loss": 1.2517,
+      "step": 8117
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3925379503303662e-05,
+      "loss": 1.1744,
+      "step": 8118
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3923946701548049e-05,
+      "loss": 1.0284,
+      "step": 8119
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3922513804569784e-05,
+      "loss": 1.1763,
+      "step": 8120
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3921080812403648e-05,
+      "loss": 1.0708,
+      "step": 8121
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3919647725084411e-05,
+      "loss": 1.0188,
+      "step": 8122
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3918214542646853e-05,
+      "loss": 1.1627,
+      "step": 8123
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.391678126512575e-05,
+      "loss": 1.1116,
+      "step": 8124
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3915347892555887e-05,
+      "loss": 1.1149,
+      "step": 8125
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.391391442497204e-05,
+      "loss": 1.1549,
+      "step": 8126
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3912480862409004e-05,
+      "loss": 1.2648,
+      "step": 8127
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.391104720490156e-05,
+      "loss": 1.0624,
+      "step": 8128
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3909613452484508e-05,
+      "loss": 1.159,
+      "step": 8129
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3908179605192631e-05,
+      "loss": 1.2539,
+      "step": 8130
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.390674566306073e-05,
+      "loss": 1.1295,
+      "step": 8131
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.39053116261236e-05,
+      "loss": 1.2004,
+      "step": 8132
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3903877494416043e-05,
+      "loss": 1.1079,
+      "step": 8133
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3902443267972858e-05,
+      "loss": 0.9638,
+      "step": 8134
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3901008946828856e-05,
+      "loss": 1.0624,
+      "step": 8135
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3899574531018836e-05,
+      "loss": 0.9943,
+      "step": 8136
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3898140020577611e-05,
+      "loss": 1.1193,
+      "step": 8137
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.389670541553999e-05,
+      "loss": 1.1699,
+      "step": 8138
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3895270715940792e-05,
+      "loss": 1.2,
+      "step": 8139
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3893835921814826e-05,
+      "loss": 1.0911,
+      "step": 8140
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3892401033196916e-05,
+      "loss": 1.0733,
+      "step": 8141
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3890966050121879e-05,
+      "loss": 1.1715,
+      "step": 8142
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.388953097262454e-05,
+      "loss": 1.1505,
+      "step": 8143
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.388809580073972e-05,
+      "loss": 1.2203,
+      "step": 8144
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.388666053450225e-05,
+      "loss": 1.0054,
+      "step": 8145
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3885225173946961e-05,
+      "loss": 1.1636,
+      "step": 8146
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3883789719108682e-05,
+      "loss": 1.284,
+      "step": 8147
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3882354170022247e-05,
+      "loss": 1.1348,
+      "step": 8148
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3880918526722497e-05,
+      "loss": 1.1177,
+      "step": 8149
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3879482789244267e-05,
+      "loss": 1.1758,
+      "step": 8150
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.38780469576224e-05,
+      "loss": 1.1502,
+      "step": 8151
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3876611031891733e-05,
+      "loss": 1.182,
+      "step": 8152
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3875175012087123e-05,
+      "loss": 1.1068,
+      "step": 8153
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3873738898243407e-05,
+      "loss": 1.0746,
+      "step": 8154
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3872302690395445e-05,
+      "loss": 1.1425,
+      "step": 8155
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.387086638857808e-05,
+      "loss": 1.1211,
+      "step": 8156
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3869429992826175e-05,
+      "loss": 1.1571,
+      "step": 8157
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3867993503174584e-05,
+      "loss": 1.2008,
+      "step": 8158
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3866556919658165e-05,
+      "loss": 1.181,
+      "step": 8159
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3865120242311778e-05,
+      "loss": 1.1367,
+      "step": 8160
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3863683471170296e-05,
+      "loss": 1.09,
+      "step": 8161
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3862246606268572e-05,
+      "loss": 1.222,
+      "step": 8162
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3860809647641484e-05,
+      "loss": 1.2005,
+      "step": 8163
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3859372595323902e-05,
+      "loss": 1.1293,
+      "step": 8164
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3857935449350695e-05,
+      "loss": 1.1701,
+      "step": 8165
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3856498209756738e-05,
+      "loss": 1.1517,
+      "step": 8166
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3855060876576915e-05,
+      "loss": 1.1925,
+      "step": 8167
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3853623449846102e-05,
+      "loss": 1.1133,
+      "step": 8168
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3852185929599178e-05,
+      "loss": 1.1366,
+      "step": 8169
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3850748315871031e-05,
+      "loss": 1.1999,
+      "step": 8170
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3849310608696552e-05,
+      "loss": 1.204,
+      "step": 8171
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3847872808110618e-05,
+      "loss": 1.2443,
+      "step": 8172
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.384643491414813e-05,
+      "loss": 1.0997,
+      "step": 8173
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3844996926843979e-05,
+      "loss": 1.1159,
+      "step": 8174
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3843558846233061e-05,
+      "loss": 1.0905,
+      "step": 8175
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.384212067235027e-05,
+      "loss": 1.245,
+      "step": 8176
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3840682405230515e-05,
+      "loss": 1.1522,
+      "step": 8177
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3839244044908689e-05,
+      "loss": 1.2574,
+      "step": 8178
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.38378055914197e-05,
+      "loss": 1.1502,
+      "step": 8179
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3836367044798458e-05,
+      "loss": 1.3372,
+      "step": 8180
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.383492840507987e-05,
+      "loss": 1.1095,
+      "step": 8181
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3833489672298845e-05,
+      "loss": 1.0035,
+      "step": 8182
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3832050846490303e-05,
+      "loss": 1.0802,
+      "step": 8183
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3830611927689157e-05,
+      "loss": 1.1469,
+      "step": 8184
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3829172915930322e-05,
+      "loss": 1.1973,
+      "step": 8185
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3827733811248724e-05,
+      "loss": 1.1022,
+      "step": 8186
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.382629461367928e-05,
+      "loss": 1.0752,
+      "step": 8187
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.382485532325692e-05,
+      "loss": 1.1365,
+      "step": 8188
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3823415940016573e-05,
+      "loss": 1.0086,
+      "step": 8189
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3821976463993161e-05,
+      "loss": 1.0293,
+      "step": 8190
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3820536895221625e-05,
+      "loss": 1.1567,
+      "step": 8191
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3819097233736888e-05,
+      "loss": 1.118,
+      "step": 8192
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.38176574795739e-05,
+      "loss": 1.0901,
+      "step": 8193
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3816217632767588e-05,
+      "loss": 1.2705,
+      "step": 8194
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3814777693352899e-05,
+      "loss": 1.0058,
+      "step": 8195
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3813337661364772e-05,
+      "loss": 1.2349,
+      "step": 8196
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3811897536838155e-05,
+      "loss": 1.192,
+      "step": 8197
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3810457319807996e-05,
+      "loss": 1.2058,
+      "step": 8198
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3809017010309244e-05,
+      "loss": 1.109,
+      "step": 8199
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3807576608376851e-05,
+      "loss": 1.2041,
+      "step": 8200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3806136114045771e-05,
+      "loss": 1.0967,
+      "step": 8201
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3804695527350963e-05,
+      "loss": 1.2496,
+      "step": 8202
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.380325484832738e-05,
+      "loss": 1.0098,
+      "step": 8203
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3801814077009988e-05,
+      "loss": 1.2581,
+      "step": 8204
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3800373213433748e-05,
+      "loss": 1.0612,
+      "step": 8205
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3798932257633631e-05,
+      "loss": 1.0757,
+      "step": 8206
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3797491209644598e-05,
+      "loss": 1.0496,
+      "step": 8207
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3796050069501621e-05,
+      "loss": 1.0889,
+      "step": 8208
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3794608837239671e-05,
+      "loss": 1.1978,
+      "step": 8209
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3793167512893728e-05,
+      "loss": 1.167,
+      "step": 8210
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.379172609649876e-05,
+      "loss": 1.0749,
+      "step": 8211
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3790284588089754e-05,
+      "loss": 1.1029,
+      "step": 8212
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.378884298770169e-05,
+      "loss": 1.2258,
+      "step": 8213
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3787401295369547e-05,
+      "loss": 1.0647,
+      "step": 8214
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3785959511128311e-05,
+      "loss": 1.0211,
+      "step": 8215
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3784517635012974e-05,
+      "loss": 1.0699,
+      "step": 8216
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.378307566705852e-05,
+      "loss": 1.1173,
+      "step": 8217
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3781633607299952e-05,
+      "loss": 1.239,
+      "step": 8218
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3780191455772252e-05,
+      "loss": 1.1002,
+      "step": 8219
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3778749212510426e-05,
+      "loss": 1.1947,
+      "step": 8220
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3777306877549465e-05,
+      "loss": 1.1158,
+      "step": 8221
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3775864450924379e-05,
+      "loss": 1.1413,
+      "step": 8222
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3774421932670166e-05,
+      "loss": 1.1273,
+      "step": 8223
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3772979322821833e-05,
+      "loss": 1.197,
+      "step": 8224
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3771536621414387e-05,
+      "loss": 1.1521,
+      "step": 8225
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.377009382848284e-05,
+      "loss": 1.2088,
+      "step": 8226
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3768650944062203e-05,
+      "loss": 1.2401,
+      "step": 8227
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.376720796818749e-05,
+      "loss": 1.0982,
+      "step": 8228
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3765764900893717e-05,
+      "loss": 1.2412,
+      "step": 8229
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3764321742215907e-05,
+      "loss": 1.1599,
+      "step": 8230
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3762878492189078e-05,
+      "loss": 1.2266,
+      "step": 8231
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3761435150848254e-05,
+      "loss": 1.2132,
+      "step": 8232
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3759991718228461e-05,
+      "loss": 1.0247,
+      "step": 8233
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3758548194364729e-05,
+      "loss": 1.2587,
+      "step": 8234
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3757104579292082e-05,
+      "loss": 1.0521,
+      "step": 8235
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.375566087304556e-05,
+      "loss": 1.1531,
+      "step": 8236
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.375421707566019e-05,
+      "loss": 1.1693,
+      "step": 8237
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3752773187171016e-05,
+      "loss": 1.1453,
+      "step": 8238
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.375132920761307e-05,
+      "loss": 1.2695,
+      "step": 8239
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3749885137021399e-05,
+      "loss": 1.2292,
+      "step": 8240
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3748440975431041e-05,
+      "loss": 1.1312,
+      "step": 8241
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3746996722877048e-05,
+      "loss": 1.0832,
+      "step": 8242
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.374555237939446e-05,
+      "loss": 1.172,
+      "step": 8243
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3744107945018334e-05,
+      "loss": 1.0756,
+      "step": 8244
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3742663419783717e-05,
+      "loss": 1.2065,
+      "step": 8245
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.374121880372567e-05,
+      "loss": 1.118,
+      "step": 8246
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.373977409687924e-05,
+      "loss": 1.0537,
+      "step": 8247
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3738329299279491e-05,
+      "loss": 1.1606,
+      "step": 8248
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3736884410961485e-05,
+      "loss": 1.0,
+      "step": 8249
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3735439431960287e-05,
+      "loss": 1.0257,
+      "step": 8250
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3733994362310957e-05,
+      "loss": 1.0372,
+      "step": 8251
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3732549202048566e-05,
+      "loss": 1.2394,
+      "step": 8252
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.373110395120818e-05,
+      "loss": 1.0487,
+      "step": 8253
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3729658609824879e-05,
+      "loss": 1.1585,
+      "step": 8254
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3728213177933727e-05,
+      "loss": 1.221,
+      "step": 8255
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.372676765556981e-05,
+      "loss": 1.1763,
+      "step": 8256
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3725322042768199e-05,
+      "loss": 1.3115,
+      "step": 8257
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3723876339563979e-05,
+      "loss": 1.217,
+      "step": 8258
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3722430545992232e-05,
+      "loss": 1.1649,
+      "step": 8259
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3720984662088042e-05,
+      "loss": 1.2693,
+      "step": 8260
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3719538687886495e-05,
+      "loss": 1.0888,
+      "step": 8261
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3718092623422688e-05,
+      "loss": 1.309,
+      "step": 8262
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3716646468731703e-05,
+      "loss": 1.0445,
+      "step": 8263
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3715200223848642e-05,
+      "loss": 1.0158,
+      "step": 8264
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3713753888808594e-05,
+      "loss": 1.1197,
+      "step": 8265
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3712307463646664e-05,
+      "loss": 1.0924,
+      "step": 8266
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3710860948397948e-05,
+      "loss": 1.2265,
+      "step": 8267
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.370941434309755e-05,
+      "loss": 0.9848,
+      "step": 8268
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3707967647780576e-05,
+      "loss": 1.0904,
+      "step": 8269
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3706520862482128e-05,
+      "loss": 1.1704,
+      "step": 8270
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3705073987237323e-05,
+      "loss": 1.188,
+      "step": 8271
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3703627022081269e-05,
+      "loss": 1.1795,
+      "step": 8272
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3702179967049076e-05,
+      "loss": 1.3064,
+      "step": 8273
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3700732822175866e-05,
+      "loss": 1.207,
+      "step": 8274
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3699285587496752e-05,
+      "loss": 1.1649,
+      "step": 8275
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3697838263046856e-05,
+      "loss": 1.0308,
+      "step": 8276
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3696390848861301e-05,
+      "loss": 1.007,
+      "step": 8277
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3694943344975214e-05,
+      "loss": 0.9981,
+      "step": 8278
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3693495751423714e-05,
+      "loss": 1.0781,
+      "step": 8279
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3692048068241936e-05,
+      "loss": 1.1807,
+      "step": 8280
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3690600295465006e-05,
+      "loss": 1.1727,
+      "step": 8281
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3689152433128066e-05,
+      "loss": 1.1457,
+      "step": 8282
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3687704481266242e-05,
+      "loss": 1.216,
+      "step": 8283
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3686256439914678e-05,
+      "loss": 1.2447,
+      "step": 8284
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3684808309108508e-05,
+      "loss": 1.1761,
+      "step": 8285
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.368336008888288e-05,
+      "loss": 1.2008,
+      "step": 8286
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.368191177927293e-05,
+      "loss": 1.1017,
+      "step": 8287
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3680463380313818e-05,
+      "loss": 1.0985,
+      "step": 8288
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3679014892040675e-05,
+      "loss": 1.1887,
+      "step": 8289
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3677566314488664e-05,
+      "loss": 1.1149,
+      "step": 8290
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3676117647692934e-05,
+      "loss": 1.1047,
+      "step": 8291
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3674668891688638e-05,
+      "loss": 1.0916,
+      "step": 8292
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3673220046510938e-05,
+      "loss": 1.2614,
+      "step": 8293
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3671771112194986e-05,
+      "loss": 1.1742,
+      "step": 8294
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3670322088775952e-05,
+      "loss": 1.0745,
+      "step": 8295
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3668872976288991e-05,
+      "loss": 1.1628,
+      "step": 8296
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3667423774769275e-05,
+      "loss": 0.857,
+      "step": 8297
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3665974484251966e-05,
+      "loss": 1.0696,
+      "step": 8298
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3664525104772244e-05,
+      "loss": 1.231,
+      "step": 8299
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3663075636365268e-05,
+      "loss": 1.2286,
+      "step": 8300
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3661626079066224e-05,
+      "loss": 1.1057,
+      "step": 8301
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3660176432910279e-05,
+      "loss": 1.189,
+      "step": 8302
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.365872669793262e-05,
+      "loss": 1.0878,
+      "step": 8303
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3657276874168422e-05,
+      "loss": 1.2542,
+      "step": 8304
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3655826961652871e-05,
+      "loss": 1.2235,
+      "step": 8305
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.365437696042115e-05,
+      "loss": 1.294,
+      "step": 8306
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3652926870508449e-05,
+      "loss": 1.3912,
+      "step": 8307
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3651476691949955e-05,
+      "loss": 1.139,
+      "step": 8308
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.365002642478086e-05,
+      "loss": 1.214,
+      "step": 8309
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3648576069036356e-05,
+      "loss": 1.1133,
+      "step": 8310
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3647125624751643e-05,
+      "loss": 1.1918,
+      "step": 8311
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3645675091961914e-05,
+      "loss": 1.0814,
+      "step": 8312
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3644224470702376e-05,
+      "loss": 1.0573,
+      "step": 8313
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3642773761008226e-05,
+      "loss": 1.0337,
+      "step": 8314
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3641322962914669e-05,
+      "loss": 1.2357,
+      "step": 8315
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3639872076456912e-05,
+      "loss": 1.2183,
+      "step": 8316
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3638421101670166e-05,
+      "loss": 1.2032,
+      "step": 8317
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3636970038589642e-05,
+      "loss": 1.1741,
+      "step": 8318
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3635518887250547e-05,
+      "loss": 1.2004,
+      "step": 8319
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.36340676476881e-05,
+      "loss": 1.064,
+      "step": 8320
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3632616319937522e-05,
+      "loss": 1.1842,
+      "step": 8321
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3631164904034027e-05,
+      "loss": 1.1679,
+      "step": 8322
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.362971340001284e-05,
+      "loss": 1.1571,
+      "step": 8323
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3628261807909183e-05,
+      "loss": 1.2245,
+      "step": 8324
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3626810127758283e-05,
+      "loss": 1.1602,
+      "step": 8325
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3625358359595364e-05,
+      "loss": 1.0918,
+      "step": 8326
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3623906503455664e-05,
+      "loss": 1.2482,
+      "step": 8327
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3622454559374405e-05,
+      "loss": 1.0327,
+      "step": 8328
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3621002527386832e-05,
+      "loss": 0.9871,
+      "step": 8329
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3619550407528175e-05,
+      "loss": 1.0523,
+      "step": 8330
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3618098199833675e-05,
+      "loss": 1.0829,
+      "step": 8331
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.361664590433857e-05,
+      "loss": 1.1198,
+      "step": 8332
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3615193521078107e-05,
+      "loss": 1.193,
+      "step": 8333
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3613741050087528e-05,
+      "loss": 0.9471,
+      "step": 8334
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3612288491402084e-05,
+      "loss": 1.2681,
+      "step": 8335
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3610835845057019e-05,
+      "loss": 1.2152,
+      "step": 8336
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3609383111087589e-05,
+      "loss": 1.2184,
+      "step": 8337
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3607930289529044e-05,
+      "loss": 1.1399,
+      "step": 8338
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.360647738041664e-05,
+      "loss": 1.1705,
+      "step": 8339
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.360502438378564e-05,
+      "loss": 1.0747,
+      "step": 8340
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3603571299671297e-05,
+      "loss": 1.0397,
+      "step": 8341
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3602118128108877e-05,
+      "loss": 1.0171,
+      "step": 8342
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3600664869133646e-05,
+      "loss": 1.1779,
+      "step": 8343
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3599211522780865e-05,
+      "loss": 1.2246,
+      "step": 8344
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3597758089085807e-05,
+      "loss": 1.1912,
+      "step": 8345
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3596304568083739e-05,
+      "loss": 1.0751,
+      "step": 8346
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3594850959809937e-05,
+      "loss": 1.1698,
+      "step": 8347
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3593397264299673e-05,
+      "loss": 1.0307,
+      "step": 8348
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3591943481588228e-05,
+      "loss": 1.065,
+      "step": 8349
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3590489611710874e-05,
+      "loss": 1.1177,
+      "step": 8350
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.35890356547029e-05,
+      "loss": 1.2384,
+      "step": 8351
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3587581610599583e-05,
+      "loss": 1.2017,
+      "step": 8352
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3586127479436212e-05,
+      "loss": 1.1274,
+      "step": 8353
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3584673261248071e-05,
+      "loss": 1.0884,
+      "step": 8354
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3583218956070457e-05,
+      "loss": 1.1934,
+      "step": 8355
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3581764563938654e-05,
+      "loss": 1.0565,
+      "step": 8356
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.358031008488796e-05,
+      "loss": 1.1222,
+      "step": 8357
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3578855518953666e-05,
+      "loss": 1.019,
+      "step": 8358
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.357740086617108e-05,
+      "loss": 1.2511,
+      "step": 8359
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.357594612657549e-05,
+      "loss": 1.1541,
+      "step": 8360
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3574491300202206e-05,
+      "loss": 1.1186,
+      "step": 8361
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.357303638708653e-05,
+      "loss": 0.9305,
+      "step": 8362
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.357158138726377e-05,
+      "loss": 1.0974,
+      "step": 8363
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3570126300769233e-05,
+      "loss": 1.1508,
+      "step": 8364
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3568671127638231e-05,
+      "loss": 1.1827,
+      "step": 8365
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3567215867906072e-05,
+      "loss": 1.1422,
+      "step": 8366
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3565760521608078e-05,
+      "loss": 1.247,
+      "step": 8367
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3564305088779561e-05,
+      "loss": 1.0772,
+      "step": 8368
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3562849569455843e-05,
+      "loss": 1.168,
+      "step": 8369
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3561393963672243e-05,
+      "loss": 1.1754,
+      "step": 8370
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3559938271464086e-05,
+      "loss": 1.0918,
+      "step": 8371
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3558482492866696e-05,
+      "loss": 1.0829,
+      "step": 8372
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3557026627915402e-05,
+      "loss": 1.2386,
+      "step": 8373
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3555570676645532e-05,
+      "loss": 0.956,
+      "step": 8374
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3554114639092414e-05,
+      "loss": 1.2328,
+      "step": 8375
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3552658515291392e-05,
+      "loss": 1.26,
+      "step": 8376
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3551202305277795e-05,
+      "loss": 1.1235,
+      "step": 8377
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3549746009086962e-05,
+      "loss": 1.2191,
+      "step": 8378
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.354828962675423e-05,
+      "loss": 1.1466,
+      "step": 8379
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3546833158314944e-05,
+      "loss": 1.0738,
+      "step": 8380
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.354537660380445e-05,
+      "loss": 0.9218,
+      "step": 8381
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3543919963258092e-05,
+      "loss": 1.1526,
+      "step": 8382
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.354246323671122e-05,
+      "loss": 1.0953,
+      "step": 8383
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3541006424199182e-05,
+      "loss": 1.0796,
+      "step": 8384
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.353954952575733e-05,
+      "loss": 1.0577,
+      "step": 8385
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3538092541421022e-05,
+      "loss": 1.1065,
+      "step": 8386
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3536635471225612e-05,
+      "loss": 1.1561,
+      "step": 8387
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3535178315206462e-05,
+      "loss": 1.2934,
+      "step": 8388
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3533721073398932e-05,
+      "loss": 1.0592,
+      "step": 8389
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3532263745838381e-05,
+      "loss": 1.1539,
+      "step": 8390
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3530806332560177e-05,
+      "loss": 1.2386,
+      "step": 8391
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.352934883359969e-05,
+      "loss": 1.1957,
+      "step": 8392
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3527891248992284e-05,
+      "loss": 1.15,
+      "step": 8393
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3526433578773332e-05,
+      "loss": 1.2702,
+      "step": 8394
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3524975822978206e-05,
+      "loss": 1.2018,
+      "step": 8395
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3523517981642287e-05,
+      "loss": 1.3439,
+      "step": 8396
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3522060054800947e-05,
+      "loss": 1.1138,
+      "step": 8397
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.352060204248957e-05,
+      "loss": 1.0881,
+      "step": 8398
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.351914394474353e-05,
+      "loss": 1.1075,
+      "step": 8399
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3517685761598223e-05,
+      "loss": 1.1958,
+      "step": 8400
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3516227493089022e-05,
+      "loss": 1.1116,
+      "step": 8401
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3514769139251325e-05,
+      "loss": 1.105,
+      "step": 8402
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3513310700120514e-05,
+      "loss": 1.2282,
+      "step": 8403
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3511852175731988e-05,
+      "loss": 1.1458,
+      "step": 8404
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3510393566121134e-05,
+      "loss": 1.0656,
+      "step": 8405
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3508934871323353e-05,
+      "loss": 1.1859,
+      "step": 8406
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3507476091374042e-05,
+      "loss": 1.0548,
+      "step": 8407
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3506017226308603e-05,
+      "loss": 1.1936,
+      "step": 8408
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3504558276162434e-05,
+      "loss": 1.0532,
+      "step": 8409
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3503099240970945e-05,
+      "loss": 1.0675,
+      "step": 8410
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3501640120769537e-05,
+      "loss": 1.136,
+      "step": 8411
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3500180915593624e-05,
+      "loss": 1.1894,
+      "step": 8412
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3498721625478611e-05,
+      "loss": 1.158,
+      "step": 8413
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3497262250459917e-05,
+      "loss": 1.1442,
+      "step": 8414
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3495802790572949e-05,
+      "loss": 1.2003,
+      "step": 8415
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3494343245853132e-05,
+      "loss": 1.0941,
+      "step": 8416
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.349288361633588e-05,
+      "loss": 1.0466,
+      "step": 8417
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3491423902056613e-05,
+      "loss": 1.3118,
+      "step": 8418
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3489964103050756e-05,
+      "loss": 1.1906,
+      "step": 8419
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3488504219353736e-05,
+      "loss": 1.1326,
+      "step": 8420
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3487044251000975e-05,
+      "loss": 1.0714,
+      "step": 8421
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3485584198027909e-05,
+      "loss": 1.1463,
+      "step": 8422
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.348412406046996e-05,
+      "loss": 0.9597,
+      "step": 8423
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3482663838362573e-05,
+      "loss": 1.218,
+      "step": 8424
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.348120353174117e-05,
+      "loss": 1.0703,
+      "step": 8425
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.34797431406412e-05,
+      "loss": 1.2145,
+      "step": 8426
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3478282665098095e-05,
+      "loss": 1.2491,
+      "step": 8427
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.34768221051473e-05,
+      "loss": 1.1712,
+      "step": 8428
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3475361460824256e-05,
+      "loss": 1.1398,
+      "step": 8429
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3473900732164414e-05,
+      "loss": 1.1914,
+      "step": 8430
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3472439919203212e-05,
+      "loss": 1.3011,
+      "step": 8431
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3470979021976112e-05,
+      "loss": 1.1716,
+      "step": 8432
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3469518040518552e-05,
+      "loss": 1.2286,
+      "step": 8433
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3468056974865998e-05,
+      "loss": 1.1774,
+      "step": 8434
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3466595825053896e-05,
+      "loss": 1.093,
+      "step": 8435
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3465134591117713e-05,
+      "loss": 1.1893,
+      "step": 8436
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.34636732730929e-05,
+      "loss": 1.1299,
+      "step": 8437
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3462211871014923e-05,
+      "loss": 1.191,
+      "step": 8438
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3460750384919247e-05,
+      "loss": 1.172,
+      "step": 8439
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3459288814841339e-05,
+      "loss": 1.1429,
+      "step": 8440
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3457827160816661e-05,
+      "loss": 1.1934,
+      "step": 8441
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.345636542288069e-05,
+      "loss": 1.2746,
+      "step": 8442
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3454903601068891e-05,
+      "loss": 1.2052,
+      "step": 8443
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3453441695416746e-05,
+      "loss": 1.2039,
+      "step": 8444
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3451979705959729e-05,
+      "loss": 1.2488,
+      "step": 8445
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3450517632733313e-05,
+      "loss": 1.058,
+      "step": 8446
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.344905547577298e-05,
+      "loss": 1.1898,
+      "step": 8447
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3447593235114217e-05,
+      "loss": 1.1551,
+      "step": 8448
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3446130910792506e-05,
+      "loss": 1.142,
+      "step": 8449
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.344466850284333e-05,
+      "loss": 1.1958,
+      "step": 8450
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3443206011302181e-05,
+      "loss": 1.1456,
+      "step": 8451
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3441743436204549e-05,
+      "loss": 1.2961,
+      "step": 8452
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3440280777585923e-05,
+      "loss": 1.1125,
+      "step": 8453
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3438818035481802e-05,
+      "loss": 1.1133,
+      "step": 8454
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.343735520992768e-05,
+      "loss": 1.0853,
+      "step": 8455
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3435892300959057e-05,
+      "loss": 1.1593,
+      "step": 8456
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3434429308611432e-05,
+      "loss": 1.1266,
+      "step": 8457
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3432966232920308e-05,
+      "loss": 1.3343,
+      "step": 8458
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3431503073921189e-05,
+      "loss": 1.306,
+      "step": 8459
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3430039831649579e-05,
+      "loss": 1.0954,
+      "step": 8460
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3428576506140995e-05,
+      "loss": 1.2626,
+      "step": 8461
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.342711309743094e-05,
+      "loss": 1.0467,
+      "step": 8462
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.342564960555493e-05,
+      "loss": 1.1326,
+      "step": 8463
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3424186030548476e-05,
+      "loss": 1.0399,
+      "step": 8464
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3422722372447098e-05,
+      "loss": 1.0932,
+      "step": 8465
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3421258631286315e-05,
+      "loss": 1.2049,
+      "step": 8466
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3419794807101647e-05,
+      "loss": 0.996,
+      "step": 8467
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3418330899928615e-05,
+      "loss": 1.1146,
+      "step": 8468
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3416866909802746e-05,
+      "loss": 1.0329,
+      "step": 8469
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3415402836759563e-05,
+      "loss": 1.2071,
+      "step": 8470
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3413938680834601e-05,
+      "loss": 1.1055,
+      "step": 8471
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3412474442063386e-05,
+      "loss": 1.1532,
+      "step": 8472
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3411010120481453e-05,
+      "loss": 1.1173,
+      "step": 8473
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3409545716124335e-05,
+      "loss": 1.2687,
+      "step": 8474
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.340808122902757e-05,
+      "loss": 1.2721,
+      "step": 8475
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3406616659226695e-05,
+      "loss": 1.2447,
+      "step": 8476
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3405152006757253e-05,
+      "loss": 1.205,
+      "step": 8477
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3403687271654784e-05,
+      "loss": 1.1862,
+      "step": 8478
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3402222453954837e-05,
+      "loss": 1.0029,
+      "step": 8479
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3400757553692955e-05,
+      "loss": 1.0515,
+      "step": 8480
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3399292570904691e-05,
+      "loss": 0.8456,
+      "step": 8481
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.339782750562559e-05,
+      "loss": 1.1409,
+      "step": 8482
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3396362357891212e-05,
+      "loss": 1.2976,
+      "step": 8483
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3394897127737105e-05,
+      "loss": 1.1561,
+      "step": 8484
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3393431815198831e-05,
+      "loss": 1.0408,
+      "step": 8485
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3391966420311943e-05,
+      "loss": 1.1602,
+      "step": 8486
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.339050094311201e-05,
+      "loss": 1.205,
+      "step": 8487
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3389035383634587e-05,
+      "loss": 1.1057,
+      "step": 8488
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3387569741915244e-05,
+      "loss": 1.0814,
+      "step": 8489
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3386104017989543e-05,
+      "loss": 1.0602,
+      "step": 8490
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.338463821189306e-05,
+      "loss": 1.2246,
+      "step": 8491
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3383172323661358e-05,
+      "loss": 1.2018,
+      "step": 8492
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3381706353330015e-05,
+      "loss": 1.0448,
+      "step": 8493
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3380240300934601e-05,
+      "loss": 0.999,
+      "step": 8494
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.33787741665107e-05,
+      "loss": 1.0123,
+      "step": 8495
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3377307950093883e-05,
+      "loss": 1.0734,
+      "step": 8496
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3375841651719739e-05,
+      "loss": 1.2667,
+      "step": 8497
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3374375271423841e-05,
+      "loss": 1.1907,
+      "step": 8498
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3372908809241782e-05,
+      "loss": 1.1266,
+      "step": 8499
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3371442265209145e-05,
+      "loss": 1.2108,
+      "step": 8500
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3369975639361518e-05,
+      "loss": 1.1466,
+      "step": 8501
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3368508931734495e-05,
+      "loss": 0.9289,
+      "step": 8502
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3367042142363666e-05,
+      "loss": 1.2112,
+      "step": 8503
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3365575271284624e-05,
+      "loss": 1.1229,
+      "step": 8504
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.336410831853297e-05,
+      "loss": 1.1808,
+      "step": 8505
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.33626412841443e-05,
+      "loss": 0.965,
+      "step": 8506
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3361174168154216e-05,
+      "loss": 1.159,
+      "step": 8507
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3359706970598319e-05,
+      "loss": 1.0462,
+      "step": 8508
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3358239691512216e-05,
+      "loss": 1.1019,
+      "step": 8509
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.335677233093151e-05,
+      "loss": 1.1935,
+      "step": 8510
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3355304888891813e-05,
+      "loss": 1.1622,
+      "step": 8511
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3353837365428732e-05,
+      "loss": 1.162,
+      "step": 8512
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3352369760577883e-05,
+      "loss": 1.1098,
+      "step": 8513
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3350902074374878e-05,
+      "loss": 1.0739,
+      "step": 8514
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3349434306855335e-05,
+      "loss": 1.1532,
+      "step": 8515
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.334796645805487e-05,
+      "loss": 1.05,
+      "step": 8516
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3346498528009109e-05,
+      "loss": 1.2015,
+      "step": 8517
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3345030516753665e-05,
+      "loss": 1.2051,
+      "step": 8518
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3343562424324172e-05,
+      "loss": 1.0792,
+      "step": 8519
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3342094250756248e-05,
+      "loss": 1.2327,
+      "step": 8520
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.334062599608553e-05,
+      "loss": 1.1682,
+      "step": 8521
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.333915766034764e-05,
+      "loss": 1.1115,
+      "step": 8522
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3337689243578216e-05,
+      "loss": 1.1339,
+      "step": 8523
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3336220745812889e-05,
+      "loss": 0.9826,
+      "step": 8524
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3334752167087295e-05,
+      "loss": 1.107,
+      "step": 8525
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3333283507437074e-05,
+      "loss": 1.096,
+      "step": 8526
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3331814766897865e-05,
+      "loss": 1.2458,
+      "step": 8527
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.333034594550531e-05,
+      "loss": 1.1701,
+      "step": 8528
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3328877043295055e-05,
+      "loss": 1.2082,
+      "step": 8529
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.332740806030274e-05,
+      "loss": 1.1429,
+      "step": 8530
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.332593899656402e-05,
+      "loss": 1.1257,
+      "step": 8531
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3324469852114541e-05,
+      "loss": 1.0823,
+      "step": 8532
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3323000626989957e-05,
+      "loss": 1.3169,
+      "step": 8533
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3321531321225918e-05,
+      "loss": 1.2089,
+      "step": 8534
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3320061934858083e-05,
+      "loss": 1.176,
+      "step": 8535
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.331859246792211e-05,
+      "loss": 0.9728,
+      "step": 8536
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3317122920453655e-05,
+      "loss": 1.2618,
+      "step": 8537
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3315653292488385e-05,
+      "loss": 1.0778,
+      "step": 8538
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3314183584061959e-05,
+      "loss": 1.2139,
+      "step": 8539
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3312713795210041e-05,
+      "loss": 1.1922,
+      "step": 8540
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3311243925968307e-05,
+      "loss": 1.0386,
+      "step": 8541
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3309773976372417e-05,
+      "loss": 0.9565,
+      "step": 8542
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.330830394645805e-05,
+      "loss": 1.2705,
+      "step": 8543
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3306833836260871e-05,
+      "loss": 1.1675,
+      "step": 8544
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3305363645816564e-05,
+      "loss": 1.24,
+      "step": 8545
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3303893375160798e-05,
+      "loss": 1.0547,
+      "step": 8546
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3302423024329259e-05,
+      "loss": 1.1924,
+      "step": 8547
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3300952593357625e-05,
+      "loss": 1.2023,
+      "step": 8548
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3299482082281578e-05,
+      "loss": 1.0706,
+      "step": 8549
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3298011491136808e-05,
+      "loss": 1.2112,
+      "step": 8550
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3296540819958994e-05,
+      "loss": 1.3085,
+      "step": 8551
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3295070068783831e-05,
+      "loss": 1.0711,
+      "step": 8552
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3293599237647006e-05,
+      "loss": 1.1069,
+      "step": 8553
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3292128326584218e-05,
+      "loss": 1.1977,
+      "step": 8554
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3290657335631153e-05,
+      "loss": 1.2657,
+      "step": 8555
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3289186264823514e-05,
+      "loss": 1.0199,
+      "step": 8556
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3287715114196998e-05,
+      "loss": 1.1433,
+      "step": 8557
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3286243883787302e-05,
+      "loss": 1.1033,
+      "step": 8558
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3284772573630133e-05,
+      "loss": 1.2434,
+      "step": 8559
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3283301183761194e-05,
+      "loss": 1.1285,
+      "step": 8560
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3281829714216192e-05,
+      "loss": 1.148,
+      "step": 8561
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3280358165030833e-05,
+      "loss": 1.181,
+      "step": 8562
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3278886536240829e-05,
+      "loss": 1.2073,
+      "step": 8563
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.327741482788189e-05,
+      "loss": 1.048,
+      "step": 8564
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3275943039989735e-05,
+      "loss": 1.1376,
+      "step": 8565
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3274471172600073e-05,
+      "loss": 0.9682,
+      "step": 8566
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3272999225748628e-05,
+      "loss": 1.1233,
+      "step": 8567
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3271527199471116e-05,
+      "loss": 1.0093,
+      "step": 8568
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3270055093803263e-05,
+      "loss": 1.1367,
+      "step": 8569
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3268582908780788e-05,
+      "loss": 1.0602,
+      "step": 8570
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3267110644439418e-05,
+      "loss": 1.084,
+      "step": 8571
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.326563830081488e-05,
+      "loss": 1.1907,
+      "step": 8572
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3264165877942906e-05,
+      "loss": 0.994,
+      "step": 8573
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3262693375859225e-05,
+      "loss": 1.1964,
+      "step": 8574
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3261220794599569e-05,
+      "loss": 1.2906,
+      "step": 8575
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.325974813419968e-05,
+      "loss": 1.1976,
+      "step": 8576
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3258275394695286e-05,
+      "loss": 1.2838,
+      "step": 8577
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3256802576122134e-05,
+      "loss": 1.1934,
+      "step": 8578
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3255329678515959e-05,
+      "loss": 1.1012,
+      "step": 8579
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3253856701912507e-05,
+      "loss": 1.031,
+      "step": 8580
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.325238364634752e-05,
+      "loss": 1.2515,
+      "step": 8581
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.325091051185675e-05,
+      "loss": 1.1763,
+      "step": 8582
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.324943729847594e-05,
+      "loss": 1.1913,
+      "step": 8583
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3247964006240845e-05,
+      "loss": 1.0957,
+      "step": 8584
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3246490635187214e-05,
+      "loss": 1.1346,
+      "step": 8585
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3245017185350803e-05,
+      "loss": 1.1501,
+      "step": 8586
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3243543656767367e-05,
+      "loss": 1.096,
+      "step": 8587
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3242070049472671e-05,
+      "loss": 1.1762,
+      "step": 8588
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3240596363502462e-05,
+      "loss": 1.0311,
+      "step": 8589
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3239122598892516e-05,
+      "loss": 1.1302,
+      "step": 8590
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3237648755678585e-05,
+      "loss": 1.0824,
+      "step": 8591
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3236174833896445e-05,
+      "loss": 1.0156,
+      "step": 8592
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3234700833581855e-05,
+      "loss": 1.299,
+      "step": 8593
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3233226754770592e-05,
+      "loss": 0.9294,
+      "step": 8594
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3231752597498421e-05,
+      "loss": 1.0962,
+      "step": 8595
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.323027836180112e-05,
+      "loss": 1.0484,
+      "step": 8596
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3228804047714462e-05,
+      "loss": 1.1083,
+      "step": 8597
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.322732965527423e-05,
+      "loss": 1.1171,
+      "step": 8598
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3225855184516194e-05,
+      "loss": 1.1084,
+      "step": 8599
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3224380635476142e-05,
+      "loss": 1.1141,
+      "step": 8600
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3222906008189851e-05,
+      "loss": 1.0641,
+      "step": 8601
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3221431302693116e-05,
+      "loss": 1.0659,
+      "step": 8602
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.321995651902171e-05,
+      "loss": 1.1423,
+      "step": 8603
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3218481657211431e-05,
+      "loss": 1.0503,
+      "step": 8604
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3217006717298066e-05,
+      "loss": 1.1475,
+      "step": 8605
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3215531699317412e-05,
+      "loss": 1.1474,
+      "step": 8606
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3214056603305258e-05,
+      "loss": 1.2278,
+      "step": 8607
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3212581429297407e-05,
+      "loss": 1.2261,
+      "step": 8608
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3211106177329647e-05,
+      "loss": 1.1556,
+      "step": 8609
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3209630847437787e-05,
+      "loss": 1.0348,
+      "step": 8610
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3208155439657624e-05,
+      "loss": 1.1731,
+      "step": 8611
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3206679954024963e-05,
+      "loss": 1.2822,
+      "step": 8612
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.320520439057561e-05,
+      "loss": 1.0438,
+      "step": 8613
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3203728749345375e-05,
+      "loss": 0.8586,
+      "step": 8614
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3202253030370062e-05,
+      "loss": 1.2854,
+      "step": 8615
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3200777233685488e-05,
+      "loss": 1.2311,
+      "step": 8616
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3199301359327462e-05,
+      "loss": 1.2345,
+      "step": 8617
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3197825407331801e-05,
+      "loss": 1.1622,
+      "step": 8618
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3196349377734321e-05,
+      "loss": 1.1683,
+      "step": 8619
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3194873270570845e-05,
+      "loss": 1.1196,
+      "step": 8620
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3193397085877186e-05,
+      "loss": 1.171,
+      "step": 8621
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3191920823689178e-05,
+      "loss": 1.1446,
+      "step": 8622
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3190444484042631e-05,
+      "loss": 1.2391,
+      "step": 8623
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3188968066973385e-05,
+      "loss": 1.0571,
+      "step": 8624
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3187491572517257e-05,
+      "loss": 1.1026,
+      "step": 8625
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3186015000710086e-05,
+      "loss": 1.2234,
+      "step": 8626
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.31845383515877e-05,
+      "loss": 1.238,
+      "step": 8627
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3183061625185936e-05,
+      "loss": 1.1951,
+      "step": 8628
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3181584821540622e-05,
+      "loss": 1.1431,
+      "step": 8629
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3180107940687607e-05,
+      "loss": 1.1919,
+      "step": 8630
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.317863098266272e-05,
+      "loss": 1.2114,
+      "step": 8631
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3177153947501808e-05,
+      "loss": 1.2132,
+      "step": 8632
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3175676835240716e-05,
+      "loss": 1.1931,
+      "step": 8633
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3174199645915283e-05,
+      "loss": 1.1867,
+      "step": 8634
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3172722379561363e-05,
+      "loss": 1.1718,
+      "step": 8635
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3171245036214799e-05,
+      "loss": 1.2054,
+      "step": 8636
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3169767615911446e-05,
+      "loss": 1.2538,
+      "step": 8637
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3168290118687153e-05,
+      "loss": 1.0733,
+      "step": 8638
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3166812544577779e-05,
+      "loss": 1.2516,
+      "step": 8639
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3165334893619175e-05,
+      "loss": 1.0398,
+      "step": 8640
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3163857165847205e-05,
+      "loss": 1.1199,
+      "step": 8641
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3162379361297721e-05,
+      "loss": 1.0134,
+      "step": 8642
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3160901480006596e-05,
+      "loss": 1.355,
+      "step": 8643
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3159423522009684e-05,
+      "loss": 0.9416,
+      "step": 8644
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3157945487342855e-05,
+      "loss": 1.1091,
+      "step": 8645
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.315646737604198e-05,
+      "loss": 1.1705,
+      "step": 8646
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3154989188142921e-05,
+      "loss": 1.0836,
+      "step": 8647
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3153510923681554e-05,
+      "loss": 1.2411,
+      "step": 8648
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.315203258269375e-05,
+      "loss": 1.1506,
+      "step": 8649
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3150554165215387e-05,
+      "loss": 1.0395,
+      "step": 8650
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.314907567128234e-05,
+      "loss": 1.0525,
+      "step": 8651
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3147597100930482e-05,
+      "loss": 1.1921,
+      "step": 8652
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3146118454195706e-05,
+      "loss": 1.3208,
+      "step": 8653
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3144639731113884e-05,
+      "loss": 1.2364,
+      "step": 8654
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3143160931720906e-05,
+      "loss": 1.1606,
+      "step": 8655
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3141682056052653e-05,
+      "loss": 1.0036,
+      "step": 8656
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.314020310414502e-05,
+      "loss": 1.1991,
+      "step": 8657
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3138724076033888e-05,
+      "loss": 1.0536,
+      "step": 8658
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3137244971755156e-05,
+      "loss": 1.1731,
+      "step": 8659
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3135765791344715e-05,
+      "loss": 1.1915,
+      "step": 8660
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3134286534838459e-05,
+      "loss": 1.2439,
+      "step": 8661
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3132807202272286e-05,
+      "loss": 1.1812,
+      "step": 8662
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3131327793682099e-05,
+      "loss": 1.0634,
+      "step": 8663
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.312984830910379e-05,
+      "loss": 1.0988,
+      "step": 8664
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3128368748573272e-05,
+      "loss": 1.1294,
+      "step": 8665
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3126889112126442e-05,
+      "loss": 1.192,
+      "step": 8666
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.312540939979921e-05,
+      "loss": 1.2995,
+      "step": 8667
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3123929611627481e-05,
+      "loss": 1.2461,
+      "step": 8668
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3122449747647172e-05,
+      "loss": 1.1811,
+      "step": 8669
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3120969807894186e-05,
+      "loss": 1.1027,
+      "step": 8670
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3119489792404445e-05,
+      "loss": 1.1773,
+      "step": 8671
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3118009701213859e-05,
+      "loss": 1.1247,
+      "step": 8672
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3116529534358346e-05,
+      "loss": 1.2275,
+      "step": 8673
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3115049291873826e-05,
+      "loss": 1.037,
+      "step": 8674
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3113568973796224e-05,
+      "loss": 1.0985,
+      "step": 8675
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.311208858016146e-05,
+      "loss": 1.1346,
+      "step": 8676
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3110608111005456e-05,
+      "loss": 1.0098,
+      "step": 8677
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3109127566364142e-05,
+      "loss": 1.1884,
+      "step": 8678
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3107646946273444e-05,
+      "loss": 1.1696,
+      "step": 8679
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3106166250769297e-05,
+      "loss": 1.2134,
+      "step": 8680
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3104685479887628e-05,
+      "loss": 1.082,
+      "step": 8681
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3103204633664373e-05,
+      "loss": 1.0275,
+      "step": 8682
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3101723712135468e-05,
+      "loss": 1.2335,
+      "step": 8683
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3100242715336851e-05,
+      "loss": 1.2095,
+      "step": 8684
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3098761643304461e-05,
+      "loss": 1.2478,
+      "step": 8685
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3097280496074237e-05,
+      "loss": 1.2455,
+      "step": 8686
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3095799273682124e-05,
+      "loss": 1.178,
+      "step": 8687
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3094317976164066e-05,
+      "loss": 1.2546,
+      "step": 8688
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3092836603556016e-05,
+      "loss": 1.3144,
+      "step": 8689
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3091355155893912e-05,
+      "loss": 1.2032,
+      "step": 8690
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3089873633213714e-05,
+      "loss": 1.0782,
+      "step": 8691
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3088392035551367e-05,
+      "loss": 1.054,
+      "step": 8692
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3086910362942829e-05,
+      "loss": 1.2505,
+      "step": 8693
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3085428615424055e-05,
+      "loss": 1.2583,
+      "step": 8694
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3083946793031003e-05,
+      "loss": 1.1983,
+      "step": 8695
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.308246489579963e-05,
+      "loss": 1.0242,
+      "step": 8696
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3080982923765902e-05,
+      "loss": 1.1073,
+      "step": 8697
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3079500876965776e-05,
+      "loss": 1.1303,
+      "step": 8698
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3078018755435223e-05,
+      "loss": 1.1792,
+      "step": 8699
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3076536559210204e-05,
+      "loss": 1.0185,
+      "step": 8700
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3075054288326694e-05,
+      "loss": 0.962,
+      "step": 8701
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3073571942820658e-05,
+      "loss": 1.1075,
+      "step": 8702
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3072089522728069e-05,
+      "loss": 1.1961,
+      "step": 8703
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3070607028084903e-05,
+      "loss": 1.0508,
+      "step": 8704
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3069124458927136e-05,
+      "loss": 1.2168,
+      "step": 8705
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3067641815290742e-05,
+      "loss": 1.1812,
+      "step": 8706
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3066159097211703e-05,
+      "loss": 1.1686,
+      "step": 8707
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3064676304726001e-05,
+      "loss": 1.1854,
+      "step": 8708
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.306319343786962e-05,
+      "loss": 1.1506,
+      "step": 8709
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.306171049667854e-05,
+      "loss": 1.2482,
+      "step": 8710
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.306022748118875e-05,
+      "loss": 1.0211,
+      "step": 8711
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.305874439143624e-05,
+      "loss": 1.2058,
+      "step": 8712
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3057261227457e-05,
+      "loss": 1.2047,
+      "step": 8713
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3055777989287018e-05,
+      "loss": 1.1464,
+      "step": 8714
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3054294676962294e-05,
+      "loss": 1.074,
+      "step": 8715
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.305281129051882e-05,
+      "loss": 0.9262,
+      "step": 8716
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3051327829992593e-05,
+      "loss": 1.1189,
+      "step": 8717
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3049844295419614e-05,
+      "loss": 1.0736,
+      "step": 8718
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3048360686835882e-05,
+      "loss": 1.1517,
+      "step": 8719
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.30468770042774e-05,
+      "loss": 1.1335,
+      "step": 8720
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3045393247780176e-05,
+      "loss": 1.1862,
+      "step": 8721
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3043909417380213e-05,
+      "loss": 0.9541,
+      "step": 8722
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.304242551311352e-05,
+      "loss": 1.2785,
+      "step": 8723
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.304094153501611e-05,
+      "loss": 1.0214,
+      "step": 8724
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.303945748312399e-05,
+      "loss": 1.1511,
+      "step": 8725
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3037973357473175e-05,
+      "loss": 1.2533,
+      "step": 8726
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3036489158099682e-05,
+      "loss": 0.9548,
+      "step": 8727
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3035004885039526e-05,
+      "loss": 1.2882,
+      "step": 8728
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3033520538328728e-05,
+      "loss": 1.211,
+      "step": 8729
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.303203611800331e-05,
+      "loss": 1.1953,
+      "step": 8730
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3030551624099288e-05,
+      "loss": 1.0872,
+      "step": 8731
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.302906705665269e-05,
+      "loss": 1.1197,
+      "step": 8732
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3027582415699547e-05,
+      "loss": 1.0112,
+      "step": 8733
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3026097701275881e-05,
+      "loss": 1.2256,
+      "step": 8734
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3024612913417725e-05,
+      "loss": 1.1527,
+      "step": 8735
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3023128052161104e-05,
+      "loss": 1.0973,
+      "step": 8736
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.302164311754206e-05,
+      "loss": 1.2123,
+      "step": 8737
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3020158109596622e-05,
+      "loss": 1.2667,
+      "step": 8738
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3018673028360829e-05,
+      "loss": 1.1525,
+      "step": 8739
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3017187873870721e-05,
+      "loss": 0.98,
+      "step": 8740
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3015702646162333e-05,
+      "loss": 1.2096,
+      "step": 8741
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3014217345271713e-05,
+      "loss": 1.227,
+      "step": 8742
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3012731971234903e-05,
+      "loss": 1.2886,
+      "step": 8743
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3011246524087946e-05,
+      "loss": 1.2397,
+      "step": 8744
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3009761003866892e-05,
+      "loss": 1.1793,
+      "step": 8745
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3008275410607792e-05,
+      "loss": 1.124,
+      "step": 8746
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3006789744346693e-05,
+      "loss": 1.1947,
+      "step": 8747
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.300530400511965e-05,
+      "loss": 1.2333,
+      "step": 8748
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3003818192962716e-05,
+      "loss": 1.0333,
+      "step": 8749
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3002332307911952e-05,
+      "loss": 1.0156,
+      "step": 8750
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.300084635000341e-05,
+      "loss": 0.9691,
+      "step": 8751
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.299936031927315e-05,
+      "loss": 1.0733,
+      "step": 8752
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2997874215757237e-05,
+      "loss": 1.1107,
+      "step": 8753
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2996388039491736e-05,
+      "loss": 1.175,
+      "step": 8754
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2994901790512708e-05,
+      "loss": 1.0943,
+      "step": 8755
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.299341546885622e-05,
+      "loss": 1.1065,
+      "step": 8756
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2991929074558343e-05,
+      "loss": 1.2132,
+      "step": 8757
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2990442607655146e-05,
+      "loss": 1.2521,
+      "step": 8758
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2988956068182702e-05,
+      "loss": 1.21,
+      "step": 8759
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2987469456177084e-05,
+      "loss": 1.0906,
+      "step": 8760
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2985982771674369e-05,
+      "loss": 1.274,
+      "step": 8761
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2984496014710633e-05,
+      "loss": 1.2527,
+      "step": 8762
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2983009185321953e-05,
+      "loss": 1.3385,
+      "step": 8763
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2981522283544416e-05,
+      "loss": 1.2306,
+      "step": 8764
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.29800353094141e-05,
+      "loss": 1.0483,
+      "step": 8765
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2978548262967095e-05,
+      "loss": 1.2754,
+      "step": 8766
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2977061144239479e-05,
+      "loss": 1.1348,
+      "step": 8767
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2975573953267345e-05,
+      "loss": 1.2769,
+      "step": 8768
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.297408669008678e-05,
+      "loss": 0.9807,
+      "step": 8769
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2972599354733879e-05,
+      "loss": 1.017,
+      "step": 8770
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2971111947244735e-05,
+      "loss": 1.2251,
+      "step": 8771
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2969624467655441e-05,
+      "loss": 1.1712,
+      "step": 8772
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2968136916002094e-05,
+      "loss": 1.2395,
+      "step": 8773
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2966649292320797e-05,
+      "loss": 1.0987,
+      "step": 8774
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2965161596647639e-05,
+      "loss": 1.1914,
+      "step": 8775
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2963673829018735e-05,
+      "loss": 1.1284,
+      "step": 8776
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2962185989470182e-05,
+      "loss": 1.2504,
+      "step": 8777
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2960698078038085e-05,
+      "loss": 1.2298,
+      "step": 8778
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2959210094758554e-05,
+      "loss": 1.1471,
+      "step": 8779
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2957722039667698e-05,
+      "loss": 1.0381,
+      "step": 8780
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2956233912801623e-05,
+      "loss": 1.1641,
+      "step": 8781
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2954745714196448e-05,
+      "loss": 1.1449,
+      "step": 8782
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2953257443888281e-05,
+      "loss": 1.2716,
+      "step": 8783
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2951769101913245e-05,
+      "loss": 1.2488,
+      "step": 8784
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.295028068830745e-05,
+      "loss": 1.141,
+      "step": 8785
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2948792203107022e-05,
+      "loss": 1.2047,
+      "step": 8786
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2947303646348078e-05,
+      "loss": 1.1496,
+      "step": 8787
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2945815018066743e-05,
+      "loss": 1.2404,
+      "step": 8788
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2944326318299139e-05,
+      "loss": 1.0589,
+      "step": 8789
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2942837547081396e-05,
+      "loss": 1.2085,
+      "step": 8790
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2941348704449639e-05,
+      "loss": 1.2538,
+      "step": 8791
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2939859790440002e-05,
+      "loss": 1.1642,
+      "step": 8792
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2938370805088608e-05,
+      "loss": 1.1024,
+      "step": 8793
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2936881748431601e-05,
+      "loss": 1.1526,
+      "step": 8794
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.293539262050511e-05,
+      "loss": 1.0197,
+      "step": 8795
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2933903421345272e-05,
+      "loss": 1.1439,
+      "step": 8796
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2932414150988226e-05,
+      "loss": 1.2483,
+      "step": 8797
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2930924809470115e-05,
+      "loss": 1.097,
+      "step": 8798
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2929435396827076e-05,
+      "loss": 1.1149,
+      "step": 8799
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2927945913095256e-05,
+      "loss": 0.9846,
+      "step": 8800
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2926456358310797e-05,
+      "loss": 1.0647,
+      "step": 8801
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2924966732509854e-05,
+      "loss": 1.0965,
+      "step": 8802
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2923477035728564e-05,
+      "loss": 1.3419,
+      "step": 8803
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2921987268003087e-05,
+      "loss": 1.3034,
+      "step": 8804
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2920497429369574e-05,
+      "loss": 1.1575,
+      "step": 8805
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2919007519864173e-05,
+      "loss": 1.1784,
+      "step": 8806
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2917517539523048e-05,
+      "loss": 1.13,
+      "step": 8807
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.291602748838235e-05,
+      "loss": 1.094,
+      "step": 8808
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2914537366478243e-05,
+      "loss": 1.054,
+      "step": 8809
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2913047173846884e-05,
+      "loss": 1.0492,
+      "step": 8810
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2911556910524436e-05,
+      "loss": 1.0897,
+      "step": 8811
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2910066576547067e-05,
+      "loss": 1.1656,
+      "step": 8812
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2908576171950937e-05,
+      "loss": 1.0802,
+      "step": 8813
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2907085696772219e-05,
+      "loss": 1.0894,
+      "step": 8814
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.290559515104708e-05,
+      "loss": 1.267,
+      "step": 8815
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2904104534811692e-05,
+      "loss": 1.1747,
+      "step": 8816
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2902613848102228e-05,
+      "loss": 1.1168,
+      "step": 8817
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2901123090954864e-05,
+      "loss": 1.2145,
+      "step": 8818
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.289963226340577e-05,
+      "loss": 1.1283,
+      "step": 8819
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2898141365491132e-05,
+      "loss": 0.942,
+      "step": 8820
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2896650397247125e-05,
+      "loss": 1.4168,
+      "step": 8821
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2895159358709933e-05,
+      "loss": 1.1872,
+      "step": 8822
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2893668249915737e-05,
+      "loss": 1.2225,
+      "step": 8823
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.289217707090072e-05,
+      "loss": 1.0641,
+      "step": 8824
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2890685821701074e-05,
+      "loss": 1.1359,
+      "step": 8825
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2889194502352981e-05,
+      "loss": 1.1429,
+      "step": 8826
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2887703112892638e-05,
+      "loss": 1.1204,
+      "step": 8827
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.288621165335623e-05,
+      "loss": 1.1986,
+      "step": 8828
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2884720123779957e-05,
+      "loss": 1.2211,
+      "step": 8829
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2883228524200006e-05,
+      "loss": 1.027,
+      "step": 8830
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2881736854652582e-05,
+      "loss": 1.0349,
+      "step": 8831
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2880245115173877e-05,
+      "loss": 1.2278,
+      "step": 8832
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2878753305800094e-05,
+      "loss": 1.184,
+      "step": 8833
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2877261426567433e-05,
+      "loss": 1.0642,
+      "step": 8834
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2875769477512101e-05,
+      "loss": 0.9961,
+      "step": 8835
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2874277458670301e-05,
+      "loss": 1.1753,
+      "step": 8836
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.287278537007824e-05,
+      "loss": 1.0914,
+      "step": 8837
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2871293211772124e-05,
+      "loss": 1.0988,
+      "step": 8838
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.286980098378817e-05,
+      "loss": 1.0835,
+      "step": 8839
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2868308686162583e-05,
+      "loss": 0.9981,
+      "step": 8840
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2866816318931582e-05,
+      "loss": 1.122,
+      "step": 8841
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2865323882131377e-05,
+      "loss": 1.046,
+      "step": 8842
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.286383137579819e-05,
+      "loss": 1.095,
+      "step": 8843
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2862338799968237e-05,
+      "loss": 0.9908,
+      "step": 8844
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.286084615467774e-05,
+      "loss": 1.3048,
+      "step": 8845
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2859353439962918e-05,
+      "loss": 1.1687,
+      "step": 8846
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.285786065586e-05,
+      "loss": 1.2407,
+      "step": 8847
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2856367802405205e-05,
+      "loss": 1.1381,
+      "step": 8848
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.285487487963477e-05,
+      "loss": 1.0643,
+      "step": 8849
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.285338188758491e-05,
+      "loss": 1.3291,
+      "step": 8850
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.285188882629187e-05,
+      "loss": 1.1721,
+      "step": 8851
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2850395695791868e-05,
+      "loss": 1.1506,
+      "step": 8852
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2848902496121152e-05,
+      "loss": 1.2149,
+      "step": 8853
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2847409227315943e-05,
+      "loss": 1.1122,
+      "step": 8854
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2845915889412492e-05,
+      "loss": 1.2104,
+      "step": 8855
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.284442248244703e-05,
+      "loss": 1.1463,
+      "step": 8856
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2842929006455802e-05,
+      "loss": 1.2487,
+      "step": 8857
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2841435461475045e-05,
+      "loss": 1.0981,
+      "step": 8858
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2839941847541007e-05,
+      "loss": 1.1039,
+      "step": 8859
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2838448164689932e-05,
+      "loss": 1.1059,
+      "step": 8860
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2836954412958067e-05,
+      "loss": 0.9732,
+      "step": 8861
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2835460592381662e-05,
+      "loss": 1.1602,
+      "step": 8862
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2833966702996968e-05,
+      "loss": 1.0571,
+      "step": 8863
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2832472744840234e-05,
+      "loss": 1.0503,
+      "step": 8864
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.283097871794772e-05,
+      "loss": 0.9463,
+      "step": 8865
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2829484622355675e-05,
+      "loss": 1.0919,
+      "step": 8866
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.282799045810036e-05,
+      "loss": 1.1775,
+      "step": 8867
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2826496225218035e-05,
+      "loss": 1.1198,
+      "step": 8868
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.282500192374496e-05,
+      "loss": 1.0978,
+      "step": 8869
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2823507553717394e-05,
+      "loss": 1.1383,
+      "step": 8870
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2822013115171603e-05,
+      "loss": 1.1682,
+      "step": 8871
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2820518608143855e-05,
+      "loss": 1.073,
+      "step": 8872
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2819024032670416e-05,
+      "loss": 1.1753,
+      "step": 8873
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2817529388787548e-05,
+      "loss": 1.275,
+      "step": 8874
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2816034676531534e-05,
+      "loss": 1.0726,
+      "step": 8875
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2814539895938638e-05,
+      "loss": 1.1569,
+      "step": 8876
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2813045047045139e-05,
+      "loss": 0.9832,
+      "step": 8877
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2811550129887304e-05,
+      "loss": 1.1794,
+      "step": 8878
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2810055144501422e-05,
+      "loss": 1.198,
+      "step": 8879
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.280856009092376e-05,
+      "loss": 1.1718,
+      "step": 8880
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2807064969190605e-05,
+      "loss": 1.2633,
+      "step": 8881
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.280556977933824e-05,
+      "loss": 1.3034,
+      "step": 8882
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2804074521402949e-05,
+      "loss": 1.2694,
+      "step": 8883
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2802579195421014e-05,
+      "loss": 1.2246,
+      "step": 8884
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2801083801428724e-05,
+      "loss": 1.1273,
+      "step": 8885
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2799588339462368e-05,
+      "loss": 0.9716,
+      "step": 8886
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2798092809558238e-05,
+      "loss": 1.186,
+      "step": 8887
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.279659721175262e-05,
+      "loss": 1.1331,
+      "step": 8888
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2795101546081814e-05,
+      "loss": 1.1612,
+      "step": 8889
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2793605812582114e-05,
+      "loss": 1.0742,
+      "step": 8890
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2792110011289817e-05,
+      "loss": 1.138,
+      "step": 8891
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2790614142241224e-05,
+      "loss": 1.2185,
+      "step": 8892
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2789118205472627e-05,
+      "loss": 1.127,
+      "step": 8893
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.278762220102034e-05,
+      "loss": 1.2784,
+      "step": 8894
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2786126128920655e-05,
+      "loss": 1.0553,
+      "step": 8895
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2784629989209884e-05,
+      "loss": 1.166,
+      "step": 8896
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2783133781924335e-05,
+      "loss": 0.9655,
+      "step": 8897
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.278163750710031e-05,
+      "loss": 1.0791,
+      "step": 8898
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2780141164774128e-05,
+      "loss": 1.0339,
+      "step": 8899
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2778644754982091e-05,
+      "loss": 1.1517,
+      "step": 8900
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2777148277760521e-05,
+      "loss": 1.2927,
+      "step": 8901
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2775651733145727e-05,
+      "loss": 1.2959,
+      "step": 8902
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2774155121174032e-05,
+      "loss": 1.1192,
+      "step": 8903
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2772658441881746e-05,
+      "loss": 1.1993,
+      "step": 8904
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2771161695305198e-05,
+      "loss": 1.1223,
+      "step": 8905
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2769664881480704e-05,
+      "loss": 1.1181,
+      "step": 8906
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2768168000444587e-05,
+      "loss": 1.1702,
+      "step": 8907
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2766671052233174e-05,
+      "loss": 1.1204,
+      "step": 8908
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2765174036882791e-05,
+      "loss": 1.1008,
+      "step": 8909
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2763676954429766e-05,
+      "loss": 0.9329,
+      "step": 8910
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2762179804910431e-05,
+      "loss": 1.2146,
+      "step": 8911
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2760682588361115e-05,
+      "loss": 1.1474,
+      "step": 8912
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2759185304818149e-05,
+      "loss": 1.0472,
+      "step": 8913
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2757687954317874e-05,
+      "loss": 1.1734,
+      "step": 8914
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.275619053689662e-05,
+      "loss": 1.1015,
+      "step": 8915
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2754693052590728e-05,
+      "loss": 1.0659,
+      "step": 8916
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2753195501436535e-05,
+      "loss": 1.3271,
+      "step": 8917
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2751697883470386e-05,
+      "loss": 1.1699,
+      "step": 8918
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2750200198728622e-05,
+      "loss": 1.0664,
+      "step": 8919
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2748702447247587e-05,
+      "loss": 1.2645,
+      "step": 8920
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2747204629063626e-05,
+      "loss": 1.0234,
+      "step": 8921
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2745706744213089e-05,
+      "loss": 1.0288,
+      "step": 8922
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2744208792732324e-05,
+      "loss": 1.166,
+      "step": 8923
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2742710774657683e-05,
+      "loss": 1.0312,
+      "step": 8924
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2741212690025514e-05,
+      "loss": 1.0912,
+      "step": 8925
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2739714538872179e-05,
+      "loss": 1.0952,
+      "step": 8926
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2738216321234026e-05,
+      "loss": 1.1113,
+      "step": 8927
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.273671803714742e-05,
+      "loss": 0.9184,
+      "step": 8928
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.273521968664871e-05,
+      "loss": 1.2318,
+      "step": 8929
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2733721269774267e-05,
+      "loss": 1.1074,
+      "step": 8930
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2732222786560446e-05,
+      "loss": 1.1113,
+      "step": 8931
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2730724237043614e-05,
+      "loss": 1.1165,
+      "step": 8932
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2729225621260134e-05,
+      "loss": 0.9761,
+      "step": 8933
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2727726939246376e-05,
+      "loss": 1.2405,
+      "step": 8934
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2726228191038705e-05,
+      "loss": 1.1634,
+      "step": 8935
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2724729376673496e-05,
+      "loss": 1.1714,
+      "step": 8936
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2723230496187116e-05,
+      "loss": 1.1632,
+      "step": 8937
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2721731549615943e-05,
+      "loss": 1.3238,
+      "step": 8938
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2720232536996347e-05,
+      "loss": 1.2422,
+      "step": 8939
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.271873345836471e-05,
+      "loss": 1.1427,
+      "step": 8940
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2717234313757404e-05,
+      "loss": 1.1695,
+      "step": 8941
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2715735103210816e-05,
+      "loss": 1.1815,
+      "step": 8942
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2714235826761321e-05,
+      "loss": 1.0462,
+      "step": 8943
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2712736484445306e-05,
+      "loss": 1.256,
+      "step": 8944
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2711237076299155e-05,
+      "loss": 1.1148,
+      "step": 8945
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2709737602359253e-05,
+      "loss": 1.0871,
+      "step": 8946
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2708238062661988e-05,
+      "loss": 1.0073,
+      "step": 8947
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.270673845724375e-05,
+      "loss": 1.2409,
+      "step": 8948
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.270523878614093e-05,
+      "loss": 1.2119,
+      "step": 8949
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.270373904938992e-05,
+      "loss": 1.0416,
+      "step": 8950
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2702239247027115e-05,
+      "loss": 1.2705,
+      "step": 8951
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2700739379088913e-05,
+      "loss": 1.1374,
+      "step": 8952
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2699239445611706e-05,
+      "loss": 1.1831,
+      "step": 8953
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2697739446631898e-05,
+      "loss": 1.0346,
+      "step": 8954
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2696239382185884e-05,
+      "loss": 1.1289,
+      "step": 8955
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.269473925231007e-05,
+      "loss": 1.2291,
+      "step": 8956
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2693239057040863e-05,
+      "loss": 1.1761,
+      "step": 8957
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2691738796414663e-05,
+      "loss": 1.2134,
+      "step": 8958
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2690238470467877e-05,
+      "loss": 1.0248,
+      "step": 8959
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2688738079236917e-05,
+      "loss": 1.2297,
+      "step": 8960
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2687237622758188e-05,
+      "loss": 1.1794,
+      "step": 8961
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2685737101068108e-05,
+      "loss": 1.1354,
+      "step": 8962
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2684236514203085e-05,
+      "loss": 1.0129,
+      "step": 8963
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2682735862199539e-05,
+      "loss": 1.1618,
+      "step": 8964
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2681235145093878e-05,
+      "loss": 1.0818,
+      "step": 8965
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.267973436292253e-05,
+      "loss": 1.0929,
+      "step": 8966
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2678233515721907e-05,
+      "loss": 1.2095,
+      "step": 8967
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2676732603528435e-05,
+      "loss": 1.2554,
+      "step": 8968
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2675231626378532e-05,
+      "loss": 1.1097,
+      "step": 8969
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2673730584308629e-05,
+      "loss": 1.1228,
+      "step": 8970
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2672229477355144e-05,
+      "loss": 1.0635,
+      "step": 8971
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2670728305554509e-05,
+      "loss": 1.2681,
+      "step": 8972
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2669227068943154e-05,
+      "loss": 1.2106,
+      "step": 8973
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2667725767557508e-05,
+      "loss": 1.2886,
+      "step": 8974
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2666224401434003e-05,
+      "loss": 1.1076,
+      "step": 8975
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2664722970609069e-05,
+      "loss": 1.1683,
+      "step": 8976
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2663221475119149e-05,
+      "loss": 1.114,
+      "step": 8977
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2661719915000675e-05,
+      "loss": 1.0676,
+      "step": 8978
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2660218290290085e-05,
+      "loss": 1.0793,
+      "step": 8979
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2658716601023824e-05,
+      "loss": 1.2295,
+      "step": 8980
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2657214847238329e-05,
+      "loss": 1.1302,
+      "step": 8981
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2655713028970045e-05,
+      "loss": 1.1851,
+      "step": 8982
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2654211146255412e-05,
+      "loss": 1.0997,
+      "step": 8983
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2652709199130884e-05,
+      "loss": 1.0846,
+      "step": 8984
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2651207187632903e-05,
+      "loss": 1.2473,
+      "step": 8985
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2649705111797923e-05,
+      "loss": 1.0183,
+      "step": 8986
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2648202971662391e-05,
+      "loss": 1.1936,
+      "step": 8987
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2646700767262761e-05,
+      "loss": 0.8686,
+      "step": 8988
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2645198498635486e-05,
+      "loss": 1.2838,
+      "step": 8989
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2643696165817026e-05,
+      "loss": 1.2143,
+      "step": 8990
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.264219376884383e-05,
+      "loss": 1.1894,
+      "step": 8991
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2640691307752367e-05,
+      "loss": 1.226,
+      "step": 8992
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2639188782579086e-05,
+      "loss": 1.1649,
+      "step": 8993
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2637686193360459e-05,
+      "loss": 1.0643,
+      "step": 8994
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2636183540132945e-05,
+      "loss": 1.0261,
+      "step": 8995
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2634680822933005e-05,
+      "loss": 1.2902,
+      "step": 8996
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2633178041797116e-05,
+      "loss": 1.1063,
+      "step": 8997
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2631675196761735e-05,
+      "loss": 1.0372,
+      "step": 8998
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.263017228786334e-05,
+      "loss": 1.1223,
+      "step": 8999
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2628669315138395e-05,
+      "loss": 0.9575,
+      "step": 9000
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2627166278623379e-05,
+      "loss": 1.1268,
+      "step": 9001
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2625663178354763e-05,
+      "loss": 1.2064,
+      "step": 9002
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2624160014369025e-05,
+      "loss": 1.0059,
+      "step": 9003
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2622656786702638e-05,
+      "loss": 1.0861,
+      "step": 9004
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2621153495392086e-05,
+      "loss": 1.0511,
+      "step": 9005
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2619650140473845e-05,
+      "loss": 1.0974,
+      "step": 9006
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.26181467219844e-05,
+      "loss": 1.1118,
+      "step": 9007
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2616643239960236e-05,
+      "loss": 1.1618,
+      "step": 9008
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2615139694437835e-05,
+      "loss": 1.1198,
+      "step": 9009
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2613636085453682e-05,
+      "loss": 1.1431,
+      "step": 9010
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.261213241304427e-05,
+      "loss": 1.3086,
+      "step": 9011
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2610628677246083e-05,
+      "loss": 1.1969,
+      "step": 9012
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.260912487809562e-05,
+      "loss": 1.1601,
+      "step": 9013
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2607621015629367e-05,
+      "loss": 1.083,
+      "step": 9014
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2606117089883822e-05,
+      "loss": 1.1821,
+      "step": 9015
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2604613100895478e-05,
+      "loss": 1.2283,
+      "step": 9016
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2603109048700835e-05,
+      "loss": 1.1859,
+      "step": 9017
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.260160493333639e-05,
+      "loss": 1.2288,
+      "step": 9018
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2600100754838647e-05,
+      "loss": 1.2153,
+      "step": 9019
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2598596513244099e-05,
+      "loss": 1.1239,
+      "step": 9020
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2597092208589261e-05,
+      "loss": 1.1303,
+      "step": 9021
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.259558784091063e-05,
+      "loss": 1.1789,
+      "step": 9022
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.259408341024472e-05,
+      "loss": 1.0846,
+      "step": 9023
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.259257891662803e-05,
+      "loss": 1.2633,
+      "step": 9024
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2591074360097078e-05,
+      "loss": 1.0951,
+      "step": 9025
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2589569740688368e-05,
+      "loss": 1.022,
+      "step": 9026
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2588065058438419e-05,
+      "loss": 1.0334,
+      "step": 9027
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2586560313383738e-05,
+      "loss": 1.0871,
+      "step": 9028
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2585055505560848e-05,
+      "loss": 1.1323,
+      "step": 9029
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2583550635006263e-05,
+      "loss": 1.2526,
+      "step": 9030
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2582045701756503e-05,
+      "loss": 1.1259,
+      "step": 9031
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2580540705848085e-05,
+      "loss": 1.2593,
+      "step": 9032
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2579035647317537e-05,
+      "loss": 1.0886,
+      "step": 9033
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2577530526201373e-05,
+      "loss": 1.1896,
+      "step": 9034
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2576025342536129e-05,
+      "loss": 1.1035,
+      "step": 9035
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.257452009635832e-05,
+      "loss": 0.8714,
+      "step": 9036
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2573014787704484e-05,
+      "loss": 1.2274,
+      "step": 9037
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2571509416611145e-05,
+      "loss": 1.0924,
+      "step": 9038
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2570003983114836e-05,
+      "loss": 1.186,
+      "step": 9039
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2568498487252086e-05,
+      "loss": 1.065,
+      "step": 9040
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2566992929059433e-05,
+      "loss": 0.872,
+      "step": 9041
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.256548730857341e-05,
+      "loss": 1.0164,
+      "step": 9042
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2563981625830557e-05,
+      "loss": 1.0674,
+      "step": 9043
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2562475880867406e-05,
+      "loss": 1.0906,
+      "step": 9044
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2560970073720506e-05,
+      "loss": 1.0957,
+      "step": 9045
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.255946420442639e-05,
+      "loss": 1.2829,
+      "step": 9046
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2557958273021608e-05,
+      "loss": 1.2303,
+      "step": 9047
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2556452279542702e-05,
+      "loss": 0.9842,
+      "step": 9048
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2554946224026215e-05,
+      "loss": 1.1523,
+      "step": 9049
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2553440106508696e-05,
+      "loss": 0.9213,
+      "step": 9050
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2551933927026698e-05,
+      "loss": 1.1181,
+      "step": 9051
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2550427685616767e-05,
+      "loss": 1.2972,
+      "step": 9052
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2548921382315458e-05,
+      "loss": 1.198,
+      "step": 9053
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.254741501715932e-05,
+      "loss": 1.0101,
+      "step": 9054
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2545908590184914e-05,
+      "loss": 1.1225,
+      "step": 9055
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2544402101428792e-05,
+      "loss": 1.0869,
+      "step": 9056
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2542895550927516e-05,
+      "loss": 1.0178,
+      "step": 9057
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.254138893871764e-05,
+      "loss": 1.1304,
+      "step": 9058
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2539882264835733e-05,
+      "loss": 0.9923,
+      "step": 9059
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.253837552931835e-05,
+      "loss": 1.1208,
+      "step": 9060
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2536868732202057e-05,
+      "loss": 1.0417,
+      "step": 9061
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.253536187352342e-05,
+      "loss": 1.1778,
+      "step": 9062
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2533854953319009e-05,
+      "loss": 1.1674,
+      "step": 9063
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2532347971625389e-05,
+      "loss": 1.1454,
+      "step": 9064
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2530840928479132e-05,
+      "loss": 1.2028,
+      "step": 9065
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2529333823916807e-05,
+      "loss": 1.1507,
+      "step": 9066
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2527826657974994e-05,
+      "loss": 1.3644,
+      "step": 9067
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2526319430690254e-05,
+      "loss": 1.1703,
+      "step": 9068
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2524812142099176e-05,
+      "loss": 1.0821,
+      "step": 9069
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2523304792238332e-05,
+      "loss": 1.0313,
+      "step": 9070
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.25217973811443e-05,
+      "loss": 1.1533,
+      "step": 9071
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2520289908853665e-05,
+      "loss": 1.2359,
+      "step": 9072
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2518782375403005e-05,
+      "loss": 1.2028,
+      "step": 9073
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2517274780828901e-05,
+      "loss": 1.0781,
+      "step": 9074
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2515767125167944e-05,
+      "loss": 1.1729,
+      "step": 9075
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2514259408456715e-05,
+      "loss": 1.0513,
+      "step": 9076
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2512751630731808e-05,
+      "loss": 1.2894,
+      "step": 9077
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2511243792029803e-05,
+      "loss": 1.2076,
+      "step": 9078
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2509735892387303e-05,
+      "loss": 1.0475,
+      "step": 9079
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2508227931840888e-05,
+      "loss": 1.1268,
+      "step": 9080
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.250671991042716e-05,
+      "loss": 1.1117,
+      "step": 9081
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2505211828182712e-05,
+      "loss": 1.0078,
+      "step": 9082
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2503703685144141e-05,
+      "loss": 1.1508,
+      "step": 9083
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2502195481348045e-05,
+      "loss": 1.2174,
+      "step": 9084
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2500687216831022e-05,
+      "loss": 1.0575,
+      "step": 9085
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2499178891629674e-05,
+      "loss": 1.0725,
+      "step": 9086
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2497670505780605e-05,
+      "loss": 1.0111,
+      "step": 9087
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.249616205932042e-05,
+      "loss": 1.1585,
+      "step": 9088
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2494653552285721e-05,
+      "loss": 1.244,
+      "step": 9089
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2493144984713119e-05,
+      "loss": 1.1215,
+      "step": 9090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2491636356639217e-05,
+      "loss": 1.268,
+      "step": 9091
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2490127668100632e-05,
+      "loss": 1.1006,
+      "step": 9092
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2488618919133966e-05,
+      "loss": 1.1726,
+      "step": 9093
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2487110109775841e-05,
+      "loss": 1.236,
+      "step": 9094
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2485601240062868e-05,
+      "loss": 1.2507,
+      "step": 9095
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2484092310031663e-05,
+      "loss": 1.1308,
+      "step": 9096
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2482583319718842e-05,
+      "loss": 1.081,
+      "step": 9097
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2481074269161026e-05,
+      "loss": 1.3467,
+      "step": 9098
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2479565158394833e-05,
+      "loss": 1.1238,
+      "step": 9099
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2478055987456887e-05,
+      "loss": 1.1009,
+      "step": 9100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2476546756383809e-05,
+      "loss": 1.0642,
+      "step": 9101
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2475037465212226e-05,
+      "loss": 1.2854,
+      "step": 9102
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2473528113978761e-05,
+      "loss": 1.2016,
+      "step": 9103
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2472018702720042e-05,
+      "loss": 1.0813,
+      "step": 9104
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2470509231472698e-05,
+      "loss": 1.0472,
+      "step": 9105
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2468999700273365e-05,
+      "loss": 1.2172,
+      "step": 9106
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2467490109158667e-05,
+      "loss": 1.1802,
+      "step": 9107
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2465980458165243e-05,
+      "loss": 1.0563,
+      "step": 9108
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2464470747329723e-05,
+      "loss": 1.2511,
+      "step": 9109
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.246296097668875e-05,
+      "loss": 1.135,
+      "step": 9110
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2461451146278951e-05,
+      "loss": 1.0163,
+      "step": 9111
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2459941256136976e-05,
+      "loss": 1.0956,
+      "step": 9112
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2458431306299459e-05,
+      "loss": 1.0553,
+      "step": 9113
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2456921296803047e-05,
+      "loss": 1.2959,
+      "step": 9114
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2455411227684377e-05,
+      "loss": 1.2233,
+      "step": 9115
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2453901098980102e-05,
+      "loss": 1.1198,
+      "step": 9116
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2452390910726859e-05,
+      "loss": 1.1318,
+      "step": 9117
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2450880662961307e-05,
+      "loss": 1.1482,
+      "step": 9118
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2449370355720083e-05,
+      "loss": 1.1902,
+      "step": 9119
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2447859989039848e-05,
+      "loss": 0.9002,
+      "step": 9120
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2446349562957247e-05,
+      "loss": 1.1425,
+      "step": 9121
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2444839077508939e-05,
+      "loss": 1.1768,
+      "step": 9122
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2443328532731574e-05,
+      "loss": 1.1936,
+      "step": 9123
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2441817928661814e-05,
+      "loss": 1.2102,
+      "step": 9124
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.244030726533631e-05,
+      "loss": 1.1056,
+      "step": 9125
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2438796542791729e-05,
+      "loss": 1.1565,
+      "step": 9126
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2437285761064728e-05,
+      "loss": 1.1715,
+      "step": 9127
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2435774920191968e-05,
+      "loss": 1.0375,
+      "step": 9128
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2434264020210113e-05,
+      "loss": 1.2355,
+      "step": 9129
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2432753061155833e-05,
+      "loss": 1.0938,
+      "step": 9130
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2431242043065788e-05,
+      "loss": 1.1057,
+      "step": 9131
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2429730965976647e-05,
+      "loss": 1.2052,
+      "step": 9132
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2428219829925083e-05,
+      "loss": 1.1801,
+      "step": 9133
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2426708634947767e-05,
+      "loss": 1.0894,
+      "step": 9134
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2425197381081364e-05,
+      "loss": 1.1145,
+      "step": 9135
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2423686068362556e-05,
+      "loss": 1.0424,
+      "step": 9136
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2422174696828015e-05,
+      "loss": 1.0466,
+      "step": 9137
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2420663266514419e-05,
+      "loss": 1.1205,
+      "step": 9138
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.241915177745844e-05,
+      "loss": 1.1116,
+      "step": 9139
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2417640229696766e-05,
+      "loss": 1.101,
+      "step": 9140
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.241612862326607e-05,
+      "loss": 1.2709,
+      "step": 9141
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2414616958203044e-05,
+      "loss": 1.2089,
+      "step": 9142
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2413105234544358e-05,
+      "loss": 1.1335,
+      "step": 9143
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2411593452326711e-05,
+      "loss": 1.2818,
+      "step": 9144
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2410081611586775e-05,
+      "loss": 1.1254,
+      "step": 9145
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2408569712361253e-05,
+      "loss": 1.0893,
+      "step": 9146
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2407057754686824e-05,
+      "loss": 1.1327,
+      "step": 9147
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2405545738600187e-05,
+      "loss": 1.2092,
+      "step": 9148
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2404033664138024e-05,
+      "loss": 1.2091,
+      "step": 9149
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.240252153133704e-05,
+      "loss": 1.1471,
+      "step": 9150
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2401009340233917e-05,
+      "loss": 1.2499,
+      "step": 9151
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2399497090865365e-05,
+      "loss": 1.1407,
+      "step": 9152
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2397984783268068e-05,
+      "loss": 1.2356,
+      "step": 9153
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.239647241747874e-05,
+      "loss": 1.1916,
+      "step": 9154
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2394959993534069e-05,
+      "loss": 1.1544,
+      "step": 9155
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2393447511470765e-05,
+      "loss": 1.1196,
+      "step": 9156
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2391934971325527e-05,
+      "loss": 1.1769,
+      "step": 9157
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2390422373135063e-05,
+      "loss": 1.2601,
+      "step": 9158
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2388909716936077e-05,
+      "loss": 1.1361,
+      "step": 9159
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2387397002765277e-05,
+      "loss": 1.1295,
+      "step": 9160
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2385884230659375e-05,
+      "loss": 1.0829,
+      "step": 9161
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.238437140065508e-05,
+      "loss": 1.1422,
+      "step": 9162
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.23828585127891e-05,
+      "loss": 1.1464,
+      "step": 9163
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2381345567098156e-05,
+      "loss": 1.0391,
+      "step": 9164
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2379832563618952e-05,
+      "loss": 1.1688,
+      "step": 9165
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2378319502388216e-05,
+      "loss": 1.2194,
+      "step": 9166
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2376806383442657e-05,
+      "loss": 1.1664,
+      "step": 9167
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2375293206818997e-05,
+      "loss": 1.1535,
+      "step": 9168
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.237377997255396e-05,
+      "loss": 1.2069,
+      "step": 9169
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2372266680684259e-05,
+      "loss": 0.9816,
+      "step": 9170
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2370753331246628e-05,
+      "loss": 1.0864,
+      "step": 9171
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.236923992427778e-05,
+      "loss": 1.0181,
+      "step": 9172
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.236772645981445e-05,
+      "loss": 1.2263,
+      "step": 9173
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.236621293789336e-05,
+      "loss": 1.0515,
+      "step": 9174
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2364699358551245e-05,
+      "loss": 1.1618,
+      "step": 9175
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2363185721824826e-05,
+      "loss": 1.1083,
+      "step": 9176
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2361672027750845e-05,
+      "loss": 1.0133,
+      "step": 9177
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2360158276366023e-05,
+      "loss": 1.3095,
+      "step": 9178
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2358644467707107e-05,
+      "loss": 1.1443,
+      "step": 9179
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.235713060181082e-05,
+      "loss": 1.1793,
+      "step": 9180
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2355616678713909e-05,
+      "loss": 1.1554,
+      "step": 9181
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2354102698453107e-05,
+      "loss": 1.163,
+      "step": 9182
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.235258866106516e-05,
+      "loss": 1.2031,
+      "step": 9183
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2351074566586798e-05,
+      "loss": 1.1715,
+      "step": 9184
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2349560415054775e-05,
+      "loss": 1.1918,
+      "step": 9185
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.234804620650583e-05,
+      "loss": 1.1748,
+      "step": 9186
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2346531940976707e-05,
+      "loss": 1.032,
+      "step": 9187
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2345017618504155e-05,
+      "loss": 1.1267,
+      "step": 9188
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2343503239124926e-05,
+      "loss": 1.0547,
+      "step": 9189
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.234198880287576e-05,
+      "loss": 1.2696,
+      "step": 9190
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2340474309793415e-05,
+      "loss": 1.0545,
+      "step": 9191
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2338959759914644e-05,
+      "loss": 1.1531,
+      "step": 9192
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2337445153276197e-05,
+      "loss": 1.0132,
+      "step": 9193
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.233593048991483e-05,
+      "loss": 1.1305,
+      "step": 9194
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.23344157698673e-05,
+      "loss": 1.1997,
+      "step": 9195
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2332900993170363e-05,
+      "loss": 1.1514,
+      "step": 9196
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.233138615986078e-05,
+      "loss": 1.1005,
+      "step": 9197
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2329871269975314e-05,
+      "loss": 1.0875,
+      "step": 9198
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2328356323550723e-05,
+      "loss": 1.179,
+      "step": 9199
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2326841320623769e-05,
+      "loss": 1.1586,
+      "step": 9200
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2325326261231224e-05,
+      "loss": 1.1163,
+      "step": 9201
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2323811145409843e-05,
+      "loss": 1.1289,
+      "step": 9202
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2322295973196403e-05,
+      "loss": 1.1334,
+      "step": 9203
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.232078074462767e-05,
+      "loss": 1.1975,
+      "step": 9204
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2319265459740412e-05,
+      "loss": 1.1787,
+      "step": 9205
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2317750118571403e-05,
+      "loss": 1.1721,
+      "step": 9206
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2316234721157416e-05,
+      "loss": 1.3209,
+      "step": 9207
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.231471926753522e-05,
+      "loss": 1.0812,
+      "step": 9208
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2313203757741598e-05,
+      "loss": 1.0048,
+      "step": 9209
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2311688191813323e-05,
+      "loss": 1.1039,
+      "step": 9210
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2310172569787176e-05,
+      "loss": 1.2639,
+      "step": 9211
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2308656891699931e-05,
+      "loss": 1.1684,
+      "step": 9212
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2307141157588376e-05,
+      "loss": 1.0696,
+      "step": 9213
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2305625367489288e-05,
+      "loss": 1.2404,
+      "step": 9214
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2304109521439453e-05,
+      "loss": 1.2241,
+      "step": 9215
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2302593619475656e-05,
+      "loss": 1.2196,
+      "step": 9216
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2301077661634686e-05,
+      "loss": 1.2538,
+      "step": 9217
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2299561647953327e-05,
+      "loss": 1.189,
+      "step": 9218
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.229804557846837e-05,
+      "loss": 1.0814,
+      "step": 9219
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.22965294532166e-05,
+      "loss": 1.1405,
+      "step": 9220
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2295013272234821e-05,
+      "loss": 1.0759,
+      "step": 9221
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2293497035559816e-05,
+      "loss": 1.2098,
+      "step": 9222
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2291980743228384e-05,
+      "loss": 1.1635,
+      "step": 9223
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.229046439527732e-05,
+      "loss": 1.3105,
+      "step": 9224
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.228894799174342e-05,
+      "loss": 1.2805,
+      "step": 9225
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2287431532663483e-05,
+      "loss": 0.9809,
+      "step": 9226
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2285915018074312e-05,
+      "loss": 1.3207,
+      "step": 9227
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2284398448012704e-05,
+      "loss": 1.2318,
+      "step": 9228
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2282881822515464e-05,
+      "loss": 1.0328,
+      "step": 9229
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2281365141619397e-05,
+      "loss": 1.1429,
+      "step": 9230
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2279848405361307e-05,
+      "loss": 1.1459,
+      "step": 9231
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2278331613778e-05,
+      "loss": 1.0857,
+      "step": 9232
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.227681476690629e-05,
+      "loss": 1.0843,
+      "step": 9233
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2275297864782974e-05,
+      "loss": 1.206,
+      "step": 9234
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2273780907444875e-05,
+      "loss": 1.2069,
+      "step": 9235
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2272263894928798e-05,
+      "loss": 1.0446,
+      "step": 9236
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2270746827271562e-05,
+      "loss": 1.3399,
+      "step": 9237
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2269229704509974e-05,
+      "loss": 1.1281,
+      "step": 9238
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2267712526680859e-05,
+      "loss": 1.1759,
+      "step": 9239
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2266195293821027e-05,
+      "loss": 1.1283,
+      "step": 9240
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2264678005967303e-05,
+      "loss": 1.2415,
+      "step": 9241
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2263160663156501e-05,
+      "loss": 1.1383,
+      "step": 9242
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2261643265425447e-05,
+      "loss": 1.222,
+      "step": 9243
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2260125812810961e-05,
+      "loss": 1.2009,
+      "step": 9244
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.225860830534987e-05,
+      "loss": 1.2401,
+      "step": 9245
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2257090743078997e-05,
+      "loss": 0.9871,
+      "step": 9246
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.225557312603517e-05,
+      "loss": 1.1039,
+      "step": 9247
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2254055454255213e-05,
+      "loss": 1.1686,
+      "step": 9248
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2252537727775965e-05,
+      "loss": 1.2784,
+      "step": 9249
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2251019946634246e-05,
+      "loss": 1.0211,
+      "step": 9250
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2249502110866895e-05,
+      "loss": 1.0057,
+      "step": 9251
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2247984220510742e-05,
+      "loss": 1.1859,
+      "step": 9252
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2246466275602624e-05,
+      "loss": 1.0743,
+      "step": 9253
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2244948276179378e-05,
+      "loss": 1.0037,
+      "step": 9254
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2243430222277835e-05,
+      "loss": 0.9928,
+      "step": 9255
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2241912113934842e-05,
+      "loss": 1.092,
+      "step": 9256
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2240393951187232e-05,
+      "loss": 1.082,
+      "step": 9257
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2238875734071853e-05,
+      "loss": 1.1841,
+      "step": 9258
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2237357462625543e-05,
+      "loss": 1.1311,
+      "step": 9259
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2235839136885148e-05,
+      "loss": 1.1245,
+      "step": 9260
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2234320756887512e-05,
+      "loss": 1.25,
+      "step": 9261
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2232802322669481e-05,
+      "loss": 1.1899,
+      "step": 9262
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2231283834267904e-05,
+      "loss": 1.2407,
+      "step": 9263
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2229765291719633e-05,
+      "loss": 1.3058,
+      "step": 9264
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2228246695061511e-05,
+      "loss": 1.135,
+      "step": 9265
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.22267280443304e-05,
+      "loss": 1.2206,
+      "step": 9266
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2225209339563144e-05,
+      "loss": 1.1731,
+      "step": 9267
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2223690580796606e-05,
+      "loss": 1.284,
+      "step": 9268
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2222171768067632e-05,
+      "loss": 1.1223,
+      "step": 9269
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2220652901413086e-05,
+      "loss": 1.0714,
+      "step": 9270
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2219133980869828e-05,
+      "loss": 1.0449,
+      "step": 9271
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.221761500647471e-05,
+      "loss": 1.0093,
+      "step": 9272
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.22160959782646e-05,
+      "loss": 1.2411,
+      "step": 9273
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2214576896276357e-05,
+      "loss": 1.1204,
+      "step": 9274
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2213057760546845e-05,
+      "loss": 1.1042,
+      "step": 9275
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2211538571112933e-05,
+      "loss": 1.1957,
+      "step": 9276
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.221001932801148e-05,
+      "loss": 1.2146,
+      "step": 9277
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.220850003127936e-05,
+      "loss": 1.1354,
+      "step": 9278
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2206980680953438e-05,
+      "loss": 1.1986,
+      "step": 9279
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2205461277070584e-05,
+      "loss": 1.2746,
+      "step": 9280
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2203941819667673e-05,
+      "loss": 1.0758,
+      "step": 9281
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2202422308781574e-05,
+      "loss": 0.966,
+      "step": 9282
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2200902744449164e-05,
+      "loss": 1.1937,
+      "step": 9283
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2199383126707315e-05,
+      "loss": 1.2116,
+      "step": 9284
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2197863455592906e-05,
+      "loss": 1.244,
+      "step": 9285
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2196343731142819e-05,
+      "loss": 1.098,
+      "step": 9286
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2194823953393924e-05,
+      "loss": 1.0504,
+      "step": 9287
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.219330412238311e-05,
+      "loss": 1.1316,
+      "step": 9288
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.219178423814725e-05,
+      "loss": 1.2812,
+      "step": 9289
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2190264300723238e-05,
+      "loss": 1.1346,
+      "step": 9290
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2188744310147949e-05,
+      "loss": 1.1147,
+      "step": 9291
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2187224266458276e-05,
+      "loss": 0.9752,
+      "step": 9292
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2185704169691098e-05,
+      "loss": 1.2975,
+      "step": 9293
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2184184019883311e-05,
+      "loss": 1.148,
+      "step": 9294
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2182663817071801e-05,
+      "loss": 1.2199,
+      "step": 9295
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2181143561293459e-05,
+      "loss": 1.0623,
+      "step": 9296
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2179623252585175e-05,
+      "loss": 1.2591,
+      "step": 9297
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2178102890983847e-05,
+      "loss": 1.0294,
+      "step": 9298
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2176582476526366e-05,
+      "loss": 1.1246,
+      "step": 9299
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2175062009249631e-05,
+      "loss": 1.1854,
+      "step": 9300
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2173541489190535e-05,
+      "loss": 1.1718,
+      "step": 9301
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2172020916385983e-05,
+      "loss": 1.176,
+      "step": 9302
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2170500290872869e-05,
+      "loss": 1.1576,
+      "step": 9303
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2168979612688097e-05,
+      "loss": 1.0153,
+      "step": 9304
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2167458881868566e-05,
+      "loss": 1.165,
+      "step": 9305
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2165938098451184e-05,
+      "loss": 1.0182,
+      "step": 9306
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2164417262472854e-05,
+      "loss": 1.0939,
+      "step": 9307
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2162896373970483e-05,
+      "loss": 1.0412,
+      "step": 9308
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2161375432980976e-05,
+      "loss": 1.2282,
+      "step": 9309
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2159854439541245e-05,
+      "loss": 1.0219,
+      "step": 9310
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.21583333936882e-05,
+      "loss": 1.1255,
+      "step": 9311
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.215681229545875e-05,
+      "loss": 1.0945,
+      "step": 9312
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2155291144889804e-05,
+      "loss": 1.2298,
+      "step": 9313
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2153769942018286e-05,
+      "loss": 1.0831,
+      "step": 9314
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2152248686881102e-05,
+      "loss": 1.2059,
+      "step": 9315
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2150727379515176e-05,
+      "loss": 0.9974,
+      "step": 9316
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2149206019957418e-05,
+      "loss": 1.0155,
+      "step": 9317
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2147684608244754e-05,
+      "loss": 1.1831,
+      "step": 9318
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2146163144414097e-05,
+      "loss": 0.9389,
+      "step": 9319
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2144641628502373e-05,
+      "loss": 1.0155,
+      "step": 9320
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2143120060546505e-05,
+      "loss": 1.0501,
+      "step": 9321
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2141598440583417e-05,
+      "loss": 1.0793,
+      "step": 9322
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2140076768650031e-05,
+      "loss": 1.028,
+      "step": 9323
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2138555044783279e-05,
+      "loss": 1.1745,
+      "step": 9324
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.213703326902008e-05,
+      "loss": 1.189,
+      "step": 9325
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2135511441397369e-05,
+      "loss": 1.2686,
+      "step": 9326
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2133989561952077e-05,
+      "loss": 0.9588,
+      "step": 9327
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2132467630721137e-05,
+      "loss": 1.2843,
+      "step": 9328
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2130945647741474e-05,
+      "loss": 1.1771,
+      "step": 9329
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.212942361305003e-05,
+      "loss": 1.0813,
+      "step": 9330
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2127901526683737e-05,
+      "loss": 1.2038,
+      "step": 9331
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2126379388679532e-05,
+      "loss": 1.1219,
+      "step": 9332
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2124857199074354e-05,
+      "loss": 1.118,
+      "step": 9333
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.212333495790514e-05,
+      "loss": 1.1362,
+      "step": 9334
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2121812665208827e-05,
+      "loss": 1.232,
+      "step": 9335
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2120290321022366e-05,
+      "loss": 1.2299,
+      "step": 9336
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.211876792538269e-05,
+      "loss": 1.2255,
+      "step": 9337
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2117245478326752e-05,
+      "loss": 0.9872,
+      "step": 9338
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2115722979891488e-05,
+      "loss": 1.1513,
+      "step": 9339
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2114200430113854e-05,
+      "loss": 1.1444,
+      "step": 9340
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2112677829030792e-05,
+      "loss": 1.2307,
+      "step": 9341
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.211115517667925e-05,
+      "loss": 1.1771,
+      "step": 9342
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2109632473096183e-05,
+      "loss": 1.1328,
+      "step": 9343
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2108109718318537e-05,
+      "loss": 1.1564,
+      "step": 9344
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.210658691238327e-05,
+      "loss": 1.2215,
+      "step": 9345
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2105064055327335e-05,
+      "loss": 1.2193,
+      "step": 9346
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2103541147187684e-05,
+      "loss": 1.1325,
+      "step": 9347
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2102018188001278e-05,
+      "loss": 0.9631,
+      "step": 9348
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2100495177805072e-05,
+      "loss": 1.1812,
+      "step": 9349
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2098972116636022e-05,
+      "loss": 1.1977,
+      "step": 9350
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2097449004531097e-05,
+      "loss": 1.0119,
+      "step": 9351
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2095925841527249e-05,
+      "loss": 1.1097,
+      "step": 9352
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2094402627661447e-05,
+      "loss": 1.2541,
+      "step": 9353
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2092879362970652e-05,
+      "loss": 1.0961,
+      "step": 9354
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2091356047491827e-05,
+      "loss": 1.131,
+      "step": 9355
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2089832681261943e-05,
+      "loss": 1.1746,
+      "step": 9356
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2088309264317965e-05,
+      "loss": 1.238,
+      "step": 9357
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2086785796696866e-05,
+      "loss": 1.0298,
+      "step": 9358
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.208526227843561e-05,
+      "loss": 1.0695,
+      "step": 9359
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2083738709571172e-05,
+      "loss": 1.2125,
+      "step": 9360
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2082215090140523e-05,
+      "loss": 1.1494,
+      "step": 9361
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2080691420180637e-05,
+      "loss": 1.0909,
+      "step": 9362
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.207916769972849e-05,
+      "loss": 1.222,
+      "step": 9363
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2077643928821056e-05,
+      "loss": 1.1624,
+      "step": 9364
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2076120107495317e-05,
+      "loss": 1.1545,
+      "step": 9365
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2074596235788245e-05,
+      "loss": 1.1458,
+      "step": 9366
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2073072313736826e-05,
+      "loss": 1.2345,
+      "step": 9367
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.207154834137804e-05,
+      "loss": 1.1896,
+      "step": 9368
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2070024318748866e-05,
+      "loss": 1.1862,
+      "step": 9369
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2068500245886288e-05,
+      "loss": 1.1616,
+      "step": 9370
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2066976122827296e-05,
+      "loss": 1.199,
+      "step": 9371
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2065451949608868e-05,
+      "loss": 1.0887,
+      "step": 9372
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2063927726268001e-05,
+      "loss": 1.1515,
+      "step": 9373
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2062403452841674e-05,
+      "loss": 1.1129,
+      "step": 9374
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.206087912936688e-05,
+      "loss": 1.1906,
+      "step": 9375
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.205935475588061e-05,
+      "loss": 1.1308,
+      "step": 9376
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.205783033241986e-05,
+      "loss": 1.1727,
+      "step": 9377
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2056305859021615e-05,
+      "loss": 1.2569,
+      "step": 9378
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2054781335722879e-05,
+      "loss": 1.1196,
+      "step": 9379
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2053256762560638e-05,
+      "loss": 1.2022,
+      "step": 9380
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.20517321395719e-05,
+      "loss": 1.241,
+      "step": 9381
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2050207466793651e-05,
+      "loss": 1.2076,
+      "step": 9382
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.20486827442629e-05,
+      "loss": 1.2532,
+      "step": 9383
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2047157972016641e-05,
+      "loss": 0.9027,
+      "step": 9384
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.204563315009188e-05,
+      "loss": 1.1651,
+      "step": 9385
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.204410827852562e-05,
+      "loss": 1.2993,
+      "step": 9386
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.204258335735486e-05,
+      "loss": 1.21,
+      "step": 9387
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.204105838661661e-05,
+      "loss": 1.1664,
+      "step": 9388
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2039533366347878e-05,
+      "loss": 1.2182,
+      "step": 9389
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2038008296585667e-05,
+      "loss": 1.2244,
+      "step": 9390
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2036483177366989e-05,
+      "loss": 1.0616,
+      "step": 9391
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2034958008728852e-05,
+      "loss": 1.0809,
+      "step": 9392
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.203343279070827e-05,
+      "loss": 1.2252,
+      "step": 9393
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2031907523342254e-05,
+      "loss": 1.1668,
+      "step": 9394
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2030382206667816e-05,
+      "loss": 1.0725,
+      "step": 9395
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2028856840721975e-05,
+      "loss": 1.2852,
+      "step": 9396
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2027331425541745e-05,
+      "loss": 1.0822,
+      "step": 9397
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2025805961164143e-05,
+      "loss": 1.13,
+      "step": 9398
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.202428044762619e-05,
+      "loss": 1.0381,
+      "step": 9399
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2022754884964902e-05,
+      "loss": 1.1258,
+      "step": 9400
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2021229273217303e-05,
+      "loss": 1.1362,
+      "step": 9401
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2019703612420412e-05,
+      "loss": 1.1489,
+      "step": 9402
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2018177902611256e-05,
+      "loss": 1.1602,
+      "step": 9403
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2016652143826855e-05,
+      "loss": 1.2771,
+      "step": 9404
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.201512633610424e-05,
+      "loss": 1.2621,
+      "step": 9405
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2013600479480431e-05,
+      "loss": 1.1145,
+      "step": 9406
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2012074573992465e-05,
+      "loss": 1.1323,
+      "step": 9407
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2010548619677365e-05,
+      "loss": 1.1625,
+      "step": 9408
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2009022616572162e-05,
+      "loss": 1.1552,
+      "step": 9409
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2007496564713888e-05,
+      "loss": 1.1719,
+      "step": 9410
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2005970464139579e-05,
+      "loss": 1.1331,
+      "step": 9411
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2004444314886262e-05,
+      "loss": 1.2013,
+      "step": 9412
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2002918116990979e-05,
+      "loss": 1.2212,
+      "step": 9413
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2001391870490761e-05,
+      "loss": 1.0138,
+      "step": 9414
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1999865575422652e-05,
+      "loss": 1.3016,
+      "step": 9415
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1998339231823683e-05,
+      "loss": 1.1503,
+      "step": 9416
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1996812839730901e-05,
+      "loss": 1.1318,
+      "step": 9417
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1995286399181341e-05,
+      "loss": 1.2067,
+      "step": 9418
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.199375991021205e-05,
+      "loss": 1.1336,
+      "step": 9419
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1992233372860068e-05,
+      "loss": 1.2541,
+      "step": 9420
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1990706787162444e-05,
+      "loss": 1.123,
+      "step": 9421
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1989180153156217e-05,
+      "loss": 1.1654,
+      "step": 9422
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1987653470878441e-05,
+      "loss": 1.1232,
+      "step": 9423
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1986126740366157e-05,
+      "loss": 1.1398,
+      "step": 9424
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198459996165642e-05,
+      "loss": 1.0076,
+      "step": 9425
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1983073134786281e-05,
+      "loss": 1.1956,
+      "step": 9426
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1981546259792785e-05,
+      "loss": 1.2131,
+      "step": 9427
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1980019336712992e-05,
+      "loss": 1.3535,
+      "step": 9428
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1978492365583949e-05,
+      "loss": 1.1852,
+      "step": 9429
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1976965346442716e-05,
+      "loss": 1.1823,
+      "step": 9430
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1975438279326351e-05,
+      "loss": 1.1354,
+      "step": 9431
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1973911164271905e-05,
+      "loss": 1.2294,
+      "step": 9432
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1972384001316442e-05,
+      "loss": 1.0493,
+      "step": 9433
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.197085679049702e-05,
+      "loss": 1.136,
+      "step": 9434
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1969329531850697e-05,
+      "loss": 1.0775,
+      "step": 9435
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1967802225414542e-05,
+      "loss": 1.214,
+      "step": 9436
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1966274871225612e-05,
+      "loss": 1.2578,
+      "step": 9437
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1964747469320972e-05,
+      "loss": 1.1294,
+      "step": 9438
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1963220019737691e-05,
+      "loss": 1.1595,
+      "step": 9439
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1961692522512834e-05,
+      "loss": 1.1713,
+      "step": 9440
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1960164977683469e-05,
+      "loss": 1.1151,
+      "step": 9441
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1958637385286662e-05,
+      "loss": 0.8824,
+      "step": 9442
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1957109745359487e-05,
+      "loss": 1.1248,
+      "step": 9443
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1955582057939014e-05,
+      "loss": 1.3234,
+      "step": 9444
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1954054323062317e-05,
+      "loss": 1.036,
+      "step": 9445
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1952526540766466e-05,
+      "loss": 1.1291,
+      "step": 9446
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.195099871108854e-05,
+      "loss": 1.2386,
+      "step": 9447
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1949470834065612e-05,
+      "loss": 1.0488,
+      "step": 9448
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.194794290973476e-05,
+      "loss": 1.1819,
+      "step": 9449
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1946414938133065e-05,
+      "loss": 1.0756,
+      "step": 9450
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.19448869192976e-05,
+      "loss": 1.1181,
+      "step": 9451
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1943358853265453e-05,
+      "loss": 1.094,
+      "step": 9452
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1941830740073696e-05,
+      "loss": 1.247,
+      "step": 9453
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1940302579759424e-05,
+      "loss": 1.0655,
+      "step": 9454
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.193877437235971e-05,
+      "loss": 1.2348,
+      "step": 9455
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1937246117911647e-05,
+      "loss": 1.109,
+      "step": 9456
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1935717816452315e-05,
+      "loss": 1.1368,
+      "step": 9457
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1934189468018807e-05,
+      "loss": 1.2283,
+      "step": 9458
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1932661072648207e-05,
+      "loss": 1.1725,
+      "step": 9459
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1931132630377607e-05,
+      "loss": 1.1417,
+      "step": 9460
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1929604141244097e-05,
+      "loss": 1.2751,
+      "step": 9461
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.192807560528477e-05,
+      "loss": 1.352,
+      "step": 9462
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1926547022536716e-05,
+      "loss": 1.0563,
+      "step": 9463
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1925018393037036e-05,
+      "loss": 1.1758,
+      "step": 9464
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1923489716822818e-05,
+      "loss": 1.1432,
+      "step": 9465
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1921960993931162e-05,
+      "loss": 1.1112,
+      "step": 9466
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1920432224399163e-05,
+      "loss": 1.0692,
+      "step": 9467
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1918903408263924e-05,
+      "loss": 1.3592,
+      "step": 9468
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1917374545562542e-05,
+      "loss": 1.0568,
+      "step": 9469
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.191584563633212e-05,
+      "loss": 1.0231,
+      "step": 9470
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1914316680609756e-05,
+      "loss": 1.2081,
+      "step": 9471
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1912787678432559e-05,
+      "loss": 1.1683,
+      "step": 9472
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1911258629837627e-05,
+      "loss": 1.1826,
+      "step": 9473
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1909729534862071e-05,
+      "loss": 1.1804,
+      "step": 9474
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1908200393542995e-05,
+      "loss": 1.0121,
+      "step": 9475
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1906671205917509e-05,
+      "loss": 1.182,
+      "step": 9476
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1905141972022716e-05,
+      "loss": 1.1824,
+      "step": 9477
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1903612691895734e-05,
+      "loss": 1.0788,
+      "step": 9478
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1902083365573665e-05,
+      "loss": 1.2738,
+      "step": 9479
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.190055399309363e-05,
+      "loss": 1.291,
+      "step": 9480
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1899024574492739e-05,
+      "loss": 1.343,
+      "step": 9481
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1897495109808108e-05,
+      "loss": 1.059,
+      "step": 9482
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1895965599076849e-05,
+      "loss": 1.275,
+      "step": 9483
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1894436042336082e-05,
+      "loss": 1.2372,
+      "step": 9484
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1892906439622923e-05,
+      "loss": 1.0623,
+      "step": 9485
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1891376790974491e-05,
+      "loss": 1.3095,
+      "step": 9486
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1889847096427907e-05,
+      "loss": 1.1191,
+      "step": 9487
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1888317356020293e-05,
+      "loss": 1.2526,
+      "step": 9488
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.188678756978877e-05,
+      "loss": 1.1733,
+      "step": 9489
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.188525773777046e-05,
+      "loss": 1.0779,
+      "step": 9490
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1883727860002487e-05,
+      "loss": 1.1813,
+      "step": 9491
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1882197936521985e-05,
+      "loss": 1.1678,
+      "step": 9492
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1880667967366069e-05,
+      "loss": 1.1737,
+      "step": 9493
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1879137952571876e-05,
+      "loss": 1.082,
+      "step": 9494
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1877607892176527e-05,
+      "loss": 1.219,
+      "step": 9495
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1876077786217159e-05,
+      "loss": 1.1809,
+      "step": 9496
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.18745476347309e-05,
+      "loss": 1.0778,
+      "step": 9497
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1873017437754884e-05,
+      "loss": 1.2154,
+      "step": 9498
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.187148719532624e-05,
+      "loss": 1.0567,
+      "step": 9499
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.186995690748211e-05,
+      "loss": 1.0169,
+      "step": 9500
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1868426574259621e-05,
+      "loss": 1.1108,
+      "step": 9501
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1866896195695916e-05,
+      "loss": 1.196,
+      "step": 9502
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.186536577182813e-05,
+      "loss": 1.1453,
+      "step": 9503
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1863835302693404e-05,
+      "loss": 1.1468,
+      "step": 9504
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1862304788328876e-05,
+      "loss": 1.1437,
+      "step": 9505
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1860774228771687e-05,
+      "loss": 1.165,
+      "step": 9506
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1859243624058979e-05,
+      "loss": 1.2361,
+      "step": 9507
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1857712974227897e-05,
+      "loss": 1.1598,
+      "step": 9508
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1856182279315585e-05,
+      "loss": 1.2197,
+      "step": 9509
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1854651539359186e-05,
+      "loss": 1.1403,
+      "step": 9510
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.185312075439585e-05,
+      "loss": 1.1296,
+      "step": 9511
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1851589924462723e-05,
+      "loss": 0.999,
+      "step": 9512
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1850059049596954e-05,
+      "loss": 1.2836,
+      "step": 9513
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1848528129835692e-05,
+      "loss": 1.085,
+      "step": 9514
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1846997165216089e-05,
+      "loss": 1.2121,
+      "step": 9515
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1845466155775296e-05,
+      "loss": 1.1461,
+      "step": 9516
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.184393510155047e-05,
+      "loss": 1.1536,
+      "step": 9517
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1842404002578758e-05,
+      "loss": 1.04,
+      "step": 9518
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1840872858897321e-05,
+      "loss": 1.0948,
+      "step": 9519
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1839341670543312e-05,
+      "loss": 1.1541,
+      "step": 9520
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.183781043755389e-05,
+      "loss": 1.1489,
+      "step": 9521
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1836279159966216e-05,
+      "loss": 1.1778,
+      "step": 9522
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1834747837817445e-05,
+      "loss": 1.2204,
+      "step": 9523
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1833216471144739e-05,
+      "loss": 1.1315,
+      "step": 9524
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1831685059985263e-05,
+      "loss": 1.148,
+      "step": 9525
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1830153604376177e-05,
+      "loss": 1.0392,
+      "step": 9526
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1828622104354643e-05,
+      "loss": 1.091,
+      "step": 9527
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1827090559957832e-05,
+      "loss": 1.2051,
+      "step": 9528
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1825558971222903e-05,
+      "loss": 1.0637,
+      "step": 9529
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1824027338187028e-05,
+      "loss": 1.1429,
+      "step": 9530
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1822495660887374e-05,
+      "loss": 1.0881,
+      "step": 9531
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1820963939361107e-05,
+      "loss": 1.2554,
+      "step": 9532
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1819432173645405e-05,
+      "loss": 1.2665,
+      "step": 9533
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1817900363777433e-05,
+      "loss": 1.0615,
+      "step": 9534
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1816368509794365e-05,
+      "loss": 1.0687,
+      "step": 9535
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1814836611733375e-05,
+      "loss": 1.0576,
+      "step": 9536
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.181330466963164e-05,
+      "loss": 1.075,
+      "step": 9537
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.181177268352633e-05,
+      "loss": 1.1856,
+      "step": 9538
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1810240653454628e-05,
+      "loss": 1.3135,
+      "step": 9539
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1808708579453707e-05,
+      "loss": 1.0553,
+      "step": 9540
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.180717646156075e-05,
+      "loss": 1.0004,
+      "step": 9541
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1805644299812932e-05,
+      "loss": 1.0245,
+      "step": 9542
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1804112094247439e-05,
+      "loss": 1.1711,
+      "step": 9543
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1802579844901452e-05,
+      "loss": 1.1833,
+      "step": 9544
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1801047551812153e-05,
+      "loss": 1.2417,
+      "step": 9545
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1799515215016726e-05,
+      "loss": 1.0095,
+      "step": 9546
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1797982834552358e-05,
+      "loss": 1.1588,
+      "step": 9547
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.179645041045623e-05,
+      "loss": 1.0785,
+      "step": 9548
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.179491794276554e-05,
+      "loss": 1.0914,
+      "step": 9549
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1793385431517466e-05,
+      "loss": 1.1745,
+      "step": 9550
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1791852876749204e-05,
+      "loss": 1.0276,
+      "step": 9551
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1790320278497939e-05,
+      "loss": 1.1217,
+      "step": 9552
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.178878763680087e-05,
+      "loss": 1.0914,
+      "step": 9553
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1787254951695181e-05,
+      "loss": 1.1326,
+      "step": 9554
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1785722223218077e-05,
+      "loss": 0.9697,
+      "step": 9555
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.178418945140674e-05,
+      "loss": 1.1327,
+      "step": 9556
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1782656636298378e-05,
+      "loss": 1.2059,
+      "step": 9557
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1781123777930178e-05,
+      "loss": 1.2542,
+      "step": 9558
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1779590876339341e-05,
+      "loss": 1.089,
+      "step": 9559
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.177805793156307e-05,
+      "loss": 1.1308,
+      "step": 9560
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1776524943638561e-05,
+      "loss": 1.2352,
+      "step": 9561
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1774991912603013e-05,
+      "loss": 1.0394,
+      "step": 9562
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1773458838493634e-05,
+      "loss": 1.198,
+      "step": 9563
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1771925721347625e-05,
+      "loss": 1.0677,
+      "step": 9564
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.177039256120219e-05,
+      "loss": 1.1481,
+      "step": 9565
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1768859358094533e-05,
+      "loss": 1.1808,
+      "step": 9566
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1767326112061864e-05,
+      "loss": 0.9997,
+      "step": 9567
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1765792823141385e-05,
+      "loss": 1.1855,
+      "step": 9568
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1764259491370309e-05,
+      "loss": 0.9893,
+      "step": 9569
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1762726116785843e-05,
+      "loss": 1.0714,
+      "step": 9570
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.17611926994252e-05,
+      "loss": 1.0818,
+      "step": 9571
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1759659239325585e-05,
+      "loss": 1.0146,
+      "step": 9572
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1758125736524218e-05,
+      "loss": 1.1585,
+      "step": 9573
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1756592191058308e-05,
+      "loss": 1.2943,
+      "step": 9574
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1755058602965074e-05,
+      "loss": 1.0898,
+      "step": 9575
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.175352497228173e-05,
+      "loss": 1.0336,
+      "step": 9576
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1751991299045491e-05,
+      "loss": 0.9155,
+      "step": 9577
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1750457583293574e-05,
+      "loss": 1.1728,
+      "step": 9578
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1748923825063201e-05,
+      "loss": 1.2598,
+      "step": 9579
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1747390024391587e-05,
+      "loss": 1.2677,
+      "step": 9580
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.174585618131596e-05,
+      "loss": 1.1758,
+      "step": 9581
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1744322295873535e-05,
+      "loss": 1.1742,
+      "step": 9582
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1742788368101541e-05,
+      "loss": 1.1576,
+      "step": 9583
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1741254398037194e-05,
+      "loss": 1.1993,
+      "step": 9584
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1739720385717728e-05,
+      "loss": 1.0713,
+      "step": 9585
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.173818633118036e-05,
+      "loss": 1.1407,
+      "step": 9586
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1736652234462327e-05,
+      "loss": 1.1349,
+      "step": 9587
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1735118095600848e-05,
+      "loss": 1.1029,
+      "step": 9588
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1733583914633157e-05,
+      "loss": 1.1523,
+      "step": 9589
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.173204969159648e-05,
+      "loss": 1.1869,
+      "step": 9590
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1730515426528055e-05,
+      "loss": 1.2465,
+      "step": 9591
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1728981119465109e-05,
+      "loss": 1.1516,
+      "step": 9592
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1727446770444875e-05,
+      "loss": 1.136,
+      "step": 9593
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1725912379504588e-05,
+      "loss": 1.1283,
+      "step": 9594
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1724377946681482e-05,
+      "loss": 0.9426,
+      "step": 9595
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1722843472012794e-05,
+      "loss": 1.0963,
+      "step": 9596
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1721308955535765e-05,
+      "loss": 1.0668,
+      "step": 9597
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1719774397287627e-05,
+      "loss": 1.1298,
+      "step": 9598
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1718239797305622e-05,
+      "loss": 1.0734,
+      "step": 9599
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1716705155626993e-05,
+      "loss": 1.0709,
+      "step": 9600
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1715170472288975e-05,
+      "loss": 1.1946,
+      "step": 9601
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1713635747328817e-05,
+      "loss": 1.0384,
+      "step": 9602
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1712100980783759e-05,
+      "loss": 1.1399,
+      "step": 9603
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1710566172691042e-05,
+      "loss": 1.2155,
+      "step": 9604
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1709031323087918e-05,
+      "loss": 1.1391,
+      "step": 9605
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1707496432011626e-05,
+      "loss": 1.3422,
+      "step": 9606
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1705961499499421e-05,
+      "loss": 1.0487,
+      "step": 9607
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1704426525588545e-05,
+      "loss": 1.0296,
+      "step": 9608
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1702891510316253e-05,
+      "loss": 1.1088,
+      "step": 9609
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1701356453719787e-05,
+      "loss": 1.1558,
+      "step": 9610
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.169982135583641e-05,
+      "loss": 1.2446,
+      "step": 9611
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1698286216703362e-05,
+      "loss": 1.1761,
+      "step": 9612
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1696751036357906e-05,
+      "loss": 1.218,
+      "step": 9613
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1695215814837288e-05,
+      "loss": 1.1168,
+      "step": 9614
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.169368055217877e-05,
+      "loss": 1.1823,
+      "step": 9615
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1692145248419606e-05,
+      "loss": 1.0494,
+      "step": 9616
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1690609903597053e-05,
+      "loss": 1.1366,
+      "step": 9617
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.168907451774837e-05,
+      "loss": 1.2974,
+      "step": 9618
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1687539090910816e-05,
+      "loss": 1.1097,
+      "step": 9619
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.168600362312165e-05,
+      "loss": 1.0809,
+      "step": 9620
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1684468114418134e-05,
+      "loss": 1.2366,
+      "step": 9621
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1682932564837533e-05,
+      "loss": 1.1916,
+      "step": 9622
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1681396974417108e-05,
+      "loss": 1.0294,
+      "step": 9623
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1679861343194121e-05,
+      "loss": 1.1998,
+      "step": 9624
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1678325671205841e-05,
+      "loss": 1.1403,
+      "step": 9625
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1676789958489532e-05,
+      "loss": 1.2088,
+      "step": 9626
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1675254205082463e-05,
+      "loss": 1.2535,
+      "step": 9627
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.16737184110219e-05,
+      "loss": 1.1549,
+      "step": 9628
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1672182576345113e-05,
+      "loss": 1.1166,
+      "step": 9629
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1670646701089374e-05,
+      "loss": 1.2163,
+      "step": 9630
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.166911078529195e-05,
+      "loss": 1.1643,
+      "step": 9631
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1667574828990121e-05,
+      "loss": 1.2181,
+      "step": 9632
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1666038832221152e-05,
+      "loss": 1.2262,
+      "step": 9633
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1664502795022321e-05,
+      "loss": 1.0493,
+      "step": 9634
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.16629667174309e-05,
+      "loss": 1.0684,
+      "step": 9635
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1661430599484169e-05,
+      "loss": 1.1733,
+      "step": 9636
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1659894441219403e-05,
+      "loss": 1.2201,
+      "step": 9637
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1658358242673882e-05,
+      "loss": 1.1933,
+      "step": 9638
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1656822003884883e-05,
+      "loss": 1.0057,
+      "step": 9639
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1655285724889687e-05,
+      "loss": 0.9179,
+      "step": 9640
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1653749405725572e-05,
+      "loss": 1.345,
+      "step": 9641
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1652213046429824e-05,
+      "loss": 1.2789,
+      "step": 9642
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1650676647039723e-05,
+      "loss": 1.0277,
+      "step": 9643
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1649140207592557e-05,
+      "loss": 1.0604,
+      "step": 9644
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1647603728125603e-05,
+      "loss": 1.2081,
+      "step": 9645
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1646067208676157e-05,
+      "loss": 1.1612,
+      "step": 9646
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1644530649281498e-05,
+      "loss": 0.9741,
+      "step": 9647
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.164299404997892e-05,
+      "loss": 1.0023,
+      "step": 9648
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1641457410805703e-05,
+      "loss": 1.1118,
+      "step": 9649
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1639920731799146e-05,
+      "loss": 1.1391,
+      "step": 9650
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1638384012996535e-05,
+      "loss": 1.2204,
+      "step": 9651
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1636847254435162e-05,
+      "loss": 1.1833,
+      "step": 9652
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.163531045615232e-05,
+      "loss": 1.1624,
+      "step": 9653
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1633773618185302e-05,
+      "loss": 1.1436,
+      "step": 9654
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1632236740571403e-05,
+      "loss": 0.9774,
+      "step": 9655
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1630699823347922e-05,
+      "loss": 1.0833,
+      "step": 9656
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1629162866552145e-05,
+      "loss": 1.0517,
+      "step": 9657
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1627625870221384e-05,
+      "loss": 1.0526,
+      "step": 9658
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1626088834392926e-05,
+      "loss": 1.0913,
+      "step": 9659
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1624551759104075e-05,
+      "loss": 1.1595,
+      "step": 9660
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1623014644392128e-05,
+      "loss": 1.058,
+      "step": 9661
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.162147749029439e-05,
+      "loss": 1.1636,
+      "step": 9662
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1619940296848163e-05,
+      "loss": 1.1104,
+      "step": 9663
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1618403064090748e-05,
+      "loss": 1.2819,
+      "step": 9664
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.161686579205945e-05,
+      "loss": 1.1761,
+      "step": 9665
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1615328480791576e-05,
+      "loss": 1.0608,
+      "step": 9666
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.161379113032443e-05,
+      "loss": 1.1182,
+      "step": 9667
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1612253740695318e-05,
+      "loss": 1.1001,
+      "step": 9668
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1610716311941551e-05,
+      "loss": 1.2362,
+      "step": 9669
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1609178844100435e-05,
+      "loss": 1.2045,
+      "step": 9670
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1607641337209279e-05,
+      "loss": 1.1253,
+      "step": 9671
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1606103791305398e-05,
+      "loss": 1.1672,
+      "step": 9672
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.16045662064261e-05,
+      "loss": 1.0279,
+      "step": 9673
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1603028582608703e-05,
+      "loss": 1.0934,
+      "step": 9674
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.160149091989051e-05,
+      "loss": 1.1168,
+      "step": 9675
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1599953218308848e-05,
+      "loss": 1.2041,
+      "step": 9676
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1598415477901022e-05,
+      "loss": 1.1529,
+      "step": 9677
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1596877698704357e-05,
+      "loss": 1.1795,
+      "step": 9678
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1595339880756162e-05,
+      "loss": 1.1241,
+      "step": 9679
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1593802024093765e-05,
+      "loss": 1.0947,
+      "step": 9680
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1592264128754475e-05,
+      "loss": 1.1731,
+      "step": 9681
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1590726194775621e-05,
+      "loss": 1.1425,
+      "step": 9682
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1589188222194521e-05,
+      "loss": 1.1719,
+      "step": 9683
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1587650211048493e-05,
+      "loss": 1.0089,
+      "step": 9684
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1586112161374867e-05,
+      "loss": 1.1962,
+      "step": 9685
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1584574073210961e-05,
+      "loss": 1.1702,
+      "step": 9686
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1583035946594101e-05,
+      "loss": 1.0171,
+      "step": 9687
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1581497781561616e-05,
+      "loss": 1.0813,
+      "step": 9688
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1579959578150833e-05,
+      "loss": 1.0112,
+      "step": 9689
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1578421336399076e-05,
+      "loss": 1.2564,
+      "step": 9690
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1576883056343672e-05,
+      "loss": 1.1684,
+      "step": 9691
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1575344738021958e-05,
+      "loss": 1.0672,
+      "step": 9692
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1573806381471258e-05,
+      "loss": 1.1116,
+      "step": 9693
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1572267986728906e-05,
+      "loss": 1.2369,
+      "step": 9694
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1570729553832234e-05,
+      "loss": 1.0671,
+      "step": 9695
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1569191082818577e-05,
+      "loss": 1.0579,
+      "step": 9696
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1567652573725263e-05,
+      "loss": 0.9522,
+      "step": 9697
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1566114026589634e-05,
+      "loss": 1.0739,
+      "step": 9698
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1564575441449024e-05,
+      "loss": 1.127,
+      "step": 9699
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.156303681834077e-05,
+      "loss": 1.1193,
+      "step": 9700
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1561498157302205e-05,
+      "loss": 1.0595,
+      "step": 9701
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1559959458370674e-05,
+      "loss": 1.1575,
+      "step": 9702
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1558420721583514e-05,
+      "loss": 1.0777,
+      "step": 9703
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1556881946978067e-05,
+      "loss": 1.0524,
+      "step": 9704
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1555343134591674e-05,
+      "loss": 1.0857,
+      "step": 9705
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1553804284461679e-05,
+      "loss": 0.9842,
+      "step": 9706
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1552265396625421e-05,
+      "loss": 1.1565,
+      "step": 9707
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1550726471120246e-05,
+      "loss": 1.0552,
+      "step": 9708
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1549187507983504e-05,
+      "loss": 1.2338,
+      "step": 9709
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1547648507252534e-05,
+      "loss": 1.1636,
+      "step": 9710
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1546109468964688e-05,
+      "loss": 1.1256,
+      "step": 9711
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1544570393157311e-05,
+      "loss": 1.0641,
+      "step": 9712
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1543031279867754e-05,
+      "loss": 1.1474,
+      "step": 9713
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1541492129133362e-05,
+      "loss": 1.2028,
+      "step": 9714
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1539952940991495e-05,
+      "loss": 1.1941,
+      "step": 9715
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1538413715479495e-05,
+      "loss": 1.1889,
+      "step": 9716
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.153687445263472e-05,
+      "loss": 1.1714,
+      "step": 9717
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.153533515249452e-05,
+      "loss": 0.9716,
+      "step": 9718
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1533795815096254e-05,
+      "loss": 1.0081,
+      "step": 9719
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1532256440477273e-05,
+      "loss": 1.0896,
+      "step": 9720
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1530717028674934e-05,
+      "loss": 1.2096,
+      "step": 9721
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1529177579726593e-05,
+      "loss": 1.1713,
+      "step": 9722
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1527638093669613e-05,
+      "loss": 1.0473,
+      "step": 9723
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1526098570541347e-05,
+      "loss": 1.2737,
+      "step": 9724
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1524559010379156e-05,
+      "loss": 1.2379,
+      "step": 9725
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1523019413220402e-05,
+      "loss": 1.2279,
+      "step": 9726
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1521479779102447e-05,
+      "loss": 1.1186,
+      "step": 9727
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.151994010806265e-05,
+      "loss": 1.0154,
+      "step": 9728
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1518400400138376e-05,
+      "loss": 1.1309,
+      "step": 9729
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1516860655366993e-05,
+      "loss": 1.0082,
+      "step": 9730
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1515320873785862e-05,
+      "loss": 1.1786,
+      "step": 9731
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.151378105543235e-05,
+      "loss": 1.1904,
+      "step": 9732
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1512241200343821e-05,
+      "loss": 1.191,
+      "step": 9733
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1510701308557647e-05,
+      "loss": 1.0924,
+      "step": 9734
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1509161380111198e-05,
+      "loss": 1.1612,
+      "step": 9735
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1507621415041837e-05,
+      "loss": 1.0174,
+      "step": 9736
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1506081413386938e-05,
+      "loss": 1.202,
+      "step": 9737
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1504541375183873e-05,
+      "loss": 1.1123,
+      "step": 9738
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1503001300470016e-05,
+      "loss": 1.1009,
+      "step": 9739
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1501461189282734e-05,
+      "loss": 1.0987,
+      "step": 9740
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1499921041659407e-05,
+      "loss": 1.0978,
+      "step": 9741
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1498380857637405e-05,
+      "loss": 1.0801,
+      "step": 9742
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.149684063725411e-05,
+      "loss": 1.2221,
+      "step": 9743
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.149530038054689e-05,
+      "loss": 1.162,
+      "step": 9744
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1493760087553133e-05,
+      "loss": 1.1697,
+      "step": 9745
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1492219758310205e-05,
+      "loss": 0.9944,
+      "step": 9746
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1490679392855498e-05,
+      "loss": 1.0961,
+      "step": 9747
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.148913899122638e-05,
+      "loss": 1.2037,
+      "step": 9748
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1487598553460245e-05,
+      "loss": 0.957,
+      "step": 9749
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1486058079594461e-05,
+      "loss": 1.2985,
+      "step": 9750
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1484517569666425e-05,
+      "loss": 1.2384,
+      "step": 9751
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1482977023713506e-05,
+      "loss": 1.2028,
+      "step": 9752
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1481436441773102e-05,
+      "loss": 1.1664,
+      "step": 9753
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1479895823882588e-05,
+      "loss": 1.1622,
+      "step": 9754
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1478355170079358e-05,
+      "loss": 1.2473,
+      "step": 9755
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1476814480400792e-05,
+      "loss": 1.1474,
+      "step": 9756
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1475273754884287e-05,
+      "loss": 0.9501,
+      "step": 9757
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1473732993567222e-05,
+      "loss": 0.9478,
+      "step": 9758
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1472192196486993e-05,
+      "loss": 1.2417,
+      "step": 9759
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1470651363680987e-05,
+      "loss": 0.9678,
+      "step": 9760
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1469110495186602e-05,
+      "loss": 1.0053,
+      "step": 9761
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.146756959104122e-05,
+      "loss": 1.0382,
+      "step": 9762
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1466028651282244e-05,
+      "loss": 1.1576,
+      "step": 9763
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1464487675947062e-05,
+      "loss": 1.0977,
+      "step": 9764
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1462946665073073e-05,
+      "loss": 1.2239,
+      "step": 9765
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.146140561869767e-05,
+      "loss": 1.1575,
+      "step": 9766
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1459864536858247e-05,
+      "loss": 1.0752,
+      "step": 9767
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.145832341959221e-05,
+      "loss": 0.9831,
+      "step": 9768
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.145678226693695e-05,
+      "loss": 1.1034,
+      "step": 9769
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1455241078929868e-05,
+      "loss": 1.2127,
+      "step": 9770
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1453699855608367e-05,
+      "loss": 1.1176,
+      "step": 9771
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1452158597009843e-05,
+      "loss": 1.1339,
+      "step": 9772
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1450617303171704e-05,
+      "loss": 0.9704,
+      "step": 9773
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1449075974131347e-05,
+      "loss": 1.1168,
+      "step": 9774
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1447534609926179e-05,
+      "loss": 1.1482,
+      "step": 9775
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.14459932105936e-05,
+      "loss": 1.1686,
+      "step": 9776
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1444451776171021e-05,
+      "loss": 1.103,
+      "step": 9777
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1442910306695845e-05,
+      "loss": 1.1481,
+      "step": 9778
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.144136880220548e-05,
+      "loss": 1.1596,
+      "step": 9779
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.143982726273733e-05,
+      "loss": 0.9779,
+      "step": 9780
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1438285688328811e-05,
+      "loss": 1.2337,
+      "step": 9781
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1436744079017326e-05,
+      "loss": 1.1116,
+      "step": 9782
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1435202434840287e-05,
+      "loss": 1.213,
+      "step": 9783
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1433660755835105e-05,
+      "loss": 1.188,
+      "step": 9784
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1432119042039196e-05,
+      "loss": 1.182,
+      "step": 9785
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1430577293489966e-05,
+      "loss": 1.2854,
+      "step": 9786
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1429035510224833e-05,
+      "loss": 1.0083,
+      "step": 9787
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1427493692281213e-05,
+      "loss": 1.2014,
+      "step": 9788
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1425951839696517e-05,
+      "loss": 1.1781,
+      "step": 9789
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1424409952508165e-05,
+      "loss": 1.1082,
+      "step": 9790
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1422868030753569e-05,
+      "loss": 1.0396,
+      "step": 9791
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1421326074470156e-05,
+      "loss": 1.0636,
+      "step": 9792
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1419784083695334e-05,
+      "loss": 1.1335,
+      "step": 9793
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.141824205846653e-05,
+      "loss": 1.1752,
+      "step": 9794
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.141669999882116e-05,
+      "loss": 1.2212,
+      "step": 9795
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1415157904796652e-05,
+      "loss": 1.2165,
+      "step": 9796
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.141361577643042e-05,
+      "loss": 1.0862,
+      "step": 9797
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1412073613759894e-05,
+      "loss": 1.084,
+      "step": 9798
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1410531416822491e-05,
+      "loss": 1.2055,
+      "step": 9799
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.140898918565564e-05,
+      "loss": 1.1754,
+      "step": 9800
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1407446920296763e-05,
+      "loss": 1.2208,
+      "step": 9801
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1405904620783291e-05,
+      "loss": 1.1986,
+      "step": 9802
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1404362287152646e-05,
+      "loss": 1.0266,
+      "step": 9803
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.140281991944226e-05,
+      "loss": 1.0866,
+      "step": 9804
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.140127751768956e-05,
+      "loss": 1.1362,
+      "step": 9805
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1399735081931975e-05,
+      "loss": 1.0606,
+      "step": 9806
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1398192612206936e-05,
+      "loss": 1.0804,
+      "step": 9807
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1396650108551874e-05,
+      "loss": 1.0555,
+      "step": 9808
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1395107571004223e-05,
+      "loss": 1.1761,
+      "step": 9809
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1393564999601415e-05,
+      "loss": 1.2436,
+      "step": 9810
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1392022394380877e-05,
+      "loss": 1.1071,
+      "step": 9811
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1390479755380053e-05,
+      "loss": 1.1307,
+      "step": 9812
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1388937082636373e-05,
+      "loss": 1.2121,
+      "step": 9813
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1387394376187277e-05,
+      "loss": 1.1216,
+      "step": 9814
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.13858516360702e-05,
+      "loss": 1.1504,
+      "step": 9815
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1384308862322579e-05,
+      "loss": 1.0548,
+      "step": 9816
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1382766054981851e-05,
+      "loss": 1.0633,
+      "step": 9817
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.138122321408546e-05,
+      "loss": 1.0934,
+      "step": 9818
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1379680339670843e-05,
+      "loss": 1.0777,
+      "step": 9819
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.137813743177544e-05,
+      "loss": 1.1962,
+      "step": 9820
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1376594490436697e-05,
+      "loss": 1.1286,
+      "step": 9821
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1375051515692052e-05,
+      "loss": 1.1002,
+      "step": 9822
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1373508507578951e-05,
+      "loss": 1.0253,
+      "step": 9823
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.137196546613484e-05,
+      "loss": 1.1796,
+      "step": 9824
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1370422391397158e-05,
+      "loss": 1.1607,
+      "step": 9825
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.136887928340336e-05,
+      "loss": 1.2219,
+      "step": 9826
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1367336142190881e-05,
+      "loss": 1.1247,
+      "step": 9827
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1365792967797181e-05,
+      "loss": 1.1607,
+      "step": 9828
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1364249760259698e-05,
+      "loss": 1.0315,
+      "step": 9829
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1362706519615889e-05,
+      "loss": 1.3173,
+      "step": 9830
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1361163245903196e-05,
+      "loss": 1.1223,
+      "step": 9831
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1359619939159079e-05,
+      "loss": 0.9268,
+      "step": 9832
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1358076599420981e-05,
+      "loss": 1.0586,
+      "step": 9833
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1356533226726363e-05,
+      "loss": 1.1497,
+      "step": 9834
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1354989821112667e-05,
+      "loss": 1.1365,
+      "step": 9835
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1353446382617357e-05,
+      "loss": 1.0264,
+      "step": 9836
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1351902911277878e-05,
+      "loss": 1.064,
+      "step": 9837
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1350359407131699e-05,
+      "loss": 1.0554,
+      "step": 9838
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1348815870216263e-05,
+      "loss": 1.0096,
+      "step": 9839
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1347272300569034e-05,
+      "loss": 1.1441,
+      "step": 9840
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1345728698227467e-05,
+      "loss": 1.2331,
+      "step": 9841
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1344185063229025e-05,
+      "loss": 1.1012,
+      "step": 9842
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1342641395611162e-05,
+      "loss": 1.18,
+      "step": 9843
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1341097695411345e-05,
+      "loss": 1.2792,
+      "step": 9844
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1339553962667024e-05,
+      "loss": 1.1518,
+      "step": 9845
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1338010197415675e-05,
+      "loss": 1.2365,
+      "step": 9846
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.133646639969475e-05,
+      "loss": 1.1914,
+      "step": 9847
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1334922569541715e-05,
+      "loss": 1.0343,
+      "step": 9848
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1333378706994036e-05,
+      "loss": 1.1345,
+      "step": 9849
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.133183481208918e-05,
+      "loss": 1.1659,
+      "step": 9850
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1330290884864605e-05,
+      "loss": 1.2225,
+      "step": 9851
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1328746925357785e-05,
+      "loss": 1.0745,
+      "step": 9852
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1327202933606185e-05,
+      "loss": 1.0756,
+      "step": 9853
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.132565890964727e-05,
+      "loss": 1.1375,
+      "step": 9854
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1324114853518515e-05,
+      "loss": 1.0974,
+      "step": 9855
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1322570765257387e-05,
+      "loss": 1.07,
+      "step": 9856
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1321026644901353e-05,
+      "loss": 1.1613,
+      "step": 9857
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1319482492487893e-05,
+      "loss": 1.02,
+      "step": 9858
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1317938308054468e-05,
+      "loss": 1.0787,
+      "step": 9859
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1316394091638559e-05,
+      "loss": 1.1761,
+      "step": 9860
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1314849843277633e-05,
+      "loss": 1.2675,
+      "step": 9861
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1313305563009172e-05,
+      "loss": 1.0105,
+      "step": 9862
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1311761250870645e-05,
+      "loss": 1.0856,
+      "step": 9863
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1310216906899534e-05,
+      "loss": 1.2126,
+      "step": 9864
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1308672531133306e-05,
+      "loss": 1.1653,
+      "step": 9865
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.130712812360945e-05,
+      "loss": 1.2945,
+      "step": 9866
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1305583684365436e-05,
+      "loss": 1.0683,
+      "step": 9867
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1304039213438747e-05,
+      "loss": 1.155,
+      "step": 9868
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1302494710866859e-05,
+      "loss": 1.1987,
+      "step": 9869
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1300950176687255e-05,
+      "loss": 1.2151,
+      "step": 9870
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1299405610937417e-05,
+      "loss": 1.1411,
+      "step": 9871
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1297861013654827e-05,
+      "loss": 1.1502,
+      "step": 9872
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1296316384876966e-05,
+      "loss": 1.2313,
+      "step": 9873
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1294771724641318e-05,
+      "loss": 1.1859,
+      "step": 9874
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.129322703298537e-05,
+      "loss": 1.0049,
+      "step": 9875
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1291682309946604e-05,
+      "loss": 1.3852,
+      "step": 9876
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1290137555562507e-05,
+      "loss": 1.2137,
+      "step": 9877
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1288592769870565e-05,
+      "loss": 1.1121,
+      "step": 9878
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1287047952908268e-05,
+      "loss": 1.0288,
+      "step": 9879
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.12855031047131e-05,
+      "loss": 1.107,
+      "step": 9880
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1283958225322556e-05,
+      "loss": 1.0552,
+      "step": 9881
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1282413314774117e-05,
+      "loss": 1.1522,
+      "step": 9882
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1280868373105282e-05,
+      "loss": 1.2325,
+      "step": 9883
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1279323400353536e-05,
+      "loss": 1.0901,
+      "step": 9884
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1277778396556376e-05,
+      "loss": 1.2108,
+      "step": 9885
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.127623336175129e-05,
+      "loss": 1.1489,
+      "step": 9886
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1274688295975775e-05,
+      "loss": 1.0766,
+      "step": 9887
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1273143199267323e-05,
+      "loss": 1.2125,
+      "step": 9888
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.127159807166343e-05,
+      "loss": 1.0274,
+      "step": 9889
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.127005291320159e-05,
+      "loss": 1.1991,
+      "step": 9890
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1268507723919302e-05,
+      "loss": 1.088,
+      "step": 9891
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1266962503854062e-05,
+      "loss": 1.1476,
+      "step": 9892
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1265417253043368e-05,
+      "loss": 1.0739,
+      "step": 9893
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1263871971524718e-05,
+      "loss": 1.138,
+      "step": 9894
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1262326659335611e-05,
+      "loss": 1.0397,
+      "step": 9895
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1260781316513551e-05,
+      "loss": 1.0729,
+      "step": 9896
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1259235943096036e-05,
+      "loss": 1.0997,
+      "step": 9897
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1257690539120567e-05,
+      "loss": 1.2055,
+      "step": 9898
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.125614510462465e-05,
+      "loss": 1.1216,
+      "step": 9899
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1254599639645783e-05,
+      "loss": 1.1573,
+      "step": 9900
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1253054144221474e-05,
+      "loss": 1.1427,
+      "step": 9901
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1251508618389221e-05,
+      "loss": 1.1657,
+      "step": 9902
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1249963062186541e-05,
+      "loss": 1.1398,
+      "step": 9903
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.124841747565093e-05,
+      "loss": 1.0639,
+      "step": 9904
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1246871858819902e-05,
+      "loss": 1.1318,
+      "step": 9905
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1245326211730956e-05,
+      "loss": 1.202,
+      "step": 9906
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1243780534421607e-05,
+      "loss": 1.2528,
+      "step": 9907
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.124223482692936e-05,
+      "loss": 1.1867,
+      "step": 9908
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1240689089291732e-05,
+      "loss": 1.137,
+      "step": 9909
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1239143321546226e-05,
+      "loss": 1.2295,
+      "step": 9910
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1237597523730357e-05,
+      "loss": 1.1148,
+      "step": 9911
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1236051695881634e-05,
+      "loss": 1.0871,
+      "step": 9912
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1234505838037574e-05,
+      "loss": 1.1182,
+      "step": 9913
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1232959950235682e-05,
+      "loss": 1.2558,
+      "step": 9914
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1231414032513484e-05,
+      "loss": 1.109,
+      "step": 9915
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1229868084908487e-05,
+      "loss": 0.9588,
+      "step": 9916
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1228322107458207e-05,
+      "loss": 1.1636,
+      "step": 9917
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1226776100200162e-05,
+      "loss": 1.1088,
+      "step": 9918
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1225230063171872e-05,
+      "loss": 1.1536,
+      "step": 9919
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1223683996410847e-05,
+      "loss": 1.1272,
+      "step": 9920
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1222137899954615e-05,
+      "loss": 1.293,
+      "step": 9921
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1220591773840687e-05,
+      "loss": 1.0631,
+      "step": 9922
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.121904561810659e-05,
+      "loss": 1.174,
+      "step": 9923
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1217499432789836e-05,
+      "loss": 1.1561,
+      "step": 9924
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1215953217927956e-05,
+      "loss": 1.066,
+      "step": 9925
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1214406973558462e-05,
+      "loss": 1.1911,
+      "step": 9926
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1212860699718888e-05,
+      "loss": 1.0416,
+      "step": 9927
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1211314396446746e-05,
+      "loss": 1.2655,
+      "step": 9928
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1209768063779572e-05,
+      "loss": 1.1086,
+      "step": 9929
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.120822170175488e-05,
+      "loss": 1.0951,
+      "step": 9930
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1206675310410201e-05,
+      "loss": 1.2194,
+      "step": 9931
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1205128889783063e-05,
+      "loss": 1.1078,
+      "step": 9932
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.120358243991099e-05,
+      "loss": 1.1028,
+      "step": 9933
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1202035960831512e-05,
+      "loss": 1.2414,
+      "step": 9934
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1200489452582156e-05,
+      "loss": 1.0806,
+      "step": 9935
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.119894291520045e-05,
+      "loss": 1.2448,
+      "step": 9936
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1197396348723923e-05,
+      "loss": 0.9798,
+      "step": 9937
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.119584975319011e-05,
+      "loss": 1.1922,
+      "step": 9938
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1194303128636543e-05,
+      "loss": 1.1461,
+      "step": 9939
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.119275647510075e-05,
+      "loss": 1.1824,
+      "step": 9940
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1191209792620266e-05,
+      "loss": 1.1351,
+      "step": 9941
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1189663081232622e-05,
+      "loss": 1.0702,
+      "step": 9942
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1188116340975355e-05,
+      "loss": 1.0958,
+      "step": 9943
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1186569571885996e-05,
+      "loss": 1.1885,
+      "step": 9944
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1185022774002087e-05,
+      "loss": 1.2514,
+      "step": 9945
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1183475947361158e-05,
+      "loss": 1.236,
+      "step": 9946
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.118192909200075e-05,
+      "loss": 1.0883,
+      "step": 9947
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1180382207958396e-05,
+      "loss": 1.2902,
+      "step": 9948
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.117883529527164e-05,
+      "loss": 1.1725,
+      "step": 9949
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1177288353978016e-05,
+      "loss": 1.1518,
+      "step": 9950
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1175741384115067e-05,
+      "loss": 1.0391,
+      "step": 9951
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.117419438572033e-05,
+      "loss": 1.0863,
+      "step": 9952
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1172647358831351e-05,
+      "loss": 1.0181,
+      "step": 9953
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.117110030348567e-05,
+      "loss": 1.0958,
+      "step": 9954
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1169553219720828e-05,
+      "loss": 1.2079,
+      "step": 9955
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1168006107574365e-05,
+      "loss": 1.1437,
+      "step": 9956
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1166458967083836e-05,
+      "loss": 1.017,
+      "step": 9957
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1164911798286771e-05,
+      "loss": 1.0053,
+      "step": 9958
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1163364601220725e-05,
+      "loss": 1.1448,
+      "step": 9959
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1161817375923243e-05,
+      "loss": 1.1361,
+      "step": 9960
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1160270122431869e-05,
+      "loss": 1.2669,
+      "step": 9961
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1158722840784154e-05,
+      "loss": 1.1994,
+      "step": 9962
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1157175531017639e-05,
+      "loss": 1.0539,
+      "step": 9963
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.115562819316988e-05,
+      "loss": 1.0825,
+      "step": 9964
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1154080827278422e-05,
+      "loss": 1.1577,
+      "step": 9965
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1152533433380817e-05,
+      "loss": 1.1682,
+      "step": 9966
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1150986011514616e-05,
+      "loss": 1.1406,
+      "step": 9967
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1149438561717367e-05,
+      "loss": 1.0674,
+      "step": 9968
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1147891084026625e-05,
+      "loss": 1.087,
+      "step": 9969
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1146343578479943e-05,
+      "loss": 1.1436,
+      "step": 9970
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1144796045114873e-05,
+      "loss": 1.1742,
+      "step": 9971
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1143248483968971e-05,
+      "loss": 1.0433,
+      "step": 9972
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.114170089507979e-05,
+      "loss": 1.2331,
+      "step": 9973
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1140153278484887e-05,
+      "loss": 1.1278,
+      "step": 9974
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1138605634221815e-05,
+      "loss": 1.1252,
+      "step": 9975
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1137057962328136e-05,
+      "loss": 1.0376,
+      "step": 9976
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1135510262841397e-05,
+      "loss": 1.1408,
+      "step": 9977
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.113396253579917e-05,
+      "loss": 1.2035,
+      "step": 9978
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1132414781239003e-05,
+      "loss": 1.2952,
+      "step": 9979
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1130866999198461e-05,
+      "loss": 1.3025,
+      "step": 9980
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1129319189715105e-05,
+      "loss": 1.0899,
+      "step": 9981
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1127771352826491e-05,
+      "loss": 1.1804,
+      "step": 9982
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1126223488570183e-05,
+      "loss": 1.1031,
+      "step": 9983
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1124675596983744e-05,
+      "loss": 1.4009,
+      "step": 9984
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1123127678104734e-05,
+      "loss": 1.1348,
+      "step": 9985
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1121579731970722e-05,
+      "loss": 1.1058,
+      "step": 9986
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1120031758619265e-05,
+      "loss": 1.2384,
+      "step": 9987
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1118483758087934e-05,
+      "loss": 1.2725,
+      "step": 9988
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1116935730414289e-05,
+      "loss": 1.1139,
+      "step": 9989
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.11153876756359e-05,
+      "loss": 1.1691,
+      "step": 9990
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1113839593790332e-05,
+      "loss": 1.0867,
+      "step": 9991
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1112291484915153e-05,
+      "loss": 1.219,
+      "step": 9992
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.111074334904793e-05,
+      "loss": 1.1764,
+      "step": 9993
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1109195186226234e-05,
+      "loss": 1.0764,
+      "step": 9994
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1107646996487631e-05,
+      "loss": 1.0744,
+      "step": 9995
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1106098779869694e-05,
+      "loss": 1.0323,
+      "step": 9996
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.110455053640999e-05,
+      "loss": 0.9343,
+      "step": 9997
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1103002266146096e-05,
+      "loss": 1.0906,
+      "step": 9998
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1101453969115579e-05,
+      "loss": 0.9138,
+      "step": 9999
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1099905645356016e-05,
+      "loss": 1.1961,
+      "step": 10000
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1098357294904976e-05,
+      "loss": 1.1228,
+      "step": 10001
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1096808917800035e-05,
+      "loss": 1.1756,
+      "step": 10002
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1095260514078766e-05,
+      "loss": 1.2702,
+      "step": 10003
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1093712083778748e-05,
+      "loss": 1.1958,
+      "step": 10004
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1092163626937549e-05,
+      "loss": 1.1386,
+      "step": 10005
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1090615143592755e-05,
+      "loss": 0.9504,
+      "step": 10006
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1089066633781937e-05,
+      "loss": 1.1392,
+      "step": 10007
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1087518097542677e-05,
+      "loss": 1.098,
+      "step": 10008
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1085969534912548e-05,
+      "loss": 1.0777,
+      "step": 10009
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1084420945929136e-05,
+      "loss": 1.1729,
+      "step": 10010
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1082872330630013e-05,
+      "loss": 1.1847,
+      "step": 10011
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1081323689052766e-05,
+      "loss": 1.184,
+      "step": 10012
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1079775021234969e-05,
+      "loss": 1.1826,
+      "step": 10013
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1078226327214212e-05,
+      "loss": 1.1784,
+      "step": 10014
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.107667760702807e-05,
+      "loss": 1.0288,
+      "step": 10015
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1075128860714132e-05,
+      "loss": 1.1425,
+      "step": 10016
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1073580088309977e-05,
+      "loss": 1.0321,
+      "step": 10017
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1072031289853189e-05,
+      "loss": 1.1973,
+      "step": 10018
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1070482465381356e-05,
+      "loss": 0.9389,
+      "step": 10019
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1068933614932061e-05,
+      "loss": 1.2148,
+      "step": 10020
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1067384738542888e-05,
+      "loss": 1.142,
+      "step": 10021
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.106583583625143e-05,
+      "loss": 1.0403,
+      "step": 10022
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.106428690809527e-05,
+      "loss": 1.1264,
+      "step": 10023
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1062737954112e-05,
+      "loss": 1.2057,
+      "step": 10024
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1061188974339201e-05,
+      "loss": 1.1813,
+      "step": 10025
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1059639968814469e-05,
+      "loss": 1.1862,
+      "step": 10026
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1058090937575388e-05,
+      "loss": 1.1607,
+      "step": 10027
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1056541880659554e-05,
+      "loss": 1.1979,
+      "step": 10028
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1054992798104553e-05,
+      "loss": 1.1734,
+      "step": 10029
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1053443689947984e-05,
+      "loss": 1.1133,
+      "step": 10030
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1051894556227431e-05,
+      "loss": 1.0814,
+      "step": 10031
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1050345396980493e-05,
+      "loss": 1.1051,
+      "step": 10032
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1048796212244761e-05,
+      "loss": 0.9979,
+      "step": 10033
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1047247002057828e-05,
+      "loss": 1.0931,
+      "step": 10034
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.104569776645729e-05,
+      "loss": 1.1046,
+      "step": 10035
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1044148505480744e-05,
+      "loss": 1.1097,
+      "step": 10036
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1042599219165784e-05,
+      "loss": 1.1576,
+      "step": 10037
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1041049907550008e-05,
+      "loss": 1.1914,
+      "step": 10038
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.103950057067101e-05,
+      "loss": 1.2781,
+      "step": 10039
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1037951208566392e-05,
+      "loss": 1.1264,
+      "step": 10040
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.103640182127375e-05,
+      "loss": 1.1072,
+      "step": 10041
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1034852408830683e-05,
+      "loss": 1.0805,
+      "step": 10042
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1033302971274793e-05,
+      "loss": 1.0204,
+      "step": 10043
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.103175350864368e-05,
+      "loss": 1.1853,
+      "step": 10044
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1030204020974939e-05,
+      "loss": 1.1165,
+      "step": 10045
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1028654508306178e-05,
+      "loss": 1.2244,
+      "step": 10046
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1027104970674998e-05,
+      "loss": 0.9817,
+      "step": 10047
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1025555408119e-05,
+      "loss": 1.0951,
+      "step": 10048
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1024005820675791e-05,
+      "loss": 1.0653,
+      "step": 10049
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.102245620838297e-05,
+      "loss": 1.0651,
+      "step": 10050
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1020906571278145e-05,
+      "loss": 1.111,
+      "step": 10051
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1019356909398917e-05,
+      "loss": 1.1209,
+      "step": 10052
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.10178072227829e-05,
+      "loss": 0.956,
+      "step": 10053
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.101625751146769e-05,
+      "loss": 1.1762,
+      "step": 10054
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.10147077754909e-05,
+      "loss": 1.0487,
+      "step": 10055
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1013158014890138e-05,
+      "loss": 0.9849,
+      "step": 10056
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1011608229703011e-05,
+      "loss": 1.1619,
+      "step": 10057
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1010058419967123e-05,
+      "loss": 1.0385,
+      "step": 10058
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1008508585720093e-05,
+      "loss": 1.2202,
+      "step": 10059
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1006958726999522e-05,
+      "loss": 1.2406,
+      "step": 10060
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1005408843843026e-05,
+      "loss": 1.1905,
+      "step": 10061
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1003858936288212e-05,
+      "loss": 1.1438,
+      "step": 10062
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1002309004372693e-05,
+      "loss": 1.0424,
+      "step": 10063
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1000759048134085e-05,
+      "loss": 1.0925,
+      "step": 10064
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0999209067609998e-05,
+      "loss": 1.0197,
+      "step": 10065
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0997659062838041e-05,
+      "loss": 1.1801,
+      "step": 10066
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.099610903385584e-05,
+      "loss": 1.0597,
+      "step": 10067
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0994558980700997e-05,
+      "loss": 1.053,
+      "step": 10068
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0993008903411135e-05,
+      "loss": 1.1204,
+      "step": 10069
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0991458802023864e-05,
+      "loss": 1.1791,
+      "step": 10070
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0989908676576808e-05,
+      "loss": 1.1323,
+      "step": 10071
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0988358527107576e-05,
+      "loss": 1.0753,
+      "step": 10072
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0986808353653792e-05,
+      "loss": 1.0807,
+      "step": 10073
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0985258156253069e-05,
+      "loss": 1.0567,
+      "step": 10074
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0983707934943031e-05,
+      "loss": 1.1358,
+      "step": 10075
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0982157689761293e-05,
+      "loss": 0.9558,
+      "step": 10076
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0980607420745475e-05,
+      "loss": 1.1597,
+      "step": 10077
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0979057127933199e-05,
+      "loss": 1.2041,
+      "step": 10078
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0977506811362089e-05,
+      "loss": 1.1469,
+      "step": 10079
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0975956471069758e-05,
+      "loss": 1.0313,
+      "step": 10080
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0974406107093837e-05,
+      "loss": 1.0736,
+      "step": 10081
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0972855719471943e-05,
+      "loss": 1.3332,
+      "step": 10082
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0971305308241706e-05,
+      "loss": 1.1161,
+      "step": 10083
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0969754873440743e-05,
+      "loss": 1.215,
+      "step": 10084
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0968204415106682e-05,
+      "loss": 1.1242,
+      "step": 10085
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0966653933277147e-05,
+      "loss": 1.046,
+      "step": 10086
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0965103427989766e-05,
+      "loss": 1.1344,
+      "step": 10087
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0963552899282158e-05,
+      "loss": 1.1078,
+      "step": 10088
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0962002347191958e-05,
+      "loss": 1.1989,
+      "step": 10089
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0960451771756788e-05,
+      "loss": 1.1665,
+      "step": 10090
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0958901173014283e-05,
+      "loss": 1.2412,
+      "step": 10091
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0957350551002062e-05,
+      "loss": 0.9811,
+      "step": 10092
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.095579990575776e-05,
+      "loss": 1.1542,
+      "step": 10093
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0954249237319003e-05,
+      "loss": 1.2205,
+      "step": 10094
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0952698545723428e-05,
+      "loss": 1.0291,
+      "step": 10095
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0951147831008657e-05,
+      "loss": 1.0888,
+      "step": 10096
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.094959709321233e-05,
+      "loss": 1.299,
+      "step": 10097
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.094804633237207e-05,
+      "loss": 1.2805,
+      "step": 10098
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0946495548525515e-05,
+      "loss": 1.2537,
+      "step": 10099
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0944944741710295e-05,
+      "loss": 1.2493,
+      "step": 10100
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0943393911964049e-05,
+      "loss": 1.2164,
+      "step": 10101
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0941843059324402e-05,
+      "loss": 1.1831,
+      "step": 10102
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0940292183828998e-05,
+      "loss": 1.1374,
+      "step": 10103
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0938741285515463e-05,
+      "loss": 1.0275,
+      "step": 10104
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0937190364421442e-05,
+      "loss": 1.1357,
+      "step": 10105
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0935639420584565e-05,
+      "loss": 1.2404,
+      "step": 10106
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0934088454042474e-05,
+      "loss": 1.189,
+      "step": 10107
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.09325374648328e-05,
+      "loss": 1.1787,
+      "step": 10108
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0930986452993187e-05,
+      "loss": 1.2213,
+      "step": 10109
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0929435418561268e-05,
+      "loss": 1.167,
+      "step": 10110
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0927884361574688e-05,
+      "loss": 1.087,
+      "step": 10111
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.092633328207108e-05,
+      "loss": 1.2069,
+      "step": 10112
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.092478218008809e-05,
+      "loss": 1.1477,
+      "step": 10113
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0923231055663352e-05,
+      "loss": 1.2357,
+      "step": 10114
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0921679908834518e-05,
+      "loss": 1.0762,
+      "step": 10115
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0920128739639218e-05,
+      "loss": 1.1102,
+      "step": 10116
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0918577548115101e-05,
+      "loss": 1.2193,
+      "step": 10117
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.091702633429981e-05,
+      "loss": 1.136,
+      "step": 10118
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0915475098230986e-05,
+      "loss": 1.0284,
+      "step": 10119
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.091392383994627e-05,
+      "loss": 1.2034,
+      "step": 10120
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0912372559483314e-05,
+      "loss": 0.969,
+      "step": 10121
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0910821256879758e-05,
+      "loss": 1.1624,
+      "step": 10122
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.090926993217325e-05,
+      "loss": 1.1208,
+      "step": 10123
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0907718585401432e-05,
+      "loss": 1.0701,
+      "step": 10124
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0906167216601958e-05,
+      "loss": 1.2142,
+      "step": 10125
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0904615825812466e-05,
+      "loss": 1.1549,
+      "step": 10126
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0903064413070611e-05,
+      "loss": 1.0806,
+      "step": 10127
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0901512978414037e-05,
+      "loss": 1.0207,
+      "step": 10128
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0899961521880396e-05,
+      "loss": 1.2428,
+      "step": 10129
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0898410043507335e-05,
+      "loss": 1.1912,
+      "step": 10130
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0896858543332502e-05,
+      "loss": 1.0005,
+      "step": 10131
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0895307021393553e-05,
+      "loss": 1.2473,
+      "step": 10132
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0893755477728131e-05,
+      "loss": 1.1156,
+      "step": 10133
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0892203912373898e-05,
+      "loss": 1.0496,
+      "step": 10134
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0890652325368495e-05,
+      "loss": 1.0678,
+      "step": 10135
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0889100716749584e-05,
+      "loss": 1.1094,
+      "step": 10136
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.088754908655481e-05,
+      "loss": 0.9671,
+      "step": 10137
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0885997434821831e-05,
+      "loss": 1.1438,
+      "step": 10138
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0884445761588299e-05,
+      "loss": 1.0456,
+      "step": 10139
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.088289406689187e-05,
+      "loss": 1.0169,
+      "step": 10140
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0881342350770196e-05,
+      "loss": 1.0727,
+      "step": 10141
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.087979061326094e-05,
+      "loss": 1.0949,
+      "step": 10142
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.087823885440175e-05,
+      "loss": 1.1871,
+      "step": 10143
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0876687074230288e-05,
+      "loss": 1.1294,
+      "step": 10144
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0875135272784205e-05,
+      "loss": 1.2937,
+      "step": 10145
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0873583450101166e-05,
+      "loss": 1.0841,
+      "step": 10146
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0872031606218824e-05,
+      "loss": 1.1081,
+      "step": 10147
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.087047974117484e-05,
+      "loss": 1.2079,
+      "step": 10148
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0868927855006874e-05,
+      "loss": 1.122,
+      "step": 10149
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0867375947752583e-05,
+      "loss": 1.1841,
+      "step": 10150
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0865824019449627e-05,
+      "loss": 1.1434,
+      "step": 10151
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0864272070135674e-05,
+      "loss": 1.0849,
+      "step": 10152
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0862720099848374e-05,
+      "loss": 1.193,
+      "step": 10153
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0861168108625397e-05,
+      "loss": 1.0771,
+      "step": 10154
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0859616096504399e-05,
+      "loss": 1.1469,
+      "step": 10155
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0858064063523053e-05,
+      "loss": 1.2003,
+      "step": 10156
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.085651200971901e-05,
+      "loss": 1.2059,
+      "step": 10157
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0854959935129944e-05,
+      "loss": 1.0675,
+      "step": 10158
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0853407839793508e-05,
+      "loss": 1.169,
+      "step": 10159
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.085185572374738e-05,
+      "loss": 1.0948,
+      "step": 10160
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0850303587029216e-05,
+      "loss": 0.9791,
+      "step": 10161
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0848751429676685e-05,
+      "loss": 1.083,
+      "step": 10162
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0847199251727452e-05,
+      "loss": 1.1972,
+      "step": 10163
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0845647053219185e-05,
+      "loss": 1.0511,
+      "step": 10164
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0844094834189553e-05,
+      "loss": 1.1637,
+      "step": 10165
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0842542594676221e-05,
+      "loss": 1.1978,
+      "step": 10166
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0840990334716856e-05,
+      "loss": 1.0189,
+      "step": 10167
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0839438054349131e-05,
+      "loss": 1.2352,
+      "step": 10168
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0837885753610713e-05,
+      "loss": 0.9963,
+      "step": 10169
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0836333432539272e-05,
+      "loss": 1.1437,
+      "step": 10170
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0834781091172475e-05,
+      "loss": 1.137,
+      "step": 10171
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0833228729548e-05,
+      "loss": 1.2247,
+      "step": 10172
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0831676347703512e-05,
+      "loss": 1.0519,
+      "step": 10173
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0830123945676688e-05,
+      "loss": 0.9482,
+      "step": 10174
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0828571523505192e-05,
+      "loss": 1.0486,
+      "step": 10175
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0827019081226706e-05,
+      "loss": 1.0958,
+      "step": 10176
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0825466618878896e-05,
+      "loss": 1.0826,
+      "step": 10177
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.082391413649944e-05,
+      "loss": 1.1833,
+      "step": 10178
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.082236163412601e-05,
+      "loss": 1.0484,
+      "step": 10179
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0820809111796285e-05,
+      "loss": 1.1889,
+      "step": 10180
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0819256569547933e-05,
+      "loss": 0.9207,
+      "step": 10181
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0817704007418637e-05,
+      "loss": 1.1824,
+      "step": 10182
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0816151425446066e-05,
+      "loss": 1.0902,
+      "step": 10183
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0814598823667904e-05,
+      "loss": 1.0125,
+      "step": 10184
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.081304620212182e-05,
+      "loss": 1.0343,
+      "step": 10185
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0811493560845499e-05,
+      "loss": 1.0275,
+      "step": 10186
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.080994089987661e-05,
+      "loss": 1.2206,
+      "step": 10187
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0808388219252844e-05,
+      "loss": 1.1765,
+      "step": 10188
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.080683551901187e-05,
+      "loss": 1.0989,
+      "step": 10189
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0805282799191372e-05,
+      "loss": 1.1709,
+      "step": 10190
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0803730059829027e-05,
+      "loss": 1.1789,
+      "step": 10191
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0802177300962519e-05,
+      "loss": 1.1888,
+      "step": 10192
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0800624522629529e-05,
+      "loss": 1.0748,
+      "step": 10193
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0799071724867734e-05,
+      "loss": 1.1736,
+      "step": 10194
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0797518907714817e-05,
+      "loss": 1.0677,
+      "step": 10195
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0795966071208464e-05,
+      "loss": 1.1982,
+      "step": 10196
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0794413215386352e-05,
+      "loss": 1.0838,
+      "step": 10197
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0792860340286171e-05,
+      "loss": 1.1441,
+      "step": 10198
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0791307445945598e-05,
+      "loss": 0.9754,
+      "step": 10199
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0789754532402324e-05,
+      "loss": 1.14,
+      "step": 10200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0788201599694028e-05,
+      "loss": 0.9996,
+      "step": 10201
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0786648647858397e-05,
+      "loss": 1.1601,
+      "step": 10202
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0785095676933118e-05,
+      "loss": 1.098,
+      "step": 10203
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0783542686955876e-05,
+      "loss": 1.1615,
+      "step": 10204
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0781989677964355e-05,
+      "loss": 1.0048,
+      "step": 10205
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0780436649996246e-05,
+      "loss": 1.0333,
+      "step": 10206
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0778883603089233e-05,
+      "loss": 0.8762,
+      "step": 10207
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.077733053728101e-05,
+      "loss": 1.1062,
+      "step": 10208
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0775777452609257e-05,
+      "loss": 1.2586,
+      "step": 10209
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0774224349111668e-05,
+      "loss": 1.2329,
+      "step": 10210
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.077267122682593e-05,
+      "loss": 1.102,
+      "step": 10211
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0771118085789735e-05,
+      "loss": 1.0943,
+      "step": 10212
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.076956492604077e-05,
+      "loss": 1.0809,
+      "step": 10213
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0768011747616732e-05,
+      "loss": 1.1508,
+      "step": 10214
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0766458550555301e-05,
+      "loss": 1.3088,
+      "step": 10215
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0764905334894177e-05,
+      "loss": 1.146,
+      "step": 10216
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0763352100671051e-05,
+      "loss": 1.077,
+      "step": 10217
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0761798847923615e-05,
+      "loss": 1.1586,
+      "step": 10218
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0760245576689564e-05,
+      "loss": 1.0497,
+      "step": 10219
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0758692287006583e-05,
+      "loss": 1.0917,
+      "step": 10220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0757138978912374e-05,
+      "loss": 0.9745,
+      "step": 10221
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0755585652444629e-05,
+      "loss": 1.1194,
+      "step": 10222
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0754032307641043e-05,
+      "loss": 1.0353,
+      "step": 10223
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0752478944539308e-05,
+      "loss": 1.0049,
+      "step": 10224
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0750925563177123e-05,
+      "loss": 1.1875,
+      "step": 10225
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0749372163592185e-05,
+      "loss": 1.044,
+      "step": 10226
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0747818745822186e-05,
+      "loss": 1.1063,
+      "step": 10227
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0746265309904827e-05,
+      "loss": 1.1836,
+      "step": 10228
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0744711855877802e-05,
+      "loss": 1.0552,
+      "step": 10229
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0743158383778811e-05,
+      "loss": 1.0841,
+      "step": 10230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.074160489364555e-05,
+      "loss": 1.3037,
+      "step": 10231
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0740051385515721e-05,
+      "loss": 1.12,
+      "step": 10232
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0738497859427022e-05,
+      "loss": 1.1123,
+      "step": 10233
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.073694431541715e-05,
+      "loss": 1.2056,
+      "step": 10234
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0735390753523806e-05,
+      "loss": 1.2445,
+      "step": 10235
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0733837173784694e-05,
+      "loss": 1.202,
+      "step": 10236
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.073228357623751e-05,
+      "loss": 1.1158,
+      "step": 10237
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0730729960919959e-05,
+      "loss": 0.974,
+      "step": 10238
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0729176327869738e-05,
+      "loss": 1.1654,
+      "step": 10239
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0727622677124554e-05,
+      "loss": 1.1118,
+      "step": 10240
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0726069008722109e-05,
+      "loss": 1.0011,
+      "step": 10241
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0724515322700101e-05,
+      "loss": 1.1861,
+      "step": 10242
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.072296161909624e-05,
+      "loss": 1.1044,
+      "step": 10243
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0721407897948224e-05,
+      "loss": 1.1651,
+      "step": 10244
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0719854159293761e-05,
+      "loss": 1.0087,
+      "step": 10245
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0718300403170555e-05,
+      "loss": 1.0139,
+      "step": 10246
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.071674662961631e-05,
+      "loss": 1.1081,
+      "step": 10247
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0715192838668732e-05,
+      "loss": 1.2281,
+      "step": 10248
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0713639030365528e-05,
+      "loss": 1.0633,
+      "step": 10249
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0712085204744402e-05,
+      "loss": 1.254,
+      "step": 10250
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0710531361843066e-05,
+      "loss": 1.1343,
+      "step": 10251
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0708977501699218e-05,
+      "loss": 1.1793,
+      "step": 10252
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0707423624350577e-05,
+      "loss": 1.1513,
+      "step": 10253
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0705869729834838e-05,
+      "loss": 1.1433,
+      "step": 10254
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0704315818189723e-05,
+      "loss": 1.2545,
+      "step": 10255
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.070276188945293e-05,
+      "loss": 1.313,
+      "step": 10256
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0701207943662177e-05,
+      "loss": 0.9278,
+      "step": 10257
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0699653980855164e-05,
+      "loss": 1.1083,
+      "step": 10258
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0698100001069611e-05,
+      "loss": 1.0701,
+      "step": 10259
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0696546004343222e-05,
+      "loss": 1.1202,
+      "step": 10260
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0694991990713712e-05,
+      "loss": 1.0672,
+      "step": 10261
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0693437960218787e-05,
+      "loss": 0.9517,
+      "step": 10262
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0691883912896166e-05,
+      "loss": 1.1114,
+      "step": 10263
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0690329848783554e-05,
+      "loss": 1.1193,
+      "step": 10264
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0688775767918667e-05,
+      "loss": 0.9713,
+      "step": 10265
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0687221670339219e-05,
+      "loss": 1.1149,
+      "step": 10266
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0685667556082922e-05,
+      "loss": 1.0065,
+      "step": 10267
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0684113425187487e-05,
+      "loss": 1.0461,
+      "step": 10268
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0682559277690636e-05,
+      "loss": 1.0776,
+      "step": 10269
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0681005113630072e-05,
+      "loss": 1.0717,
+      "step": 10270
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0679450933043522e-05,
+      "loss": 1.2306,
+      "step": 10271
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0677896735968694e-05,
+      "loss": 1.1093,
+      "step": 10272
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0676342522443308e-05,
+      "loss": 1.1802,
+      "step": 10273
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0674788292505073e-05,
+      "loss": 1.2002,
+      "step": 10274
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0673234046191716e-05,
+      "loss": 1.2459,
+      "step": 10275
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0671679783540947e-05,
+      "loss": 1.0866,
+      "step": 10276
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0670125504590484e-05,
+      "loss": 1.0718,
+      "step": 10277
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0668571209378048e-05,
+      "loss": 1.2936,
+      "step": 10278
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0667016897941354e-05,
+      "loss": 1.2051,
+      "step": 10279
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0665462570318121e-05,
+      "loss": 1.0295,
+      "step": 10280
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.066390822654607e-05,
+      "loss": 1.1635,
+      "step": 10281
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0662353866662915e-05,
+      "loss": 1.1177,
+      "step": 10282
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0660799490706383e-05,
+      "loss": 1.0973,
+      "step": 10283
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0659245098714188e-05,
+      "loss": 1.0692,
+      "step": 10284
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0657690690724056e-05,
+      "loss": 0.9915,
+      "step": 10285
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0656136266773704e-05,
+      "loss": 1.2108,
+      "step": 10286
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0654581826900855e-05,
+      "loss": 0.9511,
+      "step": 10287
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.065302737114323e-05,
+      "loss": 1.215,
+      "step": 10288
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.065147289953855e-05,
+      "loss": 1.1372,
+      "step": 10289
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.064991841212454e-05,
+      "loss": 1.2562,
+      "step": 10290
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0648363908938921e-05,
+      "loss": 1.2096,
+      "step": 10291
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0646809390019416e-05,
+      "loss": 1.1474,
+      "step": 10292
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.064525485540375e-05,
+      "loss": 0.9632,
+      "step": 10293
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0643700305129644e-05,
+      "loss": 1.1943,
+      "step": 10294
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.064214573923483e-05,
+      "loss": 1.1398,
+      "step": 10295
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0640591157757023e-05,
+      "loss": 1.1119,
+      "step": 10296
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0639036560733956e-05,
+      "loss": 1.0402,
+      "step": 10297
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0637481948203348e-05,
+      "loss": 0.9915,
+      "step": 10298
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0635927320202928e-05,
+      "loss": 1.1607,
+      "step": 10299
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0634372676770422e-05,
+      "loss": 1.2135,
+      "step": 10300
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0632818017943558e-05,
+      "loss": 1.0816,
+      "step": 10301
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0631263343760061e-05,
+      "loss": 1.2445,
+      "step": 10302
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0629708654257658e-05,
+      "loss": 1.1487,
+      "step": 10303
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.062815394947408e-05,
+      "loss": 1.2801,
+      "step": 10304
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0626599229447053e-05,
+      "loss": 1.0796,
+      "step": 10305
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0625044494214305e-05,
+      "loss": 1.1825,
+      "step": 10306
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0623489743813564e-05,
+      "loss": 1.1387,
+      "step": 10307
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0621934978282561e-05,
+      "loss": 1.069,
+      "step": 10308
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0620380197659026e-05,
+      "loss": 1.1723,
+      "step": 10309
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0618825401980689e-05,
+      "loss": 1.0662,
+      "step": 10310
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0617270591285278e-05,
+      "loss": 1.1338,
+      "step": 10311
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0615715765610523e-05,
+      "loss": 1.1508,
+      "step": 10312
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0614160924994161e-05,
+      "loss": 1.1476,
+      "step": 10313
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0612606069473917e-05,
+      "loss": 1.0716,
+      "step": 10314
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0611051199087527e-05,
+      "loss": 1.0851,
+      "step": 10315
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.060949631387272e-05,
+      "loss": 1.0917,
+      "step": 10316
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0607941413867231e-05,
+      "loss": 1.1292,
+      "step": 10317
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.060638649910879e-05,
+      "loss": 1.1836,
+      "step": 10318
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0604831569635134e-05,
+      "loss": 1.2324,
+      "step": 10319
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0603276625483994e-05,
+      "loss": 1.0005,
+      "step": 10320
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0601721666693105e-05,
+      "loss": 1.2338,
+      "step": 10321
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.06001666933002e-05,
+      "loss": 1.2056,
+      "step": 10322
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0598611705343011e-05,
+      "loss": 1.0989,
+      "step": 10323
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0597056702859283e-05,
+      "loss": 1.1938,
+      "step": 10324
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.059550168588674e-05,
+      "loss": 1.0096,
+      "step": 10325
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0593946654463125e-05,
+      "loss": 1.1008,
+      "step": 10326
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0592391608626169e-05,
+      "loss": 1.0445,
+      "step": 10327
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0590836548413615e-05,
+      "loss": 1.1852,
+      "step": 10328
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.058928147386319e-05,
+      "loss": 1.1301,
+      "step": 10329
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.058772638501264e-05,
+      "loss": 1.0926,
+      "step": 10330
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0586171281899698e-05,
+      "loss": 1.0431,
+      "step": 10331
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0584616164562105e-05,
+      "loss": 1.1926,
+      "step": 10332
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0583061033037594e-05,
+      "loss": 1.2138,
+      "step": 10333
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0581505887363908e-05,
+      "loss": 1.1365,
+      "step": 10334
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0579950727578783e-05,
+      "loss": 1.0965,
+      "step": 10335
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0578395553719962e-05,
+      "loss": 1.1509,
+      "step": 10336
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0576840365825178e-05,
+      "loss": 1.2462,
+      "step": 10337
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0575285163932179e-05,
+      "loss": 1.2171,
+      "step": 10338
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0573729948078699e-05,
+      "loss": 1.1269,
+      "step": 10339
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0572174718302482e-05,
+      "loss": 1.0478,
+      "step": 10340
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0570619474641263e-05,
+      "loss": 1.0763,
+      "step": 10341
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0569064217132791e-05,
+      "loss": 1.1555,
+      "step": 10342
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0567508945814804e-05,
+      "loss": 1.0371,
+      "step": 10343
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0565953660725043e-05,
+      "loss": 1.173,
+      "step": 10344
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0564398361901252e-05,
+      "loss": 1.2679,
+      "step": 10345
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0562843049381171e-05,
+      "loss": 1.0861,
+      "step": 10346
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0561287723202544e-05,
+      "loss": 1.2222,
+      "step": 10347
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0559732383403118e-05,
+      "loss": 1.1117,
+      "step": 10348
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0558177030020629e-05,
+      "loss": 1.182,
+      "step": 10349
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0556621663092828e-05,
+      "loss": 1.3095,
+      "step": 10350
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0555066282657456e-05,
+      "loss": 1.1859,
+      "step": 10351
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0553510888752256e-05,
+      "loss": 1.1648,
+      "step": 10352
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0551955481414976e-05,
+      "loss": 0.9495,
+      "step": 10353
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0550400060683356e-05,
+      "loss": 1.0664,
+      "step": 10354
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0548844626595149e-05,
+      "loss": 1.2554,
+      "step": 10355
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0547289179188099e-05,
+      "loss": 1.0712,
+      "step": 10356
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0545733718499944e-05,
+      "loss": 1.1876,
+      "step": 10357
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0544178244568442e-05,
+      "loss": 1.0581,
+      "step": 10358
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.054262275743133e-05,
+      "loss": 1.1911,
+      "step": 10359
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0541067257126364e-05,
+      "loss": 1.1063,
+      "step": 10360
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0539511743691283e-05,
+      "loss": 1.2309,
+      "step": 10361
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0537956217163843e-05,
+      "loss": 1.0981,
+      "step": 10362
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0536400677581783e-05,
+      "loss": 1.1883,
+      "step": 10363
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.053484512498286e-05,
+      "loss": 1.1976,
+      "step": 10364
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0533289559404815e-05,
+      "loss": 1.1739,
+      "step": 10365
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0531733980885406e-05,
+      "loss": 1.191,
+      "step": 10366
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.053017838946237e-05,
+      "loss": 1.1022,
+      "step": 10367
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.052862278517347e-05,
+      "loss": 1.1786,
+      "step": 10368
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0527067168056447e-05,
+      "loss": 1.0237,
+      "step": 10369
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0525511538149055e-05,
+      "loss": 1.0706,
+      "step": 10370
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0523955895489042e-05,
+      "loss": 1.1938,
+      "step": 10371
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0522400240114162e-05,
+      "loss": 1.1135,
+      "step": 10372
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.052084457206216e-05,
+      "loss": 1.0244,
+      "step": 10373
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0519288891370799e-05,
+      "loss": 1.2388,
+      "step": 10374
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0517733198077817e-05,
+      "loss": 1.0682,
+      "step": 10375
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0516177492220978e-05,
+      "loss": 1.2333,
+      "step": 10376
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0514621773838025e-05,
+      "loss": 1.1878,
+      "step": 10377
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.051306604296672e-05,
+      "loss": 1.0878,
+      "step": 10378
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0511510299644805e-05,
+      "loss": 1.1858,
+      "step": 10379
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0509954543910044e-05,
+      "loss": 1.1378,
+      "step": 10380
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0508398775800184e-05,
+      "loss": 1.1044,
+      "step": 10381
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.050684299535298e-05,
+      "loss": 1.1782,
+      "step": 10382
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0505287202606188e-05,
+      "loss": 1.2637,
+      "step": 10383
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0503731397597566e-05,
+      "loss": 1.0721,
+      "step": 10384
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0502175580364857e-05,
+      "loss": 1.0351,
+      "step": 10385
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0500619750945829e-05,
+      "loss": 1.1988,
+      "step": 10386
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.049906390937823e-05,
+      "loss": 1.1701,
+      "step": 10387
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0497508055699814e-05,
+      "loss": 1.0808,
+      "step": 10388
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0495952189948346e-05,
+      "loss": 1.1127,
+      "step": 10389
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0494396312161575e-05,
+      "loss": 1.0522,
+      "step": 10390
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.049284042237726e-05,
+      "loss": 1.118,
+      "step": 10391
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0491284520633157e-05,
+      "loss": 1.2744,
+      "step": 10392
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0489728606967022e-05,
+      "loss": 0.9657,
+      "step": 10393
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0488172681416616e-05,
+      "loss": 1.1807,
+      "step": 10394
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0486616744019691e-05,
+      "loss": 1.1561,
+      "step": 10395
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0485060794814013e-05,
+      "loss": 1.1873,
+      "step": 10396
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0483504833837332e-05,
+      "loss": 1.0692,
+      "step": 10397
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0481948861127416e-05,
+      "loss": 0.992,
+      "step": 10398
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0480392876722015e-05,
+      "loss": 1.1377,
+      "step": 10399
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0478836880658894e-05,
+      "loss": 1.1912,
+      "step": 10400
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0477280872975805e-05,
+      "loss": 1.2125,
+      "step": 10401
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0475724853710519e-05,
+      "loss": 1.0932,
+      "step": 10402
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0474168822900785e-05,
+      "loss": 1.072,
+      "step": 10403
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0472612780584373e-05,
+      "loss": 1.2015,
+      "step": 10404
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0471056726799034e-05,
+      "loss": 1.1982,
+      "step": 10405
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0469500661582536e-05,
+      "loss": 1.1784,
+      "step": 10406
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0467944584972638e-05,
+      "loss": 0.996,
+      "step": 10407
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0466388497007098e-05,
+      "loss": 1.2865,
+      "step": 10408
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0464832397723686e-05,
+      "loss": 1.0283,
+      "step": 10409
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0463276287160154e-05,
+      "loss": 1.1626,
+      "step": 10410
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0461720165354272e-05,
+      "loss": 1.0413,
+      "step": 10411
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0460164032343797e-05,
+      "loss": 1.184,
+      "step": 10412
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0458607888166496e-05,
+      "loss": 1.1898,
+      "step": 10413
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.045705173286013e-05,
+      "loss": 1.1883,
+      "step": 10414
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0455495566462463e-05,
+      "loss": 1.1172,
+      "step": 10415
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0453939389011257e-05,
+      "loss": 1.0029,
+      "step": 10416
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0452383200544278e-05,
+      "loss": 1.1372,
+      "step": 10417
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0450827001099288e-05,
+      "loss": 1.2086,
+      "step": 10418
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0449270790714055e-05,
+      "loss": 1.0525,
+      "step": 10419
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0447714569426337e-05,
+      "loss": 1.2094,
+      "step": 10420
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0446158337273908e-05,
+      "loss": 1.1252,
+      "step": 10421
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0444602094294525e-05,
+      "loss": 1.2839,
+      "step": 10422
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0443045840525958e-05,
+      "loss": 1.1671,
+      "step": 10423
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.044148957600597e-05,
+      "loss": 1.0608,
+      "step": 10424
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.043993330077233e-05,
+      "loss": 1.1863,
+      "step": 10425
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0438377014862801e-05,
+      "loss": 1.2365,
+      "step": 10426
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0436820718315152e-05,
+      "loss": 1.2028,
+      "step": 10427
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0435264411167148e-05,
+      "loss": 1.1606,
+      "step": 10428
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0433708093456559e-05,
+      "loss": 1.2815,
+      "step": 10429
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0432151765221148e-05,
+      "loss": 1.0952,
+      "step": 10430
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0430595426498686e-05,
+      "loss": 0.92,
+      "step": 10431
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0429039077326937e-05,
+      "loss": 1.3087,
+      "step": 10432
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0427482717743673e-05,
+      "loss": 1.1766,
+      "step": 10433
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0425926347786662e-05,
+      "loss": 1.114,
+      "step": 10434
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0424369967493668e-05,
+      "loss": 0.9856,
+      "step": 10435
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0422813576902463e-05,
+      "loss": 1.1365,
+      "step": 10436
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0421257176050818e-05,
+      "loss": 1.1483,
+      "step": 10437
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.04197007649765e-05,
+      "loss": 1.182,
+      "step": 10438
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0418144343717278e-05,
+      "loss": 1.0902,
+      "step": 10439
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0416587912310923e-05,
+      "loss": 1.1715,
+      "step": 10440
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0415031470795204e-05,
+      "loss": 1.1725,
+      "step": 10441
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0413475019207891e-05,
+      "loss": 1.0987,
+      "step": 10442
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0411918557586758e-05,
+      "loss": 1.3078,
+      "step": 10443
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.041036208596957e-05,
+      "loss": 1.112,
+      "step": 10444
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0408805604394103e-05,
+      "loss": 1.1367,
+      "step": 10445
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0407249112898126e-05,
+      "loss": 1.1911,
+      "step": 10446
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0405692611519411e-05,
+      "loss": 1.099,
+      "step": 10447
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0404136100295727e-05,
+      "loss": 1.1685,
+      "step": 10448
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0402579579264854e-05,
+      "loss": 1.1175,
+      "step": 10449
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.040102304846455e-05,
+      "loss": 1.1257,
+      "step": 10450
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0399466507932603e-05,
+      "loss": 1.1682,
+      "step": 10451
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0397909957706776e-05,
+      "loss": 1.2134,
+      "step": 10452
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0396353397824845e-05,
+      "loss": 1.0628,
+      "step": 10453
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0394796828324583e-05,
+      "loss": 1.128,
+      "step": 10454
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0393240249243763e-05,
+      "loss": 1.0131,
+      "step": 10455
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0391683660620158e-05,
+      "loss": 1.1571,
+      "step": 10456
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0390127062491544e-05,
+      "loss": 0.9763,
+      "step": 10457
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.038857045489569e-05,
+      "loss": 1.2455,
+      "step": 10458
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0387013837870376e-05,
+      "loss": 1.1602,
+      "step": 10459
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0385457211453374e-05,
+      "loss": 1.1738,
+      "step": 10460
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0383900575682462e-05,
+      "loss": 1.1996,
+      "step": 10461
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0382343930595407e-05,
+      "loss": 1.1364,
+      "step": 10462
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0380787276229994e-05,
+      "loss": 0.9019,
+      "step": 10463
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.037923061262399e-05,
+      "loss": 0.9882,
+      "step": 10464
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0377673939815176e-05,
+      "loss": 1.2185,
+      "step": 10465
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0376117257841326e-05,
+      "loss": 1.2215,
+      "step": 10466
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0374560566740217e-05,
+      "loss": 1.124,
+      "step": 10467
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0373003866549622e-05,
+      "loss": 1.2604,
+      "step": 10468
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0371447157307322e-05,
+      "loss": 1.1671,
+      "step": 10469
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0369890439051092e-05,
+      "loss": 1.2298,
+      "step": 10470
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.036833371181871e-05,
+      "loss": 1.084,
+      "step": 10471
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0366776975647948e-05,
+      "loss": 1.131,
+      "step": 10472
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0365220230576592e-05,
+      "loss": 1.1892,
+      "step": 10473
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0363663476642411e-05,
+      "loss": 1.211,
+      "step": 10474
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0362106713883188e-05,
+      "loss": 1.1751,
+      "step": 10475
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0360549942336701e-05,
+      "loss": 1.1998,
+      "step": 10476
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0358993162040728e-05,
+      "loss": 1.2925,
+      "step": 10477
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0357436373033048e-05,
+      "loss": 1.239,
+      "step": 10478
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0355879575351435e-05,
+      "loss": 1.4222,
+      "step": 10479
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0354322769033671e-05,
+      "loss": 1.0462,
+      "step": 10480
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0352765954117539e-05,
+      "loss": 1.202,
+      "step": 10481
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0351209130640812e-05,
+      "loss": 1.0879,
+      "step": 10482
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0349652298641273e-05,
+      "loss": 1.0787,
+      "step": 10483
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0348095458156699e-05,
+      "loss": 1.1807,
+      "step": 10484
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0346538609224877e-05,
+      "loss": 1.0361,
+      "step": 10485
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0344981751883578e-05,
+      "loss": 0.9728,
+      "step": 10486
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0343424886170588e-05,
+      "loss": 1.1093,
+      "step": 10487
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0341868012123683e-05,
+      "loss": 1.1956,
+      "step": 10488
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0340311129780652e-05,
+      "loss": 1.2061,
+      "step": 10489
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0338754239179267e-05,
+      "loss": 1.0121,
+      "step": 10490
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0337197340357315e-05,
+      "loss": 1.1355,
+      "step": 10491
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0335640433352572e-05,
+      "loss": 1.1949,
+      "step": 10492
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0334083518202824e-05,
+      "loss": 1.1177,
+      "step": 10493
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0332526594945853e-05,
+      "loss": 1.0269,
+      "step": 10494
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0330969663619438e-05,
+      "loss": 1.2132,
+      "step": 10495
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0329412724261365e-05,
+      "loss": 1.197,
+      "step": 10496
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0327855776909412e-05,
+      "loss": 1.1507,
+      "step": 10497
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0326298821601363e-05,
+      "loss": 1.3596,
+      "step": 10498
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0324741858375003e-05,
+      "loss": 1.2195,
+      "step": 10499
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0323184887268111e-05,
+      "loss": 1.1088,
+      "step": 10500
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0321627908318475e-05,
+      "loss": 1.1732,
+      "step": 10501
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0320070921563874e-05,
+      "loss": 1.1945,
+      "step": 10502
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0318513927042094e-05,
+      "loss": 1.2124,
+      "step": 10503
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0316956924790915e-05,
+      "loss": 1.0136,
+      "step": 10504
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0315399914848125e-05,
+      "loss": 1.229,
+      "step": 10505
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0313842897251509e-05,
+      "loss": 1.115,
+      "step": 10506
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0312285872038846e-05,
+      "loss": 1.2234,
+      "step": 10507
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0310728839247923e-05,
+      "loss": 1.1769,
+      "step": 10508
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0309171798916526e-05,
+      "loss": 0.9403,
+      "step": 10509
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.030761475108244e-05,
+      "loss": 1.1214,
+      "step": 10510
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0306057695783446e-05,
+      "loss": 1.1325,
+      "step": 10511
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0304500633057335e-05,
+      "loss": 1.0594,
+      "step": 10512
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0302943562941885e-05,
+      "loss": 1.1477,
+      "step": 10513
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0301386485474888e-05,
+      "loss": 1.1309,
+      "step": 10514
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0299829400694128e-05,
+      "loss": 1.1736,
+      "step": 10515
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0298272308637389e-05,
+      "loss": 1.1273,
+      "step": 10516
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0296715209342458e-05,
+      "loss": 1.0004,
+      "step": 10517
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0295158102847124e-05,
+      "loss": 1.1081,
+      "step": 10518
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0293600989189168e-05,
+      "loss": 1.2124,
+      "step": 10519
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0292043868406379e-05,
+      "loss": 1.1302,
+      "step": 10520
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0290486740536544e-05,
+      "loss": 1.1342,
+      "step": 10521
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0288929605617454e-05,
+      "loss": 1.0675,
+      "step": 10522
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0287372463686888e-05,
+      "loss": 1.1235,
+      "step": 10523
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.028581531478264e-05,
+      "loss": 1.1914,
+      "step": 10524
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0284258158942494e-05,
+      "loss": 1.0716,
+      "step": 10525
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.028270099620424e-05,
+      "loss": 1.0946,
+      "step": 10526
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0281143826605662e-05,
+      "loss": 1.0913,
+      "step": 10527
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0279586650184553e-05,
+      "loss": 0.9438,
+      "step": 10528
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0278029466978695e-05,
+      "loss": 1.0164,
+      "step": 10529
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0276472277025882e-05,
+      "loss": 1.0786,
+      "step": 10530
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0274915080363896e-05,
+      "loss": 1.1143,
+      "step": 10531
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0273357877030534e-05,
+      "loss": 1.1574,
+      "step": 10532
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0271800667063578e-05,
+      "loss": 1.2391,
+      "step": 10533
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.027024345050082e-05,
+      "loss": 1.1798,
+      "step": 10534
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0268686227380047e-05,
+      "loss": 0.9549,
+      "step": 10535
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.026712899773905e-05,
+      "loss": 1.1061,
+      "step": 10536
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0265571761615618e-05,
+      "loss": 1.1287,
+      "step": 10537
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0264014519047543e-05,
+      "loss": 1.0861,
+      "step": 10538
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0262457270072607e-05,
+      "loss": 1.1113,
+      "step": 10539
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0260900014728609e-05,
+      "loss": 1.096,
+      "step": 10540
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0259342753053332e-05,
+      "loss": 1.1044,
+      "step": 10541
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0257785485084572e-05,
+      "loss": 1.0901,
+      "step": 10542
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0256228210860113e-05,
+      "loss": 1.0974,
+      "step": 10543
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0254670930417752e-05,
+      "loss": 1.1319,
+      "step": 10544
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0253113643795273e-05,
+      "loss": 1.2249,
+      "step": 10545
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0251556351030475e-05,
+      "loss": 1.1619,
+      "step": 10546
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0249999052161142e-05,
+      "loss": 1.1606,
+      "step": 10547
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0248441747225067e-05,
+      "loss": 1.2273,
+      "step": 10548
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.024688443626004e-05,
+      "loss": 1.1638,
+      "step": 10549
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0245327119303856e-05,
+      "loss": 1.1356,
+      "step": 10550
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0243769796394303e-05,
+      "loss": 1.0027,
+      "step": 10551
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0242212467569174e-05,
+      "loss": 1.1633,
+      "step": 10552
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.024065513286626e-05,
+      "loss": 1.1378,
+      "step": 10553
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0239097792323355e-05,
+      "loss": 1.2873,
+      "step": 10554
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0237540445978248e-05,
+      "loss": 1.1181,
+      "step": 10555
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0235983093868738e-05,
+      "loss": 1.0957,
+      "step": 10556
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0234425736032607e-05,
+      "loss": 1.1652,
+      "step": 10557
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0232868372507656e-05,
+      "loss": 1.0732,
+      "step": 10558
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.023131100333167e-05,
+      "loss": 1.1263,
+      "step": 10559
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.022975362854245e-05,
+      "loss": 1.2337,
+      "step": 10560
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0228196248177784e-05,
+      "loss": 1.1126,
+      "step": 10561
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0226638862275466e-05,
+      "loss": 1.146,
+      "step": 10562
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0225081470873289e-05,
+      "loss": 1.2145,
+      "step": 10563
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0223524074009048e-05,
+      "loss": 1.0861,
+      "step": 10564
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0221966671720534e-05,
+      "loss": 1.0995,
+      "step": 10565
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0220409264045543e-05,
+      "loss": 1.1993,
+      "step": 10566
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0218851851021865e-05,
+      "loss": 1.0258,
+      "step": 10567
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.02172944326873e-05,
+      "loss": 1.1392,
+      "step": 10568
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0215737009079633e-05,
+      "loss": 1.0976,
+      "step": 10569
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0214179580236667e-05,
+      "loss": 1.2558,
+      "step": 10570
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0212622146196192e-05,
+      "loss": 1.2087,
+      "step": 10571
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0211064706996003e-05,
+      "loss": 1.1239,
+      "step": 10572
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0209507262673892e-05,
+      "loss": 1.0855,
+      "step": 10573
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.020794981326766e-05,
+      "loss": 1.0927,
+      "step": 10574
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0206392358815094e-05,
+      "loss": 1.1955,
+      "step": 10575
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0204834899353992e-05,
+      "loss": 1.1929,
+      "step": 10576
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0203277434922152e-05,
+      "loss": 1.1608,
+      "step": 10577
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0201719965557366e-05,
+      "loss": 1.0071,
+      "step": 10578
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0200162491297428e-05,
+      "loss": 1.1529,
+      "step": 10579
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0198605012180133e-05,
+      "loss": 0.9866,
+      "step": 10580
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0197047528243284e-05,
+      "loss": 1.2286,
+      "step": 10581
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0195490039524667e-05,
+      "loss": 1.0588,
+      "step": 10582
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0193932546062083e-05,
+      "loss": 1.1013,
+      "step": 10583
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0192375047893327e-05,
+      "loss": 1.1002,
+      "step": 10584
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0190817545056195e-05,
+      "loss": 1.0666,
+      "step": 10585
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.018926003758848e-05,
+      "loss": 1.0622,
+      "step": 10586
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0187702525527983e-05,
+      "loss": 1.0722,
+      "step": 10587
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0186145008912496e-05,
+      "loss": 1.0761,
+      "step": 10588
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0184587487779818e-05,
+      "loss": 0.9757,
+      "step": 10589
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0183029962167744e-05,
+      "loss": 1.0961,
+      "step": 10590
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0181472432114072e-05,
+      "loss": 1.184,
+      "step": 10591
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0179914897656595e-05,
+      "loss": 1.0576,
+      "step": 10592
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0178357358833117e-05,
+      "loss": 1.1287,
+      "step": 10593
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0176799815681427e-05,
+      "loss": 1.1549,
+      "step": 10594
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0175242268239327e-05,
+      "loss": 1.2941,
+      "step": 10595
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.017368471654461e-05,
+      "loss": 0.9675,
+      "step": 10596
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.017212716063508e-05,
+      "loss": 1.1924,
+      "step": 10597
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0170569600548525e-05,
+      "loss": 1.1946,
+      "step": 10598
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0169012036322751e-05,
+      "loss": 0.9428,
+      "step": 10599
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.016745446799555e-05,
+      "loss": 1.2935,
+      "step": 10600
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0165896895604723e-05,
+      "loss": 1.2595,
+      "step": 10601
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0164339319188066e-05,
+      "loss": 1.1305,
+      "step": 10602
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0162781738783378e-05,
+      "loss": 1.0631,
+      "step": 10603
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.016122415442845e-05,
+      "loss": 1.0621,
+      "step": 10604
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0159666566161093e-05,
+      "loss": 1.2062,
+      "step": 10605
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0158108974019094e-05,
+      "loss": 1.2853,
+      "step": 10606
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0156551378040259e-05,
+      "loss": 1.0677,
+      "step": 10607
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0154993778262379e-05,
+      "loss": 1.0498,
+      "step": 10608
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0153436174723256e-05,
+      "loss": 1.1055,
+      "step": 10609
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0151878567460689e-05,
+      "loss": 1.1261,
+      "step": 10610
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0150320956512479e-05,
+      "loss": 1.2637,
+      "step": 10611
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0148763341916417e-05,
+      "loss": 1.1892,
+      "step": 10612
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0147205723710307e-05,
+      "loss": 0.9902,
+      "step": 10613
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0145648101931948e-05,
+      "loss": 1.245,
+      "step": 10614
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.014409047661914e-05,
+      "loss": 1.2147,
+      "step": 10615
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0142532847809674e-05,
+      "loss": 1.0394,
+      "step": 10616
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.014097521554136e-05,
+      "loss": 1.1435,
+      "step": 10617
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.013941757985199e-05,
+      "loss": 1.2214,
+      "step": 10618
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0137859940779367e-05,
+      "loss": 1.2251,
+      "step": 10619
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0136302298361286e-05,
+      "loss": 1.2756,
+      "step": 10620
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0134744652635554e-05,
+      "loss": 1.1999,
+      "step": 10621
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0133187003639959e-05,
+      "loss": 1.0569,
+      "step": 10622
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0131629351412314e-05,
+      "loss": 1.1388,
+      "step": 10623
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0130071695990404e-05,
+      "loss": 1.1772,
+      "step": 10624
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0128514037412042e-05,
+      "loss": 1.0308,
+      "step": 10625
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.012695637571502e-05,
+      "loss": 1.2161,
+      "step": 10626
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0125398710937142e-05,
+      "loss": 1.0953,
+      "step": 10627
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0123841043116205e-05,
+      "loss": 1.0814,
+      "step": 10628
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0122283372290011e-05,
+      "loss": 0.9167,
+      "step": 10629
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0120725698496356e-05,
+      "loss": 1.0082,
+      "step": 10630
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0119168021773047e-05,
+      "loss": 1.1543,
+      "step": 10631
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.011761034215788e-05,
+      "loss": 1.1859,
+      "step": 10632
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0116052659688655e-05,
+      "loss": 1.0764,
+      "step": 10633
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0114494974403173e-05,
+      "loss": 1.1237,
+      "step": 10634
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0112937286339234e-05,
+      "loss": 1.2423,
+      "step": 10635
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.011137959553464e-05,
+      "loss": 1.1105,
+      "step": 10636
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0109821902027193e-05,
+      "loss": 1.2695,
+      "step": 10637
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0108264205854688e-05,
+      "loss": 0.8784,
+      "step": 10638
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0106706507054933e-05,
+      "loss": 1.0667,
+      "step": 10639
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.010514880566572e-05,
+      "loss": 1.1009,
+      "step": 10640
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0103591101724859e-05,
+      "loss": 1.2211,
+      "step": 10641
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0102033395270144e-05,
+      "loss": 1.238,
+      "step": 10642
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.010047568633938e-05,
+      "loss": 0.9245,
+      "step": 10643
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0098917974970363e-05,
+      "loss": 1.0238,
+      "step": 10644
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0097360261200902e-05,
+      "loss": 1.0549,
+      "step": 10645
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0095802545068791e-05,
+      "loss": 1.132,
+      "step": 10646
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0094244826611836e-05,
+      "loss": 1.1131,
+      "step": 10647
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0092687105867833e-05,
+      "loss": 1.0452,
+      "step": 10648
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0091129382874591e-05,
+      "loss": 1.0687,
+      "step": 10649
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0089571657669904e-05,
+      "loss": 1.0337,
+      "step": 10650
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0088013930291578e-05,
+      "loss": 1.1053,
+      "step": 10651
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0086456200777412e-05,
+      "loss": 1.1608,
+      "step": 10652
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.008489846916521e-05,
+      "loss": 1.0917,
+      "step": 10653
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0083340735492769e-05,
+      "loss": 1.1461,
+      "step": 10654
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0081782999797895e-05,
+      "loss": 1.0913,
+      "step": 10655
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0080225262118386e-05,
+      "loss": 1.1879,
+      "step": 10656
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.007866752249205e-05,
+      "loss": 1.0236,
+      "step": 10657
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.007710978095668e-05,
+      "loss": 1.0798,
+      "step": 10658
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0075552037550084e-05,
+      "loss": 1.1723,
+      "step": 10659
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0073994292310063e-05,
+      "loss": 1.0855,
+      "step": 10660
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0072436545274418e-05,
+      "loss": 1.1269,
+      "step": 10661
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0070878796480949e-05,
+      "loss": 1.169,
+      "step": 10662
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0069321045967464e-05,
+      "loss": 1.1608,
+      "step": 10663
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0067763293771757e-05,
+      "loss": 1.1938,
+      "step": 10664
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0066205539931637e-05,
+      "loss": 1.1174,
+      "step": 10665
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0064647784484903e-05,
+      "loss": 1.169,
+      "step": 10666
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0063090027469356e-05,
+      "loss": 1.11,
+      "step": 10667
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.00615322689228e-05,
+      "loss": 1.1546,
+      "step": 10668
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0059974508883037e-05,
+      "loss": 1.1524,
+      "step": 10669
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0058416747387872e-05,
+      "loss": 1.132,
+      "step": 10670
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.00568589844751e-05,
+      "loss": 1.192,
+      "step": 10671
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.005530122018253e-05,
+      "loss": 1.103,
+      "step": 10672
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.005374345454796e-05,
+      "loss": 1.0166,
+      "step": 10673
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0052185687609197e-05,
+      "loss": 1.1537,
+      "step": 10674
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0050627919404037e-05,
+      "loss": 1.1111,
+      "step": 10675
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.004907014997029e-05,
+      "loss": 1.1347,
+      "step": 10676
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0047512379345755e-05,
+      "loss": 1.1897,
+      "step": 10677
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0045954607568231e-05,
+      "loss": 1.1008,
+      "step": 10678
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0044396834675525e-05,
+      "loss": 1.162,
+      "step": 10679
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.004283906070544e-05,
+      "loss": 1.0709,
+      "step": 10680
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0041281285695776e-05,
+      "loss": 0.9515,
+      "step": 10681
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0039723509684338e-05,
+      "loss": 1.1625,
+      "step": 10682
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0038165732708925e-05,
+      "loss": 1.105,
+      "step": 10683
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0036607954807343e-05,
+      "loss": 1.3052,
+      "step": 10684
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0035050176017392e-05,
+      "loss": 1.1246,
+      "step": 10685
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0033492396376879e-05,
+      "loss": 1.0803,
+      "step": 10686
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.00319346159236e-05,
+      "loss": 1.0479,
+      "step": 10687
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0030376834695366e-05,
+      "loss": 1.1041,
+      "step": 10688
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0028819052729973e-05,
+      "loss": 1.1576,
+      "step": 10689
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0027261270065227e-05,
+      "loss": 1.1326,
+      "step": 10690
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.002570348673893e-05,
+      "loss": 0.9837,
+      "step": 10691
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0024145702788887e-05,
+      "loss": 0.9707,
+      "step": 10692
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0022587918252894e-05,
+      "loss": 1.0832,
+      "step": 10693
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0021030133168765e-05,
+      "loss": 1.1835,
+      "step": 10694
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0019472347574293e-05,
+      "loss": 1.0325,
+      "step": 10695
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0017914561507285e-05,
+      "loss": 1.2021,
+      "step": 10696
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0016356775005542e-05,
+      "loss": 1.1674,
+      "step": 10697
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0014798988106871e-05,
+      "loss": 1.1734,
+      "step": 10698
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0013241200849069e-05,
+      "loss": 1.0099,
+      "step": 10699
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0011683413269946e-05,
+      "loss": 1.0896,
+      "step": 10700
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0010125625407296e-05,
+      "loss": 1.04,
+      "step": 10701
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.000856783729893e-05,
+      "loss": 1.1082,
+      "step": 10702
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0007010048982646e-05,
+      "loss": 0.9839,
+      "step": 10703
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0005452260496251e-05,
+      "loss": 1.2514,
+      "step": 10704
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0003894471877545e-05,
+      "loss": 1.0517,
+      "step": 10705
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0002336683164331e-05,
+      "loss": 0.946,
+      "step": 10706
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0000778894394412e-05,
+      "loss": 1.0437,
+      "step": 10707
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.999221105605593e-06,
+      "loss": 1.1857,
+      "step": 10708
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.997663316835674e-06,
+      "loss": 1.1275,
+      "step": 10709
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.99610552812246e-06,
+      "loss": 1.0417,
+      "step": 10710
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.994547739503752e-06,
+      "loss": 1.0689,
+      "step": 10711
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.992989951017357e-06,
+      "loss": 1.1002,
+      "step": 10712
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.991432162701073e-06,
+      "loss": 1.0217,
+      "step": 10713
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.989874374592707e-06,
+      "loss": 1.1353,
+      "step": 10714
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.988316586730057e-06,
+      "loss": 1.1128,
+      "step": 10715
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.986758799150934e-06,
+      "loss": 1.1558,
+      "step": 10716
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.985201011893132e-06,
+      "loss": 1.1166,
+      "step": 10717
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.98364322499446e-06,
+      "loss": 1.101,
+      "step": 10718
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.982085438492719e-06,
+      "loss": 1.2378,
+      "step": 10719
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.98052765242571e-06,
+      "loss": 1.081,
+      "step": 10720
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.978969866831237e-06,
+      "loss": 1.1225,
+      "step": 10721
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.977412081747107e-06,
+      "loss": 1.0126,
+      "step": 10722
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.975854297211116e-06,
+      "loss": 1.2633,
+      "step": 10723
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.974296513261072e-06,
+      "loss": 1.1196,
+      "step": 10724
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.972738729934775e-06,
+      "loss": 1.1998,
+      "step": 10725
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.97118094727003e-06,
+      "loss": 1.1518,
+      "step": 10726
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.969623165304639e-06,
+      "loss": 1.091,
+      "step": 10727
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.968065384076403e-06,
+      "loss": 1.0365,
+      "step": 10728
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.966507603623125e-06,
+      "loss": 1.2423,
+      "step": 10729
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.964949823982613e-06,
+      "loss": 1.2194,
+      "step": 10730
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.963392045192662e-06,
+      "loss": 1.1955,
+      "step": 10731
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.961834267291079e-06,
+      "loss": 1.0738,
+      "step": 10732
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.960276490315668e-06,
+      "loss": 1.1131,
+      "step": 10733
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.958718714304227e-06,
+      "loss": 1.2598,
+      "step": 10734
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.957160939294565e-06,
+      "loss": 1.0938,
+      "step": 10735
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.955603165324478e-06,
+      "loss": 1.1594,
+      "step": 10736
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.954045392431772e-06,
+      "loss": 1.1219,
+      "step": 10737
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.95248762065425e-06,
+      "loss": 1.2487,
+      "step": 10738
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.950929850029715e-06,
+      "loss": 1.1631,
+      "step": 10739
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.949372080595964e-06,
+      "loss": 1.0431,
+      "step": 10740
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.94781431239081e-06,
+      "loss": 1.1644,
+      "step": 10741
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.946256545452043e-06,
+      "loss": 1.1575,
+      "step": 10742
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.944698779817475e-06,
+      "loss": 1.1291,
+      "step": 10743
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.943141015524905e-06,
+      "loss": 1.2016,
+      "step": 10744
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.941583252612135e-06,
+      "loss": 1.1251,
+      "step": 10745
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.940025491116965e-06,
+      "loss": 1.1182,
+      "step": 10746
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.938467731077204e-06,
+      "loss": 1.1204,
+      "step": 10747
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.936909972530647e-06,
+      "loss": 1.0459,
+      "step": 10748
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.9353522155151e-06,
+      "loss": 0.9953,
+      "step": 10749
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.933794460068364e-06,
+      "loss": 1.2452,
+      "step": 10750
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.932236706228243e-06,
+      "loss": 1.1815,
+      "step": 10751
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.930678954032539e-06,
+      "loss": 1.2218,
+      "step": 10752
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.92912120351905e-06,
+      "loss": 1.1923,
+      "step": 10753
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.927563454725584e-06,
+      "loss": 1.0596,
+      "step": 10754
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.926005707689937e-06,
+      "loss": 1.2135,
+      "step": 10755
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.924447962449917e-06,
+      "loss": 1.3205,
+      "step": 10756
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.92289021904332e-06,
+      "loss": 1.1991,
+      "step": 10757
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.921332477507955e-06,
+      "loss": 1.1801,
+      "step": 10758
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.919774737881615e-06,
+      "loss": 1.2422,
+      "step": 10759
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.918217000202108e-06,
+      "loss": 1.0518,
+      "step": 10760
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.916659264507233e-06,
+      "loss": 1.1108,
+      "step": 10761
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.915101530834795e-06,
+      "loss": 1.1251,
+      "step": 10762
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.91354379922259e-06,
+      "loss": 0.9618,
+      "step": 10763
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.911986069708424e-06,
+      "loss": 1.1561,
+      "step": 10764
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.910428342330096e-06,
+      "loss": 1.0724,
+      "step": 10765
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.90887061712541e-06,
+      "loss": 1.1792,
+      "step": 10766
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.907312894132165e-06,
+      "loss": 1.3669,
+      "step": 10767
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.905755173388167e-06,
+      "loss": 1.2214,
+      "step": 10768
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.90419745493121e-06,
+      "loss": 1.1052,
+      "step": 10769
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.902639738799102e-06,
+      "loss": 1.2472,
+      "step": 10770
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.901082025029639e-06,
+      "loss": 1.1189,
+      "step": 10771
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.899524313660623e-06,
+      "loss": 1.1688,
+      "step": 10772
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.89796660472986e-06,
+      "loss": 1.1188,
+      "step": 10773
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.896408898275143e-06,
+      "loss": 1.2225,
+      "step": 10774
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.894851194334283e-06,
+      "loss": 1.0462,
+      "step": 10775
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.89329349294507e-06,
+      "loss": 1.1271,
+      "step": 10776
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.891735794145314e-06,
+      "loss": 1.1925,
+      "step": 10777
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.89017809797281e-06,
+      "loss": 1.0922,
+      "step": 10778
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.888620404465363e-06,
+      "loss": 1.1172,
+      "step": 10779
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.887062713660767e-06,
+      "loss": 1.1861,
+      "step": 10780
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.88550502559683e-06,
+      "loss": 1.2547,
+      "step": 10781
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.883947340311347e-06,
+      "loss": 1.156,
+      "step": 10782
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.882389657842125e-06,
+      "loss": 1.0206,
+      "step": 10783
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.880831978226955e-06,
+      "loss": 1.1359,
+      "step": 10784
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.879274301503645e-06,
+      "loss": 1.227,
+      "step": 10785
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.877716627709992e-06,
+      "loss": 0.9895,
+      "step": 10786
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.876158956883799e-06,
+      "loss": 1.1724,
+      "step": 10787
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.87460128906286e-06,
+      "loss": 0.9746,
+      "step": 10788
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.873043624284983e-06,
+      "loss": 1.2588,
+      "step": 10789
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.87148596258796e-06,
+      "loss": 1.081,
+      "step": 10790
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.869928304009598e-06,
+      "loss": 1.1029,
+      "step": 10791
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.868370648587691e-06,
+      "loss": 1.0865,
+      "step": 10792
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.866812996360043e-06,
+      "loss": 1.1589,
+      "step": 10793
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.865255347364451e-06,
+      "loss": 1.0048,
+      "step": 10794
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.863697701638715e-06,
+      "loss": 0.9994,
+      "step": 10795
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.862140059220637e-06,
+      "loss": 1.1511,
+      "step": 10796
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.860582420148012e-06,
+      "loss": 1.2445,
+      "step": 10797
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.859024784458643e-06,
+      "loss": 1.1865,
+      "step": 10798
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.857467152190328e-06,
+      "loss": 1.1059,
+      "step": 10799
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.855909523380865e-06,
+      "loss": 1.0108,
+      "step": 10800
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.854351898068056e-06,
+      "loss": 1.1739,
+      "step": 10801
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.852794276289694e-06,
+      "loss": 1.1936,
+      "step": 10802
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.851236658083586e-06,
+      "loss": 1.2086,
+      "step": 10803
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.849679043487526e-06,
+      "loss": 1.2019,
+      "step": 10804
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.848121432539313e-06,
+      "loss": 1.0979,
+      "step": 10805
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.846563825276745e-06,
+      "loss": 1.1968,
+      "step": 10806
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.845006221737624e-06,
+      "loss": 1.1797,
+      "step": 10807
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.843448621959745e-06,
+      "loss": 1.0916,
+      "step": 10808
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.841891025980908e-06,
+      "loss": 1.1066,
+      "step": 10809
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.84033343383891e-06,
+      "loss": 1.0032,
+      "step": 10810
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.838775845571553e-06,
+      "loss": 1.1769,
+      "step": 10811
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.837218261216628e-06,
+      "loss": 1.1289,
+      "step": 10812
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.835660680811939e-06,
+      "loss": 0.9961,
+      "step": 10813
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.83410310439528e-06,
+      "loss": 1.0023,
+      "step": 10814
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.832545532004454e-06,
+      "loss": 1.1412,
+      "step": 10815
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.830987963677254e-06,
+      "loss": 1.1055,
+      "step": 10816
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.829430399451478e-06,
+      "loss": 1.2183,
+      "step": 10817
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.827872839364925e-06,
+      "loss": 1.047,
+      "step": 10818
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.826315283455392e-06,
+      "loss": 0.9762,
+      "step": 10819
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.824757731760678e-06,
+      "loss": 1.0266,
+      "step": 10820
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.823200184318577e-06,
+      "loss": 1.2418,
+      "step": 10821
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.82164264116689e-06,
+      "loss": 1.1896,
+      "step": 10822
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.820085102343408e-06,
+      "loss": 1.1643,
+      "step": 10823
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.818527567885933e-06,
+      "loss": 1.1643,
+      "step": 10824
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.81697003783226e-06,
+      "loss": 1.1679,
+      "step": 10825
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.815412512220187e-06,
+      "loss": 1.0675,
+      "step": 10826
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.813854991087507e-06,
+      "loss": 1.098,
+      "step": 10827
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.812297474472022e-06,
+      "loss": 1.0568,
+      "step": 10828
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.810739962411522e-06,
+      "loss": 1.1568,
+      "step": 10829
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.80918245494381e-06,
+      "loss": 1.1349,
+      "step": 10830
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.807624952106676e-06,
+      "loss": 1.0555,
+      "step": 10831
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.80606745393792e-06,
+      "loss": 1.1399,
+      "step": 10832
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.804509960475332e-06,
+      "loss": 1.1854,
+      "step": 10833
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.802952471756716e-06,
+      "loss": 1.1032,
+      "step": 10834
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.801394987819865e-06,
+      "loss": 1.2423,
+      "step": 10835
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.799837508702574e-06,
+      "loss": 0.9203,
+      "step": 10836
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.798280034442637e-06,
+      "loss": 1.1198,
+      "step": 10837
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.79672256507785e-06,
+      "loss": 1.0604,
+      "step": 10838
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.79516510064601e-06,
+      "loss": 1.143,
+      "step": 10839
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.793607641184908e-06,
+      "loss": 1.2803,
+      "step": 10840
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.792050186732344e-06,
+      "loss": 0.9395,
+      "step": 10841
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.790492737326108e-06,
+      "loss": 1.104,
+      "step": 10842
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.788935293004e-06,
+      "loss": 1.0816,
+      "step": 10843
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.78737785380381e-06,
+      "loss": 1.1135,
+      "step": 10844
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.785820419763335e-06,
+      "loss": 0.9529,
+      "step": 10845
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.784262990920367e-06,
+      "loss": 1.026,
+      "step": 10846
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.782705567312704e-06,
+      "loss": 1.2042,
+      "step": 10847
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.781148148978135e-06,
+      "loss": 1.0854,
+      "step": 10848
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.77959073595446e-06,
+      "loss": 1.2421,
+      "step": 10849
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.778033328279468e-06,
+      "loss": 1.202,
+      "step": 10850
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.776475925990955e-06,
+      "loss": 1.1694,
+      "step": 10851
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.774918529126713e-06,
+      "loss": 1.0894,
+      "step": 10852
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.773361137724537e-06,
+      "loss": 1.221,
+      "step": 10853
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.771803751822219e-06,
+      "loss": 1.1365,
+      "step": 10854
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.770246371457553e-06,
+      "loss": 1.0892,
+      "step": 10855
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.768688996668332e-06,
+      "loss": 1.1539,
+      "step": 10856
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.767131627492348e-06,
+      "loss": 1.2084,
+      "step": 10857
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.765574263967397e-06,
+      "loss": 1.1873,
+      "step": 10858
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.764016906131267e-06,
+      "loss": 1.1068,
+      "step": 10859
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.762459554021753e-06,
+      "loss": 1.1407,
+      "step": 10860
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.760902207676646e-06,
+      "loss": 1.2112,
+      "step": 10861
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.759344867133742e-06,
+      "loss": 1.1561,
+      "step": 10862
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.757787532430827e-06,
+      "loss": 1.1247,
+      "step": 10863
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.7562302036057e-06,
+      "loss": 1.1364,
+      "step": 10864
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.754672880696147e-06,
+      "loss": 1.0643,
+      "step": 10865
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.753115563739963e-06,
+      "loss": 0.9784,
+      "step": 10866
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.751558252774937e-06,
+      "loss": 1.0854,
+      "step": 10867
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.750000947838862e-06,
+      "loss": 0.9426,
+      "step": 10868
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.748443648969528e-06,
+      "loss": 1.1967,
+      "step": 10869
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.746886356204729e-06,
+      "loss": 1.1432,
+      "step": 10870
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.74532906958225e-06,
+      "loss": 1.0128,
+      "step": 10871
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.743771789139889e-06,
+      "loss": 1.1412,
+      "step": 10872
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.742214514915431e-06,
+      "loss": 1.0758,
+      "step": 10873
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.74065724694667e-06,
+      "loss": 1.2791,
+      "step": 10874
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.739099985271394e-06,
+      "loss": 0.969,
+      "step": 10875
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.737542729927394e-06,
+      "loss": 0.9788,
+      "step": 10876
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.73598548095246e-06,
+      "loss": 1.0117,
+      "step": 10877
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.734428238384385e-06,
+      "loss": 1.2408,
+      "step": 10878
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.732871002260951e-06,
+      "loss": 1.1134,
+      "step": 10879
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.731313772619956e-06,
+      "loss": 1.2044,
+      "step": 10880
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.729756549499181e-06,
+      "loss": 1.1402,
+      "step": 10881
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.728199332936425e-06,
+      "loss": 1.1416,
+      "step": 10882
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.72664212296947e-06,
+      "loss": 1.1507,
+      "step": 10883
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.725084919636106e-06,
+      "loss": 1.1595,
+      "step": 10884
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.723527722974121e-06,
+      "loss": 1.0222,
+      "step": 10885
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.721970533021307e-06,
+      "loss": 1.1536,
+      "step": 10886
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.72041334981545e-06,
+      "loss": 1.2326,
+      "step": 10887
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.718856173394342e-06,
+      "loss": 1.1881,
+      "step": 10888
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.717299003795763e-06,
+      "loss": 1.086,
+      "step": 10889
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.715741841057509e-06,
+      "loss": 1.102,
+      "step": 10890
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.714184685217361e-06,
+      "loss": 1.1593,
+      "step": 10891
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.712627536313113e-06,
+      "loss": 1.2099,
+      "step": 10892
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.711070394382549e-06,
+      "loss": 1.116,
+      "step": 10893
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.709513259463457e-06,
+      "loss": 1.057,
+      "step": 10894
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.707956131593623e-06,
+      "loss": 1.1514,
+      "step": 10895
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.706399010810837e-06,
+      "loss": 1.0276,
+      "step": 10896
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.70484189715288e-06,
+      "loss": 1.1677,
+      "step": 10897
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.703284790657547e-06,
+      "loss": 1.1584,
+      "step": 10898
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.701727691362614e-06,
+      "loss": 1.0563,
+      "step": 10899
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.700170599305877e-06,
+      "loss": 1.1916,
+      "step": 10900
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.698613514525117e-06,
+      "loss": 1.2532,
+      "step": 10901
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.69705643705812e-06,
+      "loss": 1.0599,
+      "step": 10902
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.695499366942672e-06,
+      "loss": 0.8853,
+      "step": 10903
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.693942304216557e-06,
+      "loss": 1.364,
+      "step": 10904
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.692385248917565e-06,
+      "loss": 1.2175,
+      "step": 10905
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.690828201083476e-06,
+      "loss": 1.0805,
+      "step": 10906
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.68927116075208e-06,
+      "loss": 0.9769,
+      "step": 10907
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.687714127961158e-06,
+      "loss": 1.0742,
+      "step": 10908
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.686157102748496e-06,
+      "loss": 1.1917,
+      "step": 10909
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.684600085151877e-06,
+      "loss": 1.047,
+      "step": 10910
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.683043075209088e-06,
+      "loss": 1.1765,
+      "step": 10911
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.681486072957911e-06,
+      "loss": 1.0662,
+      "step": 10912
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.679929078436131e-06,
+      "loss": 1.1541,
+      "step": 10913
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.678372091681529e-06,
+      "loss": 1.2182,
+      "step": 10914
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.676815112731892e-06,
+      "loss": 1.0419,
+      "step": 10915
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.675258141624999e-06,
+      "loss": 1.2879,
+      "step": 10916
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.673701178398635e-06,
+      "loss": 1.1533,
+      "step": 10917
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.67214422309059e-06,
+      "loss": 1.0392,
+      "step": 10918
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.670587275738635e-06,
+      "loss": 1.1336,
+      "step": 10919
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.669030336380562e-06,
+      "loss": 1.0435,
+      "step": 10920
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.667473405054147e-06,
+      "loss": 1.1595,
+      "step": 10921
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.665916481797176e-06,
+      "loss": 1.1763,
+      "step": 10922
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.664359566647428e-06,
+      "loss": 1.0657,
+      "step": 10923
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.662802659642688e-06,
+      "loss": 1.2701,
+      "step": 10924
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.661245760820734e-06,
+      "loss": 1.0043,
+      "step": 10925
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.659688870219352e-06,
+      "loss": 1.1506,
+      "step": 10926
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.658131987876317e-06,
+      "loss": 1.0045,
+      "step": 10927
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.656575113829415e-06,
+      "loss": 1.0707,
+      "step": 10928
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.655018248116424e-06,
+      "loss": 1.2603,
+      "step": 10929
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.653461390775127e-06,
+      "loss": 1.0434,
+      "step": 10930
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.651904541843301e-06,
+      "loss": 1.0797,
+      "step": 10931
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.65034770135873e-06,
+      "loss": 0.9849,
+      "step": 10932
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.64879086935919e-06,
+      "loss": 1.0323,
+      "step": 10933
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.647234045882465e-06,
+      "loss": 1.15,
+      "step": 10934
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.64567723096633e-06,
+      "loss": 1.1679,
+      "step": 10935
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.644120424648568e-06,
+      "loss": 1.279,
+      "step": 10936
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.642563626966956e-06,
+      "loss": 1.2396,
+      "step": 10937
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.641006837959274e-06,
+      "loss": 1.1319,
+      "step": 10938
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.6394500576633e-06,
+      "loss": 1.2877,
+      "step": 10939
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.637893286116814e-06,
+      "loss": 1.0262,
+      "step": 10940
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.636336523357592e-06,
+      "loss": 1.0617,
+      "step": 10941
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.634779769423412e-06,
+      "loss": 1.1805,
+      "step": 10942
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.633223024352054e-06,
+      "loss": 1.0725,
+      "step": 10943
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.631666288181293e-06,
+      "loss": 1.0973,
+      "step": 10944
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.630109560948912e-06,
+      "loss": 1.0569,
+      "step": 10945
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.62855284269268e-06,
+      "loss": 1.1485,
+      "step": 10946
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.626996133450382e-06,
+      "loss": 1.1531,
+      "step": 10947
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.625439433259786e-06,
+      "loss": 1.1868,
+      "step": 10948
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.623882742158678e-06,
+      "loss": 1.1165,
+      "step": 10949
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.622326060184827e-06,
+      "loss": 1.073,
+      "step": 10950
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.620769387376013e-06,
+      "loss": 1.1698,
+      "step": 10951
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.61921272377001e-06,
+      "loss": 1.1502,
+      "step": 10952
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.617656069404595e-06,
+      "loss": 1.2097,
+      "step": 10953
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.616099424317541e-06,
+      "loss": 1.117,
+      "step": 10954
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.614542788546627e-06,
+      "loss": 1.0252,
+      "step": 10955
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.612986162129625e-06,
+      "loss": 1.3809,
+      "step": 10956
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.611429545104313e-06,
+      "loss": 1.2246,
+      "step": 10957
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.609872937508459e-06,
+      "loss": 0.9095,
+      "step": 10958
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.608316339379847e-06,
+      "loss": 1.2245,
+      "step": 10959
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.606759750756239e-06,
+      "loss": 1.0274,
+      "step": 10960
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.60520317167542e-06,
+      "loss": 1.2121,
+      "step": 10961
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.603646602175158e-06,
+      "loss": 0.9614,
+      "step": 10962
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.602090042293228e-06,
+      "loss": 1.1789,
+      "step": 10963
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.600533492067398e-06,
+      "loss": 1.185,
+      "step": 10964
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.598976951535451e-06,
+      "loss": 1.2213,
+      "step": 10965
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.597420420735152e-06,
+      "loss": 1.1225,
+      "step": 10966
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.595863899704276e-06,
+      "loss": 1.0949,
+      "step": 10967
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.594307388480592e-06,
+      "loss": 1.1327,
+      "step": 10968
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.592750887101877e-06,
+      "loss": 1.2666,
+      "step": 10969
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.591194395605899e-06,
+      "loss": 1.0033,
+      "step": 10970
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.589637914030433e-06,
+      "loss": 1.2418,
+      "step": 10971
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.588081442413245e-06,
+      "loss": 1.1379,
+      "step": 10972
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.58652498079211e-06,
+      "loss": 1.1178,
+      "step": 10973
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.5849685292048e-06,
+      "loss": 1.0459,
+      "step": 10974
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.58341208768908e-06,
+      "loss": 1.0463,
+      "step": 10975
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.581855656282724e-06,
+      "loss": 1.2232,
+      "step": 10976
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.580299235023505e-06,
+      "loss": 1.0787,
+      "step": 10977
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.578742823949185e-06,
+      "loss": 1.073,
+      "step": 10978
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.57718642309754e-06,
+      "loss": 1.0595,
+      "step": 10979
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.575630032506335e-06,
+      "loss": 1.1108,
+      "step": 10980
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.574073652213345e-06,
+      "loss": 1.0281,
+      "step": 10981
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.57251728225633e-06,
+      "loss": 1.1187,
+      "step": 10982
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.570960922673068e-06,
+      "loss": 1.0616,
+      "step": 10983
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.56940457350132e-06,
+      "loss": 1.2428,
+      "step": 10984
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.567848234778857e-06,
+      "loss": 1.1024,
+      "step": 10985
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.566291906543446e-06,
+      "loss": 1.202,
+      "step": 10986
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.564735588832857e-06,
+      "loss": 1.1709,
+      "step": 10987
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.563179281684853e-06,
+      "loss": 1.088,
+      "step": 10988
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.561622985137202e-06,
+      "loss": 1.2438,
+      "step": 10989
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.560066699227675e-06,
+      "loss": 1.0029,
+      "step": 10990
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.558510423994033e-06,
+      "loss": 1.1772,
+      "step": 10991
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.556954159474048e-06,
+      "loss": 1.0761,
+      "step": 10992
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.55539790570548e-06,
+      "loss": 1.1467,
+      "step": 10993
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.553841662726098e-06,
+      "loss": 1.1897,
+      "step": 10994
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.552285430573665e-06,
+      "loss": 1.0828,
+      "step": 10995
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.55072920928595e-06,
+      "loss": 1.043,
+      "step": 10996
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.549172998900715e-06,
+      "loss": 1.2191,
+      "step": 10997
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.547616799455727e-06,
+      "loss": 1.2396,
+      "step": 10998
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.546060610988747e-06,
+      "loss": 1.2581,
+      "step": 10999
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.544504433537537e-06,
+      "loss": 0.9746,
+      "step": 11000
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.542948267139871e-06,
+      "loss": 1.0024,
+      "step": 11001
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.541392111833504e-06,
+      "loss": 1.0967,
+      "step": 11002
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.539835967656204e-06,
+      "loss": 1.0829,
+      "step": 11003
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.538279834645728e-06,
+      "loss": 1.1873,
+      "step": 11004
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.536723712839848e-06,
+      "loss": 1.0156,
+      "step": 11005
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.535167602276315e-06,
+      "loss": 1.1885,
+      "step": 11006
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.533611502992902e-06,
+      "loss": 1.0811,
+      "step": 11007
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.532055415027364e-06,
+      "loss": 1.0914,
+      "step": 11008
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.530499338417466e-06,
+      "loss": 1.2262,
+      "step": 11009
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.528943273200965e-06,
+      "loss": 1.0926,
+      "step": 11010
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.52738721941563e-06,
+      "loss": 1.0704,
+      "step": 11011
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.525831177099215e-06,
+      "loss": 1.0508,
+      "step": 11012
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.524275146289484e-06,
+      "loss": 1.2418,
+      "step": 11013
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.522719127024195e-06,
+      "loss": 1.0483,
+      "step": 11014
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.52116311934111e-06,
+      "loss": 1.1588,
+      "step": 11015
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.519607123277987e-06,
+      "loss": 1.0422,
+      "step": 11016
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.518051138872587e-06,
+      "loss": 1.071,
+      "step": 11017
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.516495166162667e-06,
+      "loss": 1.0534,
+      "step": 11018
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.51493920518599e-06,
+      "loss": 1.0184,
+      "step": 11019
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.51338325598031e-06,
+      "loss": 0.9959,
+      "step": 11020
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.51182731858339e-06,
+      "loss": 1.1148,
+      "step": 11021
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.510271393032981e-06,
+      "loss": 1.0968,
+      "step": 11022
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.508715479366848e-06,
+      "loss": 1.1199,
+      "step": 11023
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.507159577622743e-06,
+      "loss": 1.2434,
+      "step": 11024
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.50560368783843e-06,
+      "loss": 1.0143,
+      "step": 11025
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.504047810051656e-06,
+      "loss": 1.1644,
+      "step": 11026
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.502491944300188e-06,
+      "loss": 1.0591,
+      "step": 11027
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.500936090621774e-06,
+      "loss": 1.1209,
+      "step": 11028
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.499380249054174e-06,
+      "loss": 1.1068,
+      "step": 11029
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.497824419635145e-06,
+      "loss": 1.2275,
+      "step": 11030
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49626860240244e-06,
+      "loss": 1.1996,
+      "step": 11031
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.494712797393813e-06,
+      "loss": 1.2574,
+      "step": 11032
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49315700464702e-06,
+      "loss": 1.1523,
+      "step": 11033
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49160122419982e-06,
+      "loss": 1.1189,
+      "step": 11034
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.490045456089958e-06,
+      "loss": 1.2562,
+      "step": 11035
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.488489700355197e-06,
+      "loss": 1.0741,
+      "step": 11036
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.486933957033283e-06,
+      "loss": 1.0909,
+      "step": 11037
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.485378226161978e-06,
+      "loss": 1.0674,
+      "step": 11038
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.483822507779026e-06,
+      "loss": 1.0808,
+      "step": 11039
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.482266801922186e-06,
+      "loss": 1.3059,
+      "step": 11040
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.480711108629206e-06,
+      "loss": 1.3524,
+      "step": 11041
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.479155427937843e-06,
+      "loss": 1.1049,
+      "step": 11042
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.477599759885842e-06,
+      "loss": 1.062,
+      "step": 11043
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.476044104510961e-06,
+      "loss": 1.0579,
+      "step": 11044
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.474488461850948e-06,
+      "loss": 1.2795,
+      "step": 11045
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.472932831943556e-06,
+      "loss": 1.0619,
+      "step": 11046
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.471377214826533e-06,
+      "loss": 1.1465,
+      "step": 11047
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.469821610537631e-06,
+      "loss": 0.8923,
+      "step": 11048
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.468266019114599e-06,
+      "loss": 1.1778,
+      "step": 11049
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.466710440595186e-06,
+      "loss": 1.1268,
+      "step": 11050
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.465154875017142e-06,
+      "loss": 1.1864,
+      "step": 11051
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.463599322418218e-06,
+      "loss": 1.1279,
+      "step": 11052
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.46204378283616e-06,
+      "loss": 1.1478,
+      "step": 11053
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.460488256308718e-06,
+      "loss": 1.2115,
+      "step": 11054
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.458932742873639e-06,
+      "loss": 1.1054,
+      "step": 11055
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.457377242568671e-06,
+      "loss": 1.1042,
+      "step": 11056
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.455821755431561e-06,
+      "loss": 1.1049,
+      "step": 11057
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.454266281500059e-06,
+      "loss": 1.13,
+      "step": 11058
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.452710820811905e-06,
+      "loss": 1.2497,
+      "step": 11059
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.451155373404855e-06,
+      "loss": 0.8981,
+      "step": 11060
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.449599939316645e-06,
+      "loss": 1.1339,
+      "step": 11061
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44804451858503e-06,
+      "loss": 1.2465,
+      "step": 11062
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.446489111247749e-06,
+      "loss": 1.3421,
+      "step": 11063
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44493371734255e-06,
+      "loss": 1.1876,
+      "step": 11064
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.443378336907177e-06,
+      "loss": 1.1015,
+      "step": 11065
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.441822969979377e-06,
+      "loss": 1.1146,
+      "step": 11066
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.440267616596888e-06,
+      "loss": 1.1613,
+      "step": 11067
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.438712276797461e-06,
+      "loss": 1.1182,
+      "step": 11068
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.437156950618834e-06,
+      "loss": 1.2991,
+      "step": 11069
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.435601638098755e-06,
+      "loss": 1.0359,
+      "step": 11070
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.434046339274962e-06,
+      "loss": 1.1287,
+      "step": 11071
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.432491054185203e-06,
+      "loss": 1.0641,
+      "step": 11072
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.430935782867214e-06,
+      "loss": 0.9715,
+      "step": 11073
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.429380525358738e-06,
+      "loss": 1.1988,
+      "step": 11074
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.427825281697523e-06,
+      "loss": 1.0895,
+      "step": 11075
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.426270051921305e-06,
+      "loss": 1.0952,
+      "step": 11076
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.424714836067826e-06,
+      "loss": 1.0223,
+      "step": 11077
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.423159634174824e-06,
+      "loss": 1.1894,
+      "step": 11078
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.421604446280043e-06,
+      "loss": 1.2658,
+      "step": 11079
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.420049272421219e-06,
+      "loss": 1.2444,
+      "step": 11080
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.418494112636097e-06,
+      "loss": 1.0735,
+      "step": 11081
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.41693896696241e-06,
+      "loss": 1.2904,
+      "step": 11082
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.415383835437896e-06,
+      "loss": 1.123,
+      "step": 11083
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.413828718100304e-06,
+      "loss": 1.2017,
+      "step": 11084
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.412273614987359e-06,
+      "loss": 1.2291,
+      "step": 11085
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.410718526136811e-06,
+      "loss": 1.1248,
+      "step": 11086
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.409163451586387e-06,
+      "loss": 1.2567,
+      "step": 11087
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.407608391373831e-06,
+      "loss": 1.1862,
+      "step": 11088
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.406053345536875e-06,
+      "loss": 1.1763,
+      "step": 11089
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.40449831411326e-06,
+      "loss": 1.0131,
+      "step": 11090
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.402943297140718e-06,
+      "loss": 1.1866,
+      "step": 11091
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.401388294656987e-06,
+      "loss": 1.2631,
+      "step": 11092
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.399833306699801e-06,
+      "loss": 1.0228,
+      "step": 11093
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.398278333306898e-06,
+      "loss": 1.1007,
+      "step": 11094
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.396723374516006e-06,
+      "loss": 0.8614,
+      "step": 11095
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.395168430364869e-06,
+      "loss": 1.1967,
+      "step": 11096
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.39361350089121e-06,
+      "loss": 1.2818,
+      "step": 11097
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.392058586132772e-06,
+      "loss": 1.1868,
+      "step": 11098
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.390503686127281e-06,
+      "loss": 1.1986,
+      "step": 11099
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.388948800912476e-06,
+      "loss": 1.2067,
+      "step": 11100
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.387393930526084e-06,
+      "loss": 1.2736,
+      "step": 11101
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.385839075005842e-06,
+      "loss": 1.1508,
+      "step": 11102
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.384284234389479e-06,
+      "loss": 1.0555,
+      "step": 11103
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.382729408714725e-06,
+      "loss": 1.1405,
+      "step": 11104
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.381174598019313e-06,
+      "loss": 1.1697,
+      "step": 11105
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.379619802340976e-06,
+      "loss": 0.9756,
+      "step": 11106
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.37806502171744e-06,
+      "loss": 0.9975,
+      "step": 11107
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.37651025618644e-06,
+      "loss": 1.1196,
+      "step": 11108
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.374955505785698e-06,
+      "loss": 1.1323,
+      "step": 11109
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.37340077055295e-06,
+      "loss": 1.1605,
+      "step": 11110
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.371846050525922e-06,
+      "loss": 1.1644,
+      "step": 11111
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.370291345742343e-06,
+      "loss": 1.079,
+      "step": 11112
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.368736656239942e-06,
+      "loss": 1.0832,
+      "step": 11113
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.367181982056444e-06,
+      "loss": 1.1637,
+      "step": 11114
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.36562732322958e-06,
+      "loss": 1.2282,
+      "step": 11115
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.364072679797074e-06,
+      "loss": 1.1563,
+      "step": 11116
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.362518051796656e-06,
+      "loss": 1.0734,
+      "step": 11117
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.360963439266048e-06,
+      "loss": 1.1704,
+      "step": 11118
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.35940884224298e-06,
+      "loss": 1.2167,
+      "step": 11119
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.357854260765173e-06,
+      "loss": 1.0815,
+      "step": 11120
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.356299694870357e-06,
+      "loss": 1.0554,
+      "step": 11121
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.354745144596253e-06,
+      "loss": 1.1046,
+      "step": 11122
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.353190609980587e-06,
+      "loss": 1.1525,
+      "step": 11123
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.351636091061082e-06,
+      "loss": 1.0268,
+      "step": 11124
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.350081587875463e-06,
+      "loss": 1.0753,
+      "step": 11125
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.348527100461453e-06,
+      "loss": 1.1467,
+      "step": 11126
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.346972628856775e-06,
+      "loss": 1.1093,
+      "step": 11127
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.345418173099146e-06,
+      "loss": 1.0477,
+      "step": 11128
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.343863733226298e-06,
+      "loss": 1.1195,
+      "step": 11129
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.342309309275946e-06,
+      "loss": 1.1885,
+      "step": 11130
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.340754901285814e-06,
+      "loss": 1.132,
+      "step": 11131
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.339200509293618e-06,
+      "loss": 1.0401,
+      "step": 11132
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.337646133337087e-06,
+      "loss": 1.1495,
+      "step": 11133
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.336091773453933e-06,
+      "loss": 1.1418,
+      "step": 11134
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.334537429681882e-06,
+      "loss": 1.145,
+      "step": 11135
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.332983102058648e-06,
+      "loss": 0.9713,
+      "step": 11136
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.331428790621954e-06,
+      "loss": 1.1127,
+      "step": 11137
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.329874495409517e-06,
+      "loss": 1.2471,
+      "step": 11138
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.328320216459056e-06,
+      "loss": 1.1417,
+      "step": 11139
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.326765953808285e-06,
+      "loss": 1.0807,
+      "step": 11140
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.325211707494928e-06,
+      "loss": 1.051,
+      "step": 11141
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.323657477556695e-06,
+      "loss": 1.1705,
+      "step": 11142
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.32210326403131e-06,
+      "loss": 1.152,
+      "step": 11143
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.320549066956481e-06,
+      "loss": 1.2397,
+      "step": 11144
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.31899488636993e-06,
+      "loss": 1.2948,
+      "step": 11145
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.31744072230937e-06,
+      "loss": 1.0115,
+      "step": 11146
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.315886574812518e-06,
+      "loss": 1.0853,
+      "step": 11147
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.314332443917083e-06,
+      "loss": 1.0758,
+      "step": 11148
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.312778329660788e-06,
+      "loss": 0.8477,
+      "step": 11149
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.311224232081338e-06,
+      "loss": 1.2907,
+      "step": 11150
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.309670151216451e-06,
+      "loss": 1.0875,
+      "step": 11151
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.30811608710384e-06,
+      "loss": 1.0841,
+      "step": 11152
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.306562039781217e-06,
+      "loss": 1.2199,
+      "step": 11153
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.305008009286293e-06,
+      "loss": 1.0869,
+      "step": 11154
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.303453995656783e-06,
+      "loss": 1.267,
+      "step": 11155
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.301899998930392e-06,
+      "loss": 1.1546,
+      "step": 11156
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.30034601914484e-06,
+      "loss": 1.1741,
+      "step": 11157
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.298792056337828e-06,
+      "loss": 1.2222,
+      "step": 11158
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.297238110547075e-06,
+      "loss": 1.1165,
+      "step": 11159
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.295684181810282e-06,
+      "loss": 1.1185,
+      "step": 11160
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.294130270165163e-06,
+      "loss": 1.135,
+      "step": 11161
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.29257637564943e-06,
+      "loss": 1.0343,
+      "step": 11162
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.291022498300784e-06,
+      "loss": 1.2403,
+      "step": 11163
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.28946863815694e-06,
+      "loss": 1.0144,
+      "step": 11164
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.2879147952556e-06,
+      "loss": 1.1046,
+      "step": 11165
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.286360969634472e-06,
+      "loss": 1.1973,
+      "step": 11166
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.284807161331269e-06,
+      "loss": 1.1084,
+      "step": 11167
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.28325337038369e-06,
+      "loss": 1.0937,
+      "step": 11168
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.281699596829447e-06,
+      "loss": 1.1744,
+      "step": 11169
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.280145840706239e-06,
+      "loss": 1.1652,
+      "step": 11170
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.278592102051777e-06,
+      "loss": 1.1679,
+      "step": 11171
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.27703838090376e-06,
+      "loss": 1.3168,
+      "step": 11172
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.2754846772999e-06,
+      "loss": 1.1089,
+      "step": 11173
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.273930991277893e-06,
+      "loss": 1.27,
+      "step": 11174
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.272377322875446e-06,
+      "loss": 1.1346,
+      "step": 11175
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.270823672130262e-06,
+      "loss": 1.1625,
+      "step": 11176
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.269270039080043e-06,
+      "loss": 1.2052,
+      "step": 11177
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.267716423762489e-06,
+      "loss": 1.1007,
+      "step": 11178
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.266162826215307e-06,
+      "loss": 0.932,
+      "step": 11179
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.264609246476194e-06,
+      "loss": 1.1001,
+      "step": 11180
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.263055684582853e-06,
+      "loss": 1.1005,
+      "step": 11181
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.26150214057298e-06,
+      "loss": 1.0716,
+      "step": 11182
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.25994861448428e-06,
+      "loss": 1.0337,
+      "step": 11183
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.25839510635445e-06,
+      "loss": 1.231,
+      "step": 11184
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.256841616221192e-06,
+      "loss": 1.1301,
+      "step": 11185
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.2552881441222e-06,
+      "loss": 1.2437,
+      "step": 11186
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.253734690095178e-06,
+      "loss": 1.0396,
+      "step": 11187
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.252181254177817e-06,
+      "loss": 1.222,
+      "step": 11188
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.25062783640782e-06,
+      "loss": 1.2339,
+      "step": 11189
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.249074436822878e-06,
+      "loss": 1.0934,
+      "step": 11190
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.247521055460695e-06,
+      "loss": 1.1383,
+      "step": 11191
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.24596769235896e-06,
+      "loss": 1.161,
+      "step": 11192
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.244414347555375e-06,
+      "loss": 1.1941,
+      "step": 11193
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.242861021087628e-06,
+      "loss": 1.2499,
+      "step": 11194
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.24130771299342e-06,
+      "loss": 1.2,
+      "step": 11195
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.239754423310441e-06,
+      "loss": 1.2354,
+      "step": 11196
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.238201152076389e-06,
+      "loss": 1.1567,
+      "step": 11197
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.23664789932895e-06,
+      "loss": 1.1008,
+      "step": 11198
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.235094665105826e-06,
+      "loss": 1.2788,
+      "step": 11199
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.233541449444702e-06,
+      "loss": 1.0853,
+      "step": 11200
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.231988252383273e-06,
+      "loss": 1.2754,
+      "step": 11201
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.230435073959232e-06,
+      "loss": 1.1963,
+      "step": 11202
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.228881914210267e-06,
+      "loss": 1.1427,
+      "step": 11203
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.227328773174072e-06,
+      "loss": 1.1697,
+      "step": 11204
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.225775650888334e-06,
+      "loss": 1.2088,
+      "step": 11205
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.224222547390747e-06,
+      "loss": 1.1214,
+      "step": 11206
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.222669462718992e-06,
+      "loss": 0.9904,
+      "step": 11207
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.221116396910768e-06,
+      "loss": 1.1171,
+      "step": 11208
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.219563350003756e-06,
+      "loss": 1.0373,
+      "step": 11209
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.218010322035648e-06,
+      "loss": 1.0879,
+      "step": 11210
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.216457313044127e-06,
+      "loss": 1.2656,
+      "step": 11211
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.214904323066884e-06,
+      "loss": 1.1095,
+      "step": 11212
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.213351352141605e-06,
+      "loss": 1.1587,
+      "step": 11213
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.211798400305975e-06,
+      "loss": 1.1674,
+      "step": 11214
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.210245467597678e-06,
+      "loss": 1.2331,
+      "step": 11215
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.208692554054404e-06,
+      "loss": 1.0675,
+      "step": 11216
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.207139659713832e-06,
+      "loss": 1.0117,
+      "step": 11217
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.20558678461365e-06,
+      "loss": 1.1346,
+      "step": 11218
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.20403392879154e-06,
+      "loss": 1.1603,
+      "step": 11219
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.202481092285187e-06,
+      "loss": 1.2153,
+      "step": 11220
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.20092827513227e-06,
+      "loss": 1.1077,
+      "step": 11221
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.199375477370476e-06,
+      "loss": 0.9895,
+      "step": 11222
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.197822699037483e-06,
+      "loss": 1.0924,
+      "step": 11223
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.196269940170975e-06,
+      "loss": 1.1224,
+      "step": 11224
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.19471720080863e-06,
+      "loss": 1.1055,
+      "step": 11225
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.193164480988134e-06,
+      "loss": 1.1011,
+      "step": 11226
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.19161178074716e-06,
+      "loss": 1.036,
+      "step": 11227
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.190059100123392e-06,
+      "loss": 1.0501,
+      "step": 11228
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.188506439154506e-06,
+      "loss": 1.084,
+      "step": 11229
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.186953797878185e-06,
+      "loss": 1.1432,
+      "step": 11230
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.185401176332103e-06,
+      "loss": 0.9819,
+      "step": 11231
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.183848574553939e-06,
+      "loss": 0.9906,
+      "step": 11232
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.182295992581368e-06,
+      "loss": 1.1513,
+      "step": 11233
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.180743430452072e-06,
+      "loss": 1.2177,
+      "step": 11234
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.17919088820372e-06,
+      "loss": 1.1591,
+      "step": 11235
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.177638365873995e-06,
+      "loss": 0.9719,
+      "step": 11236
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.176085863500564e-06,
+      "loss": 1.0951,
+      "step": 11237
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.17453338112111e-06,
+      "loss": 1.0469,
+      "step": 11238
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.1729809187733e-06,
+      "loss": 1.0877,
+      "step": 11239
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.171428476494813e-06,
+      "loss": 1.1168,
+      "step": 11240
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.169876054323319e-06,
+      "loss": 1.0851,
+      "step": 11241
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.168323652296494e-06,
+      "loss": 1.0335,
+      "step": 11242
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.166771270452005e-06,
+      "loss": 1.2103,
+      "step": 11243
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.165218908827529e-06,
+      "loss": 1.0742,
+      "step": 11244
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.163666567460735e-06,
+      "loss": 1.2105,
+      "step": 11245
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.16211424638929e-06,
+      "loss": 1.087,
+      "step": 11246
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.160561945650874e-06,
+      "loss": 1.2533,
+      "step": 11247
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.159009665283146e-06,
+      "loss": 0.9741,
+      "step": 11248
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.157457405323784e-06,
+      "loss": 1.1775,
+      "step": 11249
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.155905165810449e-06,
+      "loss": 1.1036,
+      "step": 11250
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.154352946780813e-06,
+      "loss": 1.0232,
+      "step": 11251
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.152800748272548e-06,
+      "loss": 1.1869,
+      "step": 11252
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.151248570323314e-06,
+      "loss": 1.2495,
+      "step": 11253
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.149696412970785e-06,
+      "loss": 1.0867,
+      "step": 11254
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.148144276252621e-06,
+      "loss": 1.1735,
+      "step": 11255
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.146592160206492e-06,
+      "loss": 1.0488,
+      "step": 11256
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.145040064870058e-06,
+      "loss": 1.1505,
+      "step": 11257
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.14348799028099e-06,
+      "loss": 1.1859,
+      "step": 11258
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.141935936476949e-06,
+      "loss": 1.1582,
+      "step": 11259
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.1403839034956e-06,
+      "loss": 1.2669,
+      "step": 11260
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.138831891374605e-06,
+      "loss": 1.1342,
+      "step": 11261
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.137279900151627e-06,
+      "loss": 1.1057,
+      "step": 11262
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.135727929864328e-06,
+      "loss": 1.1467,
+      "step": 11263
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.134175980550373e-06,
+      "loss": 1.2526,
+      "step": 11264
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.132624052247417e-06,
+      "loss": 1.1382,
+      "step": 11265
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.131072144993128e-06,
+      "loss": 1.1976,
+      "step": 11266
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.12952025882516e-06,
+      "loss": 1.0874,
+      "step": 11267
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.127968393781178e-06,
+      "loss": 1.0314,
+      "step": 11268
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.126416549898836e-06,
+      "loss": 1.1058,
+      "step": 11269
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.124864727215796e-06,
+      "loss": 1.1666,
+      "step": 11270
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.123312925769716e-06,
+      "loss": 1.1302,
+      "step": 11271
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.121761145598253e-06,
+      "loss": 1.1675,
+      "step": 11272
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.120209386739063e-06,
+      "loss": 1.1485,
+      "step": 11273
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.118657649229806e-06,
+      "loss": 1.1329,
+      "step": 11274
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.117105933108132e-06,
+      "loss": 1.0394,
+      "step": 11275
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.115554238411704e-06,
+      "loss": 1.2001,
+      "step": 11276
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.114002565178172e-06,
+      "loss": 1.1862,
+      "step": 11277
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.112450913445195e-06,
+      "loss": 1.1522,
+      "step": 11278
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.11089928325042e-06,
+      "loss": 1.1871,
+      "step": 11279
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.109347674631508e-06,
+      "loss": 1.0707,
+      "step": 11280
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.107796087626106e-06,
+      "loss": 1.1801,
+      "step": 11281
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.10624452227187e-06,
+      "loss": 1.2017,
+      "step": 11282
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.10469297860645e-06,
+      "loss": 1.1178,
+      "step": 11283
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.103141456667502e-06,
+      "loss": 1.1489,
+      "step": 11284
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.10158995649267e-06,
+      "loss": 1.0756,
+      "step": 11285
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.100038478119605e-06,
+      "loss": 1.2316,
+      "step": 11286
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.098487021585966e-06,
+      "loss": 1.1117,
+      "step": 11287
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.096935586929392e-06,
+      "loss": 1.191,
+      "step": 11288
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.095384174187535e-06,
+      "loss": 1.1802,
+      "step": 11289
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.093832783398045e-06,
+      "loss": 0.9622,
+      "step": 11290
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.09228141459857e-06,
+      "loss": 1.0504,
+      "step": 11291
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.090730067826752e-06,
+      "loss": 1.1346,
+      "step": 11292
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.089178743120245e-06,
+      "loss": 1.1309,
+      "step": 11293
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.087627440516687e-06,
+      "loss": 1.1303,
+      "step": 11294
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.086076160053732e-06,
+      "loss": 1.2645,
+      "step": 11295
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.084524901769017e-06,
+      "loss": 1.0699,
+      "step": 11296
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.082973665700194e-06,
+      "loss": 1.2417,
+      "step": 11297
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.0814224518849e-06,
+      "loss": 0.9602,
+      "step": 11298
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.079871260360786e-06,
+      "loss": 1.2157,
+      "step": 11299
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.078320091165486e-06,
+      "loss": 1.0532,
+      "step": 11300
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.07676894433665e-06,
+      "loss": 1.1947,
+      "step": 11301
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.075217819911913e-06,
+      "loss": 1.1984,
+      "step": 11302
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.073666717928924e-06,
+      "loss": 1.1298,
+      "step": 11303
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.072115638425316e-06,
+      "loss": 1.0356,
+      "step": 11304
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.070564581438734e-06,
+      "loss": 1.186,
+      "step": 11305
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.069013547006816e-06,
+      "loss": 1.2375,
+      "step": 11306
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.067462535167204e-06,
+      "loss": 1.0158,
+      "step": 11307
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.065911545957528e-06,
+      "loss": 1.2529,
+      "step": 11308
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.064360579415437e-06,
+      "loss": 1.1342,
+      "step": 11309
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.06280963557856e-06,
+      "loss": 1.1247,
+      "step": 11310
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.06125871448454e-06,
+      "loss": 1.1635,
+      "step": 11311
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.059707816171007e-06,
+      "loss": 1.2669,
+      "step": 11312
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.058156940675603e-06,
+      "loss": 1.1558,
+      "step": 11313
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.056606088035958e-06,
+      "loss": 1.1689,
+      "step": 11314
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.05505525828971e-06,
+      "loss": 1.1201,
+      "step": 11315
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.05350445147449e-06,
+      "loss": 1.2232,
+      "step": 11316
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.051953667627936e-06,
+      "loss": 1.1221,
+      "step": 11317
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.050402906787677e-06,
+      "loss": 1.154,
+      "step": 11318
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.048852168991348e-06,
+      "loss": 1.052,
+      "step": 11319
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.047301454276577e-06,
+      "loss": 1.2175,
+      "step": 11320
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.045750762681e-06,
+      "loss": 1.1446,
+      "step": 11321
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.044200094242245e-06,
+      "loss": 1.1573,
+      "step": 11322
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.042649448997943e-06,
+      "loss": 1.1576,
+      "step": 11323
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.041098826985722e-06,
+      "loss": 1.1487,
+      "step": 11324
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.039548228243216e-06,
+      "loss": 1.1008,
+      "step": 11325
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.037997652808046e-06,
+      "loss": 0.9637,
+      "step": 11326
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.036447100717847e-06,
+      "loss": 1.0941,
+      "step": 11327
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.034896572010241e-06,
+      "loss": 1.0695,
+      "step": 11328
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.033346066722858e-06,
+      "loss": 1.1299,
+      "step": 11329
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.031795584893323e-06,
+      "loss": 0.9194,
+      "step": 11330
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.030245126559262e-06,
+      "loss": 1.0665,
+      "step": 11331
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.028694691758299e-06,
+      "loss": 1.2166,
+      "step": 11332
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.027144280528055e-06,
+      "loss": 1.0901,
+      "step": 11333
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.025593892906163e-06,
+      "loss": 1.1101,
+      "step": 11334
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.024043528930242e-06,
+      "loss": 1.1118,
+      "step": 11335
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.022493188637913e-06,
+      "loss": 1.0874,
+      "step": 11336
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.020942872066801e-06,
+      "loss": 1.1158,
+      "step": 11337
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.019392579254525e-06,
+      "loss": 1.138,
+      "step": 11338
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.017842310238709e-06,
+      "loss": 1.1401,
+      "step": 11339
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.01629206505697e-06,
+      "loss": 1.1215,
+      "step": 11340
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.014741843746931e-06,
+      "loss": 1.0434,
+      "step": 11341
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.013191646346208e-06,
+      "loss": 1.2888,
+      "step": 11342
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.011641472892424e-06,
+      "loss": 1.0742,
+      "step": 11343
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.010091323423192e-06,
+      "loss": 1.2119,
+      "step": 11344
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.008541197976136e-06,
+      "loss": 1.213,
+      "step": 11345
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.006991096588865e-06,
+      "loss": 1.0316,
+      "step": 11346
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.005441019299004e-06,
+      "loss": 1.1807,
+      "step": 11347
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.003890966144162e-06,
+      "loss": 1.1568,
+      "step": 11348
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.002340937161957e-06,
+      "loss": 1.1597,
+      "step": 11349
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.000790932390004e-06,
+      "loss": 1.171,
+      "step": 11350
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.999240951865917e-06,
+      "loss": 1.1416,
+      "step": 11351
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.997690995627307e-06,
+      "loss": 1.0606,
+      "step": 11352
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.996141063711791e-06,
+      "loss": 1.1331,
+      "step": 11353
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.994591156156977e-06,
+      "loss": 1.0744,
+      "step": 11354
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.993041273000483e-06,
+      "loss": 1.0259,
+      "step": 11355
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.991491414279912e-06,
+      "loss": 1.1441,
+      "step": 11356
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.989941580032879e-06,
+      "loss": 1.1074,
+      "step": 11357
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.988391770296994e-06,
+      "loss": 0.9849,
+      "step": 11358
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.986841985109867e-06,
+      "loss": 1.161,
+      "step": 11359
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.985292224509103e-06,
+      "loss": 1.1444,
+      "step": 11360
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.983742488532314e-06,
+      "loss": 1.2127,
+      "step": 11361
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.982192777217105e-06,
+      "loss": 1.0632,
+      "step": 11362
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.980643090601086e-06,
+      "loss": 1.1357,
+      "step": 11363
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.97909342872186e-06,
+      "loss": 1.0438,
+      "step": 11364
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.977543791617034e-06,
+      "loss": 1.1349,
+      "step": 11365
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.975994179324212e-06,
+      "loss": 1.0104,
+      "step": 11366
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.974444591881001e-06,
+      "loss": 1.16,
+      "step": 11367
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.972895029325004e-06,
+      "loss": 1.2079,
+      "step": 11368
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.971345491693827e-06,
+      "loss": 1.0227,
+      "step": 11369
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.969795979025065e-06,
+      "loss": 1.207,
+      "step": 11370
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.968246491356326e-06,
+      "loss": 1.324,
+      "step": 11371
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.966697028725212e-06,
+      "loss": 1.2196,
+      "step": 11372
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.965147591169319e-06,
+      "loss": 1.1012,
+      "step": 11373
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.963598178726254e-06,
+      "loss": 1.2472,
+      "step": 11374
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.962048791433611e-06,
+      "loss": 1.0897,
+      "step": 11375
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.960499429328994e-06,
+      "loss": 1.0463,
+      "step": 11376
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.958950092449995e-06,
+      "loss": 1.1044,
+      "step": 11377
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.957400780834219e-06,
+      "loss": 1.0914,
+      "step": 11378
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.955851494519257e-06,
+      "loss": 1.1741,
+      "step": 11379
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.954302233542712e-06,
+      "loss": 1.1951,
+      "step": 11380
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.952752997942174e-06,
+      "loss": 1.2007,
+      "step": 11381
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.951203787755242e-06,
+      "loss": 1.2136,
+      "step": 11382
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.949654603019509e-06,
+      "loss": 1.19,
+      "step": 11383
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.94810544377257e-06,
+      "loss": 0.9735,
+      "step": 11384
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.946556310052017e-06,
+      "loss": 0.9343,
+      "step": 11385
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.945007201895449e-06,
+      "loss": 1.1733,
+      "step": 11386
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.943458119340449e-06,
+      "loss": 1.1577,
+      "step": 11387
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.941909062424615e-06,
+      "loss": 1.239,
+      "step": 11388
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.940360031185536e-06,
+      "loss": 1.2627,
+      "step": 11389
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.938811025660802e-06,
+      "loss": 0.9956,
+      "step": 11390
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.937262045888004e-06,
+      "loss": 1.2022,
+      "step": 11391
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.935713091904732e-06,
+      "loss": 1.241,
+      "step": 11392
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.934164163748572e-06,
+      "loss": 1.1799,
+      "step": 11393
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.932615261457113e-06,
+      "loss": 1.0763,
+      "step": 11394
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.931066385067944e-06,
+      "loss": 1.0229,
+      "step": 11395
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.929517534618651e-06,
+      "loss": 1.2802,
+      "step": 11396
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.927968710146815e-06,
+      "loss": 0.9833,
+      "step": 11397
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.92641991169003e-06,
+      "loss": 1.2224,
+      "step": 11398
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.924871139285873e-06,
+      "loss": 1.0809,
+      "step": 11399
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.923322392971935e-06,
+      "loss": 1.1802,
+      "step": 11400
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.921773672785793e-06,
+      "loss": 1.1974,
+      "step": 11401
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.920224978765035e-06,
+      "loss": 1.2025,
+      "step": 11402
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.918676310947239e-06,
+      "loss": 1.185,
+      "step": 11403
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.917127669369992e-06,
+      "loss": 1.1667,
+      "step": 11404
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.915579054070869e-06,
+      "loss": 1.0972,
+      "step": 11405
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.914030465087457e-06,
+      "loss": 1.0741,
+      "step": 11406
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.912481902457328e-06,
+      "loss": 1.2102,
+      "step": 11407
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.910933366218068e-06,
+      "loss": 1.1772,
+      "step": 11408
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90938485640725e-06,
+      "loss": 1.152,
+      "step": 11409
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.907836373062456e-06,
+      "loss": 1.0723,
+      "step": 11410
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.906287916221259e-06,
+      "loss": 1.1222,
+      "step": 11411
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90473948592124e-06,
+      "loss": 1.2348,
+      "step": 11412
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.903191082199972e-06,
+      "loss": 1.1924,
+      "step": 11413
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.901642705095031e-06,
+      "loss": 1.1595,
+      "step": 11414
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90009435464399e-06,
+      "loss": 1.1231,
+      "step": 11415
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.898546030884421e-06,
+      "loss": 1.1736,
+      "step": 11416
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.896997733853904e-06,
+      "loss": 1.132,
+      "step": 11417
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.895449463590011e-06,
+      "loss": 1.2418,
+      "step": 11418
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.893901220130306e-06,
+      "loss": 1.195,
+      "step": 11419
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.89235300351237e-06,
+      "loss": 1.0425,
+      "step": 11420
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.890804813773767e-06,
+      "loss": 1.0568,
+      "step": 11421
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.889256650952072e-06,
+      "loss": 1.0726,
+      "step": 11422
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.887708515084848e-06,
+      "loss": 1.0834,
+      "step": 11423
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.88616040620967e-06,
+      "loss": 1.0685,
+      "step": 11424
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.8846123243641e-06,
+      "loss": 1.0565,
+      "step": 11425
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.883064269585711e-06,
+      "loss": 1.2133,
+      "step": 11426
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.881516241912066e-06,
+      "loss": 1.1241,
+      "step": 11427
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.879968241380735e-06,
+      "loss": 1.2111,
+      "step": 11428
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.878420268029278e-06,
+      "loss": 1.0736,
+      "step": 11429
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.876872321895264e-06,
+      "loss": 1.1271,
+      "step": 11430
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.875324403016255e-06,
+      "loss": 1.2444,
+      "step": 11431
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.873776511429817e-06,
+      "loss": 1.1011,
+      "step": 11432
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.872228647173509e-06,
+      "loss": 1.0764,
+      "step": 11433
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.870680810284897e-06,
+      "loss": 1.1784,
+      "step": 11434
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.869133000801539e-06,
+      "loss": 1.1631,
+      "step": 11435
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.867585218760997e-06,
+      "loss": 1.1977,
+      "step": 11436
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.866037464200832e-06,
+      "loss": 1.2139,
+      "step": 11437
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.864489737158604e-06,
+      "loss": 1.2365,
+      "step": 11438
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.86294203767187e-06,
+      "loss": 1.0379,
+      "step": 11439
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.861394365778188e-06,
+      "loss": 1.2347,
+      "step": 11440
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.859846721515116e-06,
+      "loss": 1.1565,
+      "step": 11441
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.858299104920213e-06,
+      "loss": 1.06,
+      "step": 11442
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.85675151603103e-06,
+      "loss": 1.0468,
+      "step": 11443
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.855203954885128e-06,
+      "loss": 1.0227,
+      "step": 11444
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.853656421520059e-06,
+      "loss": 1.0864,
+      "step": 11445
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.852108915973377e-06,
+      "loss": 1.0057,
+      "step": 11446
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.850561438282636e-06,
+      "loss": 1.1427,
+      "step": 11447
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.84901398848539e-06,
+      "loss": 1.001,
+      "step": 11448
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.847466566619185e-06,
+      "loss": 1.1176,
+      "step": 11449
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.845919172721581e-06,
+      "loss": 1.23,
+      "step": 11450
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.844371806830123e-06,
+      "loss": 0.9846,
+      "step": 11451
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.842824468982363e-06,
+      "loss": 1.0579,
+      "step": 11452
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.84127715921585e-06,
+      "loss": 1.1523,
+      "step": 11453
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.839729877568135e-06,
+      "loss": 1.299,
+      "step": 11454
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.838182624076759e-06,
+      "loss": 1.075,
+      "step": 11455
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.836635398779278e-06,
+      "loss": 1.055,
+      "step": 11456
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.83508820171323e-06,
+      "loss": 1.2228,
+      "step": 11457
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.833541032916168e-06,
+      "loss": 0.9919,
+      "step": 11458
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.831993892425636e-06,
+      "loss": 1.2125,
+      "step": 11459
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.830446780279175e-06,
+      "loss": 1.192,
+      "step": 11460
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.828899696514335e-06,
+      "loss": 1.0268,
+      "step": 11461
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.827352641168652e-06,
+      "loss": 0.9619,
+      "step": 11462
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.825805614279672e-06,
+      "loss": 1.1282,
+      "step": 11463
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.824258615884937e-06,
+      "loss": 1.0822,
+      "step": 11464
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.82271164602199e-06,
+      "loss": 1.0839,
+      "step": 11465
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.821164704728364e-06,
+      "loss": 1.0665,
+      "step": 11466
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.819617792041607e-06,
+      "loss": 1.0867,
+      "step": 11467
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.818070907999254e-06,
+      "loss": 1.0887,
+      "step": 11468
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.816524052638846e-06,
+      "loss": 1.1042,
+      "step": 11469
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.814977225997916e-06,
+      "loss": 1.0557,
+      "step": 11470
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.813430428114006e-06,
+      "loss": 1.0536,
+      "step": 11471
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.811883659024647e-06,
+      "loss": 1.1783,
+      "step": 11472
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.810336918767381e-06,
+      "loss": 1.1121,
+      "step": 11473
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.808790207379736e-06,
+      "loss": 1.1001,
+      "step": 11474
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.807243524899253e-06,
+      "loss": 1.2395,
+      "step": 11475
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.805696871363459e-06,
+      "loss": 1.2038,
+      "step": 11476
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.804150246809891e-06,
+      "loss": 1.0843,
+      "step": 11477
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.802603651276079e-06,
+      "loss": 1.2487,
+      "step": 11478
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.801057084799555e-06,
+      "loss": 1.2154,
+      "step": 11479
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.79951054741785e-06,
+      "loss": 1.1619,
+      "step": 11480
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.797964039168493e-06,
+      "loss": 1.1678,
+      "step": 11481
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.796417560089014e-06,
+      "loss": 1.137,
+      "step": 11482
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.794871110216942e-06,
+      "loss": 1.1478,
+      "step": 11483
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.793324689589802e-06,
+      "loss": 1.1492,
+      "step": 11484
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.791778298245126e-06,
+      "loss": 1.0883,
+      "step": 11485
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.790231936220435e-06,
+      "loss": 1.0974,
+      "step": 11486
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.788685603553259e-06,
+      "loss": 1.0547,
+      "step": 11487
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.787139300281117e-06,
+      "loss": 1.0955,
+      "step": 11488
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.785593026441543e-06,
+      "loss": 1.15,
+      "step": 11489
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.78404678207205e-06,
+      "loss": 1.2127,
+      "step": 11490
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.78250056721017e-06,
+      "loss": 1.2265,
+      "step": 11491
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.780954381893415e-06,
+      "loss": 1.1429,
+      "step": 11492
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.779408226159318e-06,
+      "loss": 1.0299,
+      "step": 11493
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.77786210004539e-06,
+      "loss": 0.9068,
+      "step": 11494
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.776316003589156e-06,
+      "loss": 1.2488,
+      "step": 11495
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.774769936828133e-06,
+      "loss": 1.2228,
+      "step": 11496
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.773223899799843e-06,
+      "loss": 1.1158,
+      "step": 11497
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.771677892541796e-06,
+      "loss": 0.9451,
+      "step": 11498
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.77013191509152e-06,
+      "loss": 1.0467,
+      "step": 11499
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.768585967486518e-06,
+      "loss": 1.1352,
+      "step": 11500
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.767040049764318e-06,
+      "loss": 1.1248,
+      "step": 11501
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.765494161962428e-06,
+      "loss": 1.1446,
+      "step": 11502
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.763948304118368e-06,
+      "loss": 1.148,
+      "step": 11503
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.762402476269645e-06,
+      "loss": 1.2413,
+      "step": 11504
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.760856678453776e-06,
+      "loss": 1.0791,
+      "step": 11505
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.759310910708268e-06,
+      "loss": 1.1592,
+      "step": 11506
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.757765173070637e-06,
+      "loss": 1.219,
+      "step": 11507
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.756219465578393e-06,
+      "loss": 1.2354,
+      "step": 11508
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.754673788269046e-06,
+      "loss": 1.0815,
+      "step": 11509
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.7531281411801e-06,
+      "loss": 1.0545,
+      "step": 11510
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.75158252434907e-06,
+      "loss": 1.1501,
+      "step": 11511
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.75003693781346e-06,
+      "loss": 1.1925,
+      "step": 11512
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.748491381610777e-06,
+      "loss": 1.2213,
+      "step": 11513
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.746945855778528e-06,
+      "loss": 1.1936,
+      "step": 11514
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.745400360354219e-06,
+      "loss": 1.1999,
+      "step": 11515
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.74385489537535e-06,
+      "loss": 1.1807,
+      "step": 11516
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.742309460879433e-06,
+      "loss": 1.0659,
+      "step": 11517
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.740764056903962e-06,
+      "loss": 1.0682,
+      "step": 11518
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.73921868348645e-06,
+      "loss": 1.1751,
+      "step": 11519
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.737673340664387e-06,
+      "loss": 1.0277,
+      "step": 11520
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.736128028475283e-06,
+      "loss": 1.0681,
+      "step": 11521
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.734582746956633e-06,
+      "loss": 1.0925,
+      "step": 11522
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.733037496145941e-06,
+      "loss": 1.2245,
+      "step": 11523
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.7314922760807e-06,
+      "loss": 1.1765,
+      "step": 11524
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.729947086798413e-06,
+      "loss": 1.229,
+      "step": 11525
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.728401928336574e-06,
+      "loss": 1.1359,
+      "step": 11526
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.726856800732682e-06,
+      "loss": 1.0993,
+      "step": 11527
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.725311704024229e-06,
+      "loss": 1.2392,
+      "step": 11528
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.723766638248714e-06,
+      "loss": 1.273,
+      "step": 11529
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.722221603443627e-06,
+      "loss": 1.1263,
+      "step": 11530
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.720676599646467e-06,
+      "loss": 1.1365,
+      "step": 11531
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.719131626894721e-06,
+      "loss": 1.0891,
+      "step": 11532
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.717586685225884e-06,
+      "loss": 1.2515,
+      "step": 11533
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.716041774677448e-06,
+      "loss": 0.9708,
+      "step": 11534
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.714496895286903e-06,
+      "loss": 1.0507,
+      "step": 11535
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.712952047091735e-06,
+      "loss": 1.1974,
+      "step": 11536
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.711407230129438e-06,
+      "loss": 1.2088,
+      "step": 11537
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.709862444437495e-06,
+      "loss": 1.1315,
+      "step": 11538
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.708317690053401e-06,
+      "loss": 1.0723,
+      "step": 11539
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.706772967014633e-06,
+      "loss": 1.0378,
+      "step": 11540
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.705228275358686e-06,
+      "loss": 1.0305,
+      "step": 11541
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.703683615123037e-06,
+      "loss": 0.8928,
+      "step": 11542
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.702138986345176e-06,
+      "loss": 0.9922,
+      "step": 11543
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.700594389062586e-06,
+      "loss": 1.1438,
+      "step": 11544
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.699049823312747e-06,
+      "loss": 1.0635,
+      "step": 11545
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.697505289133146e-06,
+      "loss": 1.1259,
+      "step": 11546
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.695960786561258e-06,
+      "loss": 1.1077,
+      "step": 11547
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.69441631563457e-06,
+      "loss": 0.9788,
+      "step": 11548
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.692871876390554e-06,
+      "loss": 1.0865,
+      "step": 11549
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.691327468866695e-06,
+      "loss": 1.0122,
+      "step": 11550
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.68978309310047e-06,
+      "loss": 1.3051,
+      "step": 11551
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.688238749129357e-06,
+      "loss": 1.2564,
+      "step": 11552
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.68669443699083e-06,
+      "loss": 1.0437,
+      "step": 11553
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.685150156722369e-06,
+      "loss": 1.1034,
+      "step": 11554
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.683605908361446e-06,
+      "loss": 1.0915,
+      "step": 11555
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.682061691945536e-06,
+      "loss": 1.2646,
+      "step": 11556
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.680517507512112e-06,
+      "loss": 1.0684,
+      "step": 11557
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.67897335509865e-06,
+      "loss": 1.1014,
+      "step": 11558
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.677429234742615e-06,
+      "loss": 1.1256,
+      "step": 11559
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.675885146481489e-06,
+      "loss": 1.1114,
+      "step": 11560
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.674341090352731e-06,
+      "loss": 1.0998,
+      "step": 11561
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.672797066393822e-06,
+      "loss": 1.082,
+      "step": 11562
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.671253074642218e-06,
+      "loss": 1.1451,
+      "step": 11563
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.6697091151354e-06,
+      "loss": 1.1204,
+      "step": 11564
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.668165187910826e-06,
+      "loss": 1.0723,
+      "step": 11565
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.66662129300597e-06,
+      "loss": 1.0655,
+      "step": 11566
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.665077430458288e-06,
+      "loss": 0.9624,
+      "step": 11567
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.663533600305256e-06,
+      "loss": 1.1068,
+      "step": 11568
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.66198980258433e-06,
+      "loss": 1.1143,
+      "step": 11569
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.660446037332979e-06,
+      "loss": 1.1393,
+      "step": 11570
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.658902304588662e-06,
+      "loss": 1.1118,
+      "step": 11571
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.657358604388843e-06,
+      "loss": 0.9265,
+      "step": 11572
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.65581493677098e-06,
+      "loss": 1.0933,
+      "step": 11573
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.654271301772537e-06,
+      "loss": 1.0554,
+      "step": 11574
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.652727699430971e-06,
+      "loss": 1.0981,
+      "step": 11575
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.651184129783742e-06,
+      "loss": 1.1063,
+      "step": 11576
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.649640592868306e-06,
+      "loss": 1.1165,
+      "step": 11577
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.648097088722125e-06,
+      "loss": 1.1576,
+      "step": 11578
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.64655361738265e-06,
+      "loss": 1.0372,
+      "step": 11579
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.645010178887338e-06,
+      "loss": 1.2554,
+      "step": 11580
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.643466773273643e-06,
+      "loss": 1.1097,
+      "step": 11581
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.641923400579022e-06,
+      "loss": 1.0834,
+      "step": 11582
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.640380060840921e-06,
+      "loss": 1.2056,
+      "step": 11583
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.638836754096802e-06,
+      "loss": 1.1531,
+      "step": 11584
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.637293480384111e-06,
+      "loss": 1.2251,
+      "step": 11585
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.635750239740302e-06,
+      "loss": 0.9988,
+      "step": 11586
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.634207032202819e-06,
+      "loss": 1.2027,
+      "step": 11587
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.632663857809119e-06,
+      "loss": 1.1722,
+      "step": 11588
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.631120716596642e-06,
+      "loss": 1.0639,
+      "step": 11589
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.629577608602842e-06,
+      "loss": 1.0177,
+      "step": 11590
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.628034533865162e-06,
+      "loss": 1.2862,
+      "step": 11591
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.62649149242105e-06,
+      "loss": 1.2095,
+      "step": 11592
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.624948484307948e-06,
+      "loss": 1.2147,
+      "step": 11593
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.623405509563305e-06,
+      "loss": 1.1361,
+      "step": 11594
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.621862568224559e-06,
+      "loss": 1.2629,
+      "step": 11595
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.620319660329159e-06,
+      "loss": 1.1662,
+      "step": 11596
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.61877678591454e-06,
+      "loss": 1.1704,
+      "step": 11597
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.61723394501815e-06,
+      "loss": 1.0435,
+      "step": 11598
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.615691137677423e-06,
+      "loss": 1.0216,
+      "step": 11599
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.614148363929802e-06,
+      "loss": 1.0198,
+      "step": 11600
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.612605623812721e-06,
+      "loss": 1.1373,
+      "step": 11601
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.611062917363627e-06,
+      "loss": 1.1728,
+      "step": 11602
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.609520244619948e-06,
+      "loss": 1.0446,
+      "step": 11603
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.607977605619125e-06,
+      "loss": 1.1667,
+      "step": 11604
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.60643500039859e-06,
+      "loss": 1.1762,
+      "step": 11605
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.604892428995782e-06,
+      "loss": 1.1604,
+      "step": 11606
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.603349891448128e-06,
+      "loss": 1.1272,
+      "step": 11607
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.601807387793068e-06,
+      "loss": 1.0329,
+      "step": 11608
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.600264918068028e-06,
+      "loss": 1.1408,
+      "step": 11609
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.598722482310445e-06,
+      "loss": 0.9774,
+      "step": 11610
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.597180080557743e-06,
+      "loss": 1.1038,
+      "step": 11611
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.595637712847357e-06,
+      "loss": 1.2655,
+      "step": 11612
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.594095379216712e-06,
+      "loss": 1.1017,
+      "step": 11613
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.59255307970324e-06,
+      "loss": 1.1476,
+      "step": 11614
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.591010814344364e-06,
+      "loss": 1.0784,
+      "step": 11615
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.589468583177514e-06,
+      "loss": 1.1711,
+      "step": 11616
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.58792638624011e-06,
+      "loss": 1.153,
+      "step": 11617
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.586384223569582e-06,
+      "loss": 1.1603,
+      "step": 11618
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.58484209520335e-06,
+      "loss": 1.073,
+      "step": 11619
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.58330000117884e-06,
+      "loss": 1.1219,
+      "step": 11620
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.581757941533471e-06,
+      "loss": 1.1938,
+      "step": 11621
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.580215916304668e-06,
+      "loss": 1.0844,
+      "step": 11622
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.578673925529848e-06,
+      "loss": 1.0852,
+      "step": 11623
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.577131969246433e-06,
+      "loss": 0.9307,
+      "step": 11624
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.575590047491839e-06,
+      "loss": 1.1717,
+      "step": 11625
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.574048160303487e-06,
+      "loss": 1.0771,
+      "step": 11626
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.57250630771879e-06,
+      "loss": 1.0976,
+      "step": 11627
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.57096448977517e-06,
+      "loss": 1.1696,
+      "step": 11628
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.569422706510037e-06,
+      "loss": 1.259,
+      "step": 11629
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.567880957960807e-06,
+      "loss": 1.239,
+      "step": 11630
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.566339244164897e-06,
+      "loss": 1.2008,
+      "step": 11631
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.564797565159715e-06,
+      "loss": 1.1771,
+      "step": 11632
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.563255920982678e-06,
+      "loss": 1.2501,
+      "step": 11633
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.561714311671192e-06,
+      "loss": 1.1952,
+      "step": 11634
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.560172737262671e-06,
+      "loss": 1.114,
+      "step": 11635
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.558631197794522e-06,
+      "loss": 0.9704,
+      "step": 11636
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.557089693304158e-06,
+      "loss": 1.1393,
+      "step": 11637
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.55554822382898e-06,
+      "loss": 1.1545,
+      "step": 11638
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.554006789406402e-06,
+      "loss": 1.1202,
+      "step": 11639
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.552465390073825e-06,
+      "loss": 1.1035,
+      "step": 11640
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.550924025868656e-06,
+      "loss": 1.0631,
+      "step": 11641
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.549382696828299e-06,
+      "loss": 1.2707,
+      "step": 11642
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.547841402990158e-06,
+      "loss": 1.1077,
+      "step": 11643
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.546300144391634e-06,
+      "loss": 1.2208,
+      "step": 11644
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.544758921070134e-06,
+      "loss": 1.1253,
+      "step": 11645
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.543217733063053e-06,
+      "loss": 0.9139,
+      "step": 11646
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.541676580407793e-06,
+      "loss": 1.1342,
+      "step": 11647
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.540135463141755e-06,
+      "loss": 1.0595,
+      "step": 11648
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.538594381302337e-06,
+      "loss": 1.1056,
+      "step": 11649
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.537053334926932e-06,
+      "loss": 1.228,
+      "step": 11650
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.535512324052943e-06,
+      "loss": 0.9891,
+      "step": 11651
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.53397134871776e-06,
+      "loss": 1.1352,
+      "step": 11652
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.532430408958785e-06,
+      "loss": 1.1962,
+      "step": 11653
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.530889504813405e-06,
+      "loss": 1.0374,
+      "step": 11654
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.529348636319017e-06,
+      "loss": 1.0537,
+      "step": 11655
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.527807803513012e-06,
+      "loss": 1.0518,
+      "step": 11656
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.526267006432785e-06,
+      "loss": 1.133,
+      "step": 11657
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.52472624511572e-06,
+      "loss": 1.1715,
+      "step": 11658
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.523185519599211e-06,
+      "loss": 1.1619,
+      "step": 11659
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.521644829920647e-06,
+      "loss": 1.0802,
+      "step": 11660
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.520104176117416e-06,
+      "loss": 1.0617,
+      "step": 11661
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.518563558226903e-06,
+      "loss": 1.0521,
+      "step": 11662
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.517022976286497e-06,
+      "loss": 1.0885,
+      "step": 11663
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.515482430333582e-06,
+      "loss": 1.1485,
+      "step": 11664
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.513941920405544e-06,
+      "loss": 1.2396,
+      "step": 11665
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.512401446539759e-06,
+      "loss": 1.2044,
+      "step": 11666
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.510861008773618e-06,
+      "loss": 1.2343,
+      "step": 11667
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.509320607144505e-06,
+      "loss": 1.0371,
+      "step": 11668
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.507780241689795e-06,
+      "loss": 1.0649,
+      "step": 11669
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.506239912446869e-06,
+      "loss": 1.1996,
+      "step": 11670
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.504699619453111e-06,
+      "loss": 1.1079,
+      "step": 11671
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.503159362745891e-06,
+      "loss": 1.1726,
+      "step": 11672
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.501619142362596e-06,
+      "loss": 1.1957,
+      "step": 11673
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.500078958340595e-06,
+      "loss": 0.9272,
+      "step": 11674
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.498538810717267e-06,
+      "loss": 1.241,
+      "step": 11675
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.496998699529986e-06,
+      "loss": 1.2156,
+      "step": 11676
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.495458624816127e-06,
+      "loss": 1.134,
+      "step": 11677
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.493918586613062e-06,
+      "loss": 1.2577,
+      "step": 11678
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.492378584958165e-06,
+      "loss": 1.1165,
+      "step": 11679
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.490838619888803e-06,
+      "loss": 1.0627,
+      "step": 11680
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.489298691442354e-06,
+      "loss": 1.0287,
+      "step": 11681
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.487758799656179e-06,
+      "loss": 1.1849,
+      "step": 11682
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.486218944567652e-06,
+      "loss": 1.156,
+      "step": 11683
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.48467912621414e-06,
+      "loss": 1.1151,
+      "step": 11684
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.483139344633009e-06,
+      "loss": 1.2031,
+      "step": 11685
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.481599599861622e-06,
+      "loss": 1.0795,
+      "step": 11686
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.480059891937352e-06,
+      "loss": 1.0976,
+      "step": 11687
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.478520220897555e-06,
+      "loss": 1.1829,
+      "step": 11688
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.476980586779601e-06,
+      "loss": 1.1734,
+      "step": 11689
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.475440989620845e-06,
+      "loss": 1.0948,
+      "step": 11690
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.473901429458657e-06,
+      "loss": 1.179,
+      "step": 11691
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.47236190633039e-06,
+      "loss": 1.0467,
+      "step": 11692
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.470822420273409e-06,
+      "loss": 1.2233,
+      "step": 11693
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.469282971325067e-06,
+      "loss": 1.1288,
+      "step": 11694
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.46774355952273e-06,
+      "loss": 1.041,
+      "step": 11695
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.46620418490375e-06,
+      "loss": 1.2399,
+      "step": 11696
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.464664847505482e-06,
+      "loss": 1.1833,
+      "step": 11697
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.463125547365283e-06,
+      "loss": 1.1806,
+      "step": 11698
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.461586284520509e-06,
+      "loss": 1.0127,
+      "step": 11699
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.460047059008508e-06,
+      "loss": 1.0986,
+      "step": 11700
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.45850787086664e-06,
+      "loss": 1.1526,
+      "step": 11701
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.45696872013225e-06,
+      "loss": 1.2045,
+      "step": 11702
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.455429606842694e-06,
+      "loss": 1.139,
+      "step": 11703
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.453890531035315e-06,
+      "loss": 1.0986,
+      "step": 11704
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.45235149274747e-06,
+      "loss": 1.1921,
+      "step": 11705
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.450812492016499e-06,
+      "loss": 1.2132,
+      "step": 11706
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.449273528879756e-06,
+      "loss": 1.0839,
+      "step": 11707
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.447734603374582e-06,
+      "loss": 1.2499,
+      "step": 11708
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.446195715538326e-06,
+      "loss": 1.0909,
+      "step": 11709
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.444656865408328e-06,
+      "loss": 1.1217,
+      "step": 11710
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.443118053021936e-06,
+      "loss": 1.1795,
+      "step": 11711
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.441579278416487e-06,
+      "loss": 1.0577,
+      "step": 11712
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.44004054162933e-06,
+      "loss": 1.1861,
+      "step": 11713
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.438501842697799e-06,
+      "loss": 1.0404,
+      "step": 11714
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.436963181659234e-06,
+      "loss": 1.0099,
+      "step": 11715
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.435424558550979e-06,
+      "loss": 1.1146,
+      "step": 11716
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.433885973410368e-06,
+      "loss": 1.0837,
+      "step": 11717
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.432347426274739e-06,
+      "loss": 1.2163,
+      "step": 11718
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.430808917181427e-06,
+      "loss": 1.1191,
+      "step": 11719
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.429270446167768e-06,
+      "loss": 1.2179,
+      "step": 11720
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.427732013271095e-06,
+      "loss": 1.1127,
+      "step": 11721
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.426193618528744e-06,
+      "loss": 0.9819,
+      "step": 11722
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.424655261978044e-06,
+      "loss": 1.2345,
+      "step": 11723
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.42311694365633e-06,
+      "loss": 1.0873,
+      "step": 11724
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.421578663600926e-06,
+      "loss": 1.024,
+      "step": 11725
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.420040421849172e-06,
+      "loss": 1.0721,
+      "step": 11726
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.418502218438385e-06,
+      "loss": 1.1662,
+      "step": 11727
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.416964053405902e-06,
+      "loss": 1.0421,
+      "step": 11728
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.415425926789042e-06,
+      "loss": 1.1076,
+      "step": 11729
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.41388783862514e-06,
+      "loss": 1.0993,
+      "step": 11730
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.41234978895151e-06,
+      "loss": 1.0505,
+      "step": 11731
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.410811777805484e-06,
+      "loss": 1.2116,
+      "step": 11732
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.409273805224382e-06,
+      "loss": 1.1913,
+      "step": 11733
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.407735871245528e-06,
+      "loss": 1.2032,
+      "step": 11734
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.40619797590624e-06,
+      "loss": 1.0533,
+      "step": 11735
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.404660119243841e-06,
+      "loss": 1.0653,
+      "step": 11736
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.403122301295648e-06,
+      "loss": 1.3283,
+      "step": 11737
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.401584522098981e-06,
+      "loss": 1.1316,
+      "step": 11738
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.400046781691159e-06,
+      "loss": 1.108,
+      "step": 11739
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.398509080109495e-06,
+      "loss": 1.0615,
+      "step": 11740
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.396971417391304e-06,
+      "loss": 1.0852,
+      "step": 11741
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.395433793573904e-06,
+      "loss": 1.1725,
+      "step": 11742
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.393896208694607e-06,
+      "loss": 1.121,
+      "step": 11743
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.392358662790724e-06,
+      "loss": 1.0266,
+      "step": 11744
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.39082115589957e-06,
+      "loss": 1.1042,
+      "step": 11745
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.389283688058454e-06,
+      "loss": 1.1058,
+      "step": 11746
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.387746259304685e-06,
+      "loss": 1.1789,
+      "step": 11747
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.386208869675576e-06,
+      "loss": 1.2177,
+      "step": 11748
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.384671519208427e-06,
+      "loss": 0.9889,
+      "step": 11749
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.383134207940548e-06,
+      "loss": 1.1517,
+      "step": 11750
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.381596935909253e-06,
+      "loss": 1.0722,
+      "step": 11751
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.380059703151837e-06,
+      "loss": 1.1753,
+      "step": 11752
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.378522509705611e-06,
+      "loss": 1.07,
+      "step": 11753
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.376985355607872e-06,
+      "loss": 1.0482,
+      "step": 11754
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.375448240895927e-06,
+      "loss": 1.0759,
+      "step": 11755
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.373911165607075e-06,
+      "loss": 1.161,
+      "step": 11756
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.372374129778617e-06,
+      "loss": 1.1284,
+      "step": 11757
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.370837133447853e-06,
+      "loss": 1.1255,
+      "step": 11758
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.36930017665208e-06,
+      "loss": 1.1088,
+      "step": 11759
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.367763259428597e-06,
+      "loss": 1.2133,
+      "step": 11760
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.366226381814698e-06,
+      "loss": 0.9576,
+      "step": 11761
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.364689543847681e-06,
+      "loss": 1.1012,
+      "step": 11762
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.363152745564838e-06,
+      "loss": 0.9498,
+      "step": 11763
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.361615987003467e-06,
+      "loss": 1.136,
+      "step": 11764
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.360079268200854e-06,
+      "loss": 1.0643,
+      "step": 11765
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.358542589194297e-06,
+      "loss": 1.0646,
+      "step": 11766
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.357005950021081e-06,
+      "loss": 0.9211,
+      "step": 11767
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.355469350718503e-06,
+      "loss": 1.1055,
+      "step": 11768
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.353932791323845e-06,
+      "loss": 0.9869,
+      "step": 11769
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.352396271874398e-06,
+      "loss": 1.038,
+      "step": 11770
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.350859792407446e-06,
+      "loss": 1.1313,
+      "step": 11771
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.349323352960279e-06,
+      "loss": 0.9836,
+      "step": 11772
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.347786953570178e-06,
+      "loss": 1.1455,
+      "step": 11773
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.346250594274433e-06,
+      "loss": 1.0482,
+      "step": 11774
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.344714275110318e-06,
+      "loss": 1.1691,
+      "step": 11775
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.343177996115122e-06,
+      "loss": 1.1134,
+      "step": 11776
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.34164175732612e-06,
+      "loss": 1.1824,
+      "step": 11777
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.3401055587806e-06,
+      "loss": 1.204,
+      "step": 11778
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.338569400515833e-06,
+      "loss": 1.2185,
+      "step": 11779
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.337033282569102e-06,
+      "loss": 1.0817,
+      "step": 11780
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.335497204977682e-06,
+      "loss": 1.0055,
+      "step": 11781
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.333961167778853e-06,
+      "loss": 1.2359,
+      "step": 11782
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.332425171009882e-06,
+      "loss": 1.206,
+      "step": 11783
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.330889214708051e-06,
+      "loss": 0.9588,
+      "step": 11784
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.329353298910629e-06,
+      "loss": 1.2082,
+      "step": 11785
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.32781742365489e-06,
+      "loss": 1.0834,
+      "step": 11786
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.326281588978102e-06,
+      "loss": 1.1767,
+      "step": 11787
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.32474579491754e-06,
+      "loss": 1.1589,
+      "step": 11788
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.32321004151047e-06,
+      "loss": 1.2113,
+      "step": 11789
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.321674328794162e-06,
+      "loss": 0.9555,
+      "step": 11790
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.32013865680588e-06,
+      "loss": 1.1683,
+      "step": 11791
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.318603025582897e-06,
+      "loss": 1.1419,
+      "step": 11792
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.317067435162469e-06,
+      "loss": 1.2208,
+      "step": 11793
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.315531885581867e-06,
+      "loss": 0.958,
+      "step": 11794
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.313996376878353e-06,
+      "loss": 1.1851,
+      "step": 11795
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.31246090908919e-06,
+      "loss": 1.1514,
+      "step": 11796
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.310925482251633e-06,
+      "loss": 1.0641,
+      "step": 11797
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.30939009640295e-06,
+      "loss": 1.2167,
+      "step": 11798
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.307854751580396e-06,
+      "loss": 1.2317,
+      "step": 11799
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.306319447821234e-06,
+      "loss": 1.0484,
+      "step": 11800
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.304784185162714e-06,
+      "loss": 1.071,
+      "step": 11801
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.303248963642098e-06,
+      "loss": 1.0376,
+      "step": 11802
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.301713783296641e-06,
+      "loss": 1.1153,
+      "step": 11803
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.300178644163593e-06,
+      "loss": 1.1744,
+      "step": 11804
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.298643546280214e-06,
+      "loss": 1.2244,
+      "step": 11805
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.297108489683749e-06,
+      "loss": 1.0439,
+      "step": 11806
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.295573474411456e-06,
+      "loss": 1.2009,
+      "step": 11807
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.29403850050058e-06,
+      "loss": 1.1252,
+      "step": 11808
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.292503567988375e-06,
+      "loss": 1.0829,
+      "step": 11809
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.290968676912085e-06,
+      "loss": 0.9678,
+      "step": 11810
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.289433827308961e-06,
+      "loss": 1.1161,
+      "step": 11811
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.287899019216245e-06,
+      "loss": 1.0793,
+      "step": 11812
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.286364252671188e-06,
+      "loss": 1.135,
+      "step": 11813
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.284829527711026e-06,
+      "loss": 1.276,
+      "step": 11814
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.283294844373012e-06,
+      "loss": 1.1553,
+      "step": 11815
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.28176020269438e-06,
+      "loss": 1.2325,
+      "step": 11816
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.280225602712378e-06,
+      "loss": 1.1377,
+      "step": 11817
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.27869104446424e-06,
+      "loss": 1.1348,
+      "step": 11818
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.27715652798721e-06,
+      "loss": 1.2072,
+      "step": 11819
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.275622053318523e-06,
+      "loss": 1.0978,
+      "step": 11820
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.274087620495419e-06,
+      "loss": 1.2132,
+      "step": 11821
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.272553229555132e-06,
+      "loss": 1.0388,
+      "step": 11822
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.271018880534898e-06,
+      "loss": 1.1802,
+      "step": 11823
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.26948457347195e-06,
+      "loss": 1.0938,
+      "step": 11824
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.267950308403523e-06,
+      "loss": 1.1544,
+      "step": 11825
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.266416085366847e-06,
+      "loss": 1.1616,
+      "step": 11826
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.264881904399157e-06,
+      "loss": 1.1426,
+      "step": 11827
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.263347765537678e-06,
+      "loss": 1.1181,
+      "step": 11828
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.261813668819644e-06,
+      "loss": 1.1517,
+      "step": 11829
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.260279614282277e-06,
+      "loss": 1.1802,
+      "step": 11830
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.258745601962811e-06,
+      "loss": 1.0179,
+      "step": 11831
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.257211631898465e-06,
+      "loss": 1.049,
+      "step": 11832
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.255677704126466e-06,
+      "loss": 1.2017,
+      "step": 11833
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.254143818684042e-06,
+      "loss": 1.1401,
+      "step": 11834
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.252609975608413e-06,
+      "loss": 1.0157,
+      "step": 11835
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.251076174936802e-06,
+      "loss": 1.0016,
+      "step": 11836
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.249542416706428e-06,
+      "loss": 1.1267,
+      "step": 11837
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.248008700954512e-06,
+      "loss": 1.1279,
+      "step": 11838
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.246475027718272e-06,
+      "loss": 1.0431,
+      "step": 11839
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.244941397034924e-06,
+      "loss": 1.1138,
+      "step": 11840
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.243407808941692e-06,
+      "loss": 1.191,
+      "step": 11841
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.241874263475782e-06,
+      "loss": 0.9321,
+      "step": 11842
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.240340760674415e-06,
+      "loss": 1.1435,
+      "step": 11843
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.238807300574802e-06,
+      "loss": 1.1159,
+      "step": 11844
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.237273883214159e-06,
+      "loss": 0.9236,
+      "step": 11845
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.23574050862969e-06,
+      "loss": 1.1399,
+      "step": 11846
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.234207176858615e-06,
+      "loss": 1.2537,
+      "step": 11847
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.232673887938136e-06,
+      "loss": 1.1241,
+      "step": 11848
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.231140641905467e-06,
+      "loss": 1.1538,
+      "step": 11849
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.22960743879781e-06,
+      "loss": 0.9106,
+      "step": 11850
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.228074278652376e-06,
+      "loss": 1.0549,
+      "step": 11851
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.226541161506364e-06,
+      "loss": 1.1569,
+      "step": 11852
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.225008087396988e-06,
+      "loss": 1.0705,
+      "step": 11853
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.223475056361442e-06,
+      "loss": 1.1826,
+      "step": 11854
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.221942068436935e-06,
+      "loss": 1.0724,
+      "step": 11855
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.22040912366066e-06,
+      "loss": 1.1526,
+      "step": 11856
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.218876222069827e-06,
+      "loss": 1.0553,
+      "step": 11857
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.217343363701627e-06,
+      "loss": 1.2005,
+      "step": 11858
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.215810548593261e-06,
+      "loss": 1.28,
+      "step": 11859
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.214277776781928e-06,
+      "loss": 1.2041,
+      "step": 11860
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.21274504830482e-06,
+      "loss": 1.1771,
+      "step": 11861
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.211212363199132e-06,
+      "loss": 1.1573,
+      "step": 11862
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.209679721502063e-06,
+      "loss": 1.1474,
+      "step": 11863
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.208147123250799e-06,
+      "loss": 1.0419,
+      "step": 11864
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.206614568482538e-06,
+      "loss": 1.0688,
+      "step": 11865
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.205082057234463e-06,
+      "loss": 0.9005,
+      "step": 11866
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.203549589543772e-06,
+      "loss": 1.2177,
+      "step": 11867
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.202017165447646e-06,
+      "loss": 1.2981,
+      "step": 11868
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.200484784983278e-06,
+      "loss": 1.2007,
+      "step": 11869
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.19895244818785e-06,
+      "loss": 1.1429,
+      "step": 11870
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.197420155098551e-06,
+      "loss": 1.0887,
+      "step": 11871
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.195887905752563e-06,
+      "loss": 1.0584,
+      "step": 11872
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.19435570018707e-06,
+      "loss": 1.2195,
+      "step": 11873
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.192823538439253e-06,
+      "loss": 1.0662,
+      "step": 11874
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.191291420546296e-06,
+      "loss": 1.0988,
+      "step": 11875
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.189759346545375e-06,
+      "loss": 1.2543,
+      "step": 11876
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.188227316473672e-06,
+      "loss": 1.1424,
+      "step": 11877
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.186695330368364e-06,
+      "loss": 1.2688,
+      "step": 11878
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.185163388266628e-06,
+      "loss": 1.0059,
+      "step": 11879
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.183631490205636e-06,
+      "loss": 1.1026,
+      "step": 11880
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.18209963622257e-06,
+      "loss": 1.1088,
+      "step": 11881
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.180567826354597e-06,
+      "loss": 1.0724,
+      "step": 11882
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.179036060638895e-06,
+      "loss": 0.9725,
+      "step": 11883
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.177504339112629e-06,
+      "loss": 1.2041,
+      "step": 11884
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.175972661812977e-06,
+      "loss": 1.2088,
+      "step": 11885
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.1744410287771e-06,
+      "loss": 1.0386,
+      "step": 11886
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.172909440042172e-06,
+      "loss": 1.1821,
+      "step": 11887
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.171377895645359e-06,
+      "loss": 1.181,
+      "step": 11888
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.169846395623826e-06,
+      "loss": 1.1599,
+      "step": 11889
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.168314940014742e-06,
+      "loss": 1.2215,
+      "step": 11890
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.166783528855263e-06,
+      "loss": 1.1419,
+      "step": 11891
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.165252162182559e-06,
+      "loss": 1.1226,
+      "step": 11892
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.163720840033788e-06,
+      "loss": 1.0807,
+      "step": 11893
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.162189562446112e-06,
+      "loss": 1.0509,
+      "step": 11894
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.160658329456692e-06,
+      "loss": 1.1289,
+      "step": 11895
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.159127141102684e-06,
+      "loss": 0.9519,
+      "step": 11896
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.157595997421246e-06,
+      "loss": 1.0998,
+      "step": 11897
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.156064898449536e-06,
+      "loss": 1.1448,
+      "step": 11898
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.154533844224706e-06,
+      "loss": 1.1404,
+      "step": 11899
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.153002834783916e-06,
+      "loss": 1.1771,
+      "step": 11900
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.151471870164311e-06,
+      "loss": 1.0625,
+      "step": 11901
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.14994095040305e-06,
+      "loss": 1.2906,
+      "step": 11902
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.14841007553728e-06,
+      "loss": 1.1751,
+      "step": 11903
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.146879245604155e-06,
+      "loss": 1.0789,
+      "step": 11904
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.145348460640817e-06,
+      "loss": 1.1045,
+      "step": 11905
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.14381772068442e-06,
+      "loss": 1.0927,
+      "step": 11906
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.142287025772107e-06,
+      "loss": 1.2491,
+      "step": 11907
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.140756375941026e-06,
+      "loss": 1.1969,
+      "step": 11908
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.139225771228318e-06,
+      "loss": 1.2123,
+      "step": 11909
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.137695211671131e-06,
+      "loss": 1.1536,
+      "step": 11910
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.1361646973066e-06,
+      "loss": 1.235,
+      "step": 11911
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.134634228171875e-06,
+      "loss": 1.21,
+      "step": 11912
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.133103804304087e-06,
+      "loss": 1.0389,
+      "step": 11913
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.131573425740384e-06,
+      "loss": 1.1551,
+      "step": 11914
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.130043092517895e-06,
+      "loss": 1.1885,
+      "step": 11915
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.12851280467376e-06,
+      "loss": 1.0604,
+      "step": 11916
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.126982562245118e-06,
+      "loss": 1.002,
+      "step": 11917
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.1254523652691e-06,
+      "loss": 0.9399,
+      "step": 11918
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.123922213782843e-06,
+      "loss": 1.1893,
+      "step": 11919
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.122392107823474e-06,
+      "loss": 1.1545,
+      "step": 11920
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.120862047428127e-06,
+      "loss": 1.0763,
+      "step": 11921
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.119332032633933e-06,
+      "loss": 0.9744,
+      "step": 11922
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.117802063478019e-06,
+      "loss": 1.2301,
+      "step": 11923
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.116272139997511e-06,
+      "loss": 1.0723,
+      "step": 11924
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.114742262229542e-06,
+      "loss": 1.2467,
+      "step": 11925
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.113212430211232e-06,
+      "loss": 1.0909,
+      "step": 11926
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.111682643979707e-06,
+      "loss": 1.0967,
+      "step": 11927
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.110152903572094e-06,
+      "loss": 1.0815,
+      "step": 11928
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.108623209025507e-06,
+      "loss": 1.1011,
+      "step": 11929
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.107093560377079e-06,
+      "loss": 1.1258,
+      "step": 11930
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.105563957663918e-06,
+      "loss": 0.9635,
+      "step": 11931
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.104034400923153e-06,
+      "loss": 0.9918,
+      "step": 11932
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.102504890191892e-06,
+      "loss": 1.0707,
+      "step": 11933
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.100975425507261e-06,
+      "loss": 1.0153,
+      "step": 11934
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.09944600690637e-06,
+      "loss": 1.1283,
+      "step": 11935
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.097916634426335e-06,
+      "loss": 1.1526,
+      "step": 11936
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.096387308104269e-06,
+      "loss": 1.1282,
+      "step": 11937
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.094858027977288e-06,
+      "loss": 1.0109,
+      "step": 11938
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.093328794082496e-06,
+      "loss": 1.1545,
+      "step": 11939
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.09179960645701e-06,
+      "loss": 1.2489,
+      "step": 11940
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.090270465137932e-06,
+      "loss": 1.2707,
+      "step": 11941
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.088741370162376e-06,
+      "loss": 1.1364,
+      "step": 11942
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.087212321567445e-06,
+      "loss": 1.1214,
+      "step": 11943
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.085683319390247e-06,
+      "loss": 1.1557,
+      "step": 11944
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.084154363667884e-06,
+      "loss": 1.1622,
+      "step": 11945
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.08262545443746e-06,
+      "loss": 1.2997,
+      "step": 11946
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.081096591736077e-06,
+      "loss": 1.2251,
+      "step": 11947
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.079567775600839e-06,
+      "loss": 1.2246,
+      "step": 11948
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.07803900606884e-06,
+      "loss": 1.1856,
+      "step": 11949
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.076510283177186e-06,
+      "loss": 1.0391,
+      "step": 11950
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.074981606962967e-06,
+      "loss": 1.0853,
+      "step": 11951
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.073452977463285e-06,
+      "loss": 1.1878,
+      "step": 11952
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.071924394715233e-06,
+      "loss": 1.1414,
+      "step": 11953
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.070395858755906e-06,
+      "loss": 1.0565,
+      "step": 11954
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.068867369622394e-06,
+      "loss": 1.1401,
+      "step": 11955
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.067338927351797e-06,
+      "loss": 1.0103,
+      "step": 11956
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.065810531981196e-06,
+      "loss": 1.0174,
+      "step": 11957
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.064282183547688e-06,
+      "loss": 1.0825,
+      "step": 11958
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.062753882088356e-06,
+      "loss": 1.0871,
+      "step": 11959
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.061225627640294e-06,
+      "loss": 1.0369,
+      "step": 11960
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.059697420240579e-06,
+      "loss": 1.2776,
+      "step": 11961
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.058169259926306e-06,
+      "loss": 1.0339,
+      "step": 11962
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.056641146734552e-06,
+      "loss": 1.1422,
+      "step": 11963
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.055113080702403e-06,
+      "loss": 1.0744,
+      "step": 11964
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.053585061866937e-06,
+      "loss": 1.1549,
+      "step": 11965
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.052057090265241e-06,
+      "loss": 1.0603,
+      "step": 11966
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.05052916593439e-06,
+      "loss": 1.1325,
+      "step": 11967
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.049001288911462e-06,
+      "loss": 1.1865,
+      "step": 11968
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.047473459233535e-06,
+      "loss": 0.9875,
+      "step": 11969
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.045945676937688e-06,
+      "loss": 1.1007,
+      "step": 11970
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.044417942060988e-06,
+      "loss": 1.1798,
+      "step": 11971
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.042890254640515e-06,
+      "loss": 1.1699,
+      "step": 11972
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.041362614713341e-06,
+      "loss": 1.0573,
+      "step": 11973
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.039835022316534e-06,
+      "loss": 1.0832,
+      "step": 11974
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.03830747748717e-06,
+      "loss": 1.1673,
+      "step": 11975
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.036779980262312e-06,
+      "loss": 1.2208,
+      "step": 11976
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.035252530679031e-06,
+      "loss": 1.0804,
+      "step": 11977
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.033725128774391e-06,
+      "loss": 1.0977,
+      "step": 11978
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.032197774585463e-06,
+      "loss": 1.1372,
+      "step": 11979
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.030670468149305e-06,
+      "loss": 1.1295,
+      "step": 11980
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.029143209502985e-06,
+      "loss": 1.0549,
+      "step": 11981
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.027615998683561e-06,
+      "loss": 1.116,
+      "step": 11982
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.026088835728098e-06,
+      "loss": 1.1603,
+      "step": 11983
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.024561720673654e-06,
+      "loss": 1.0788,
+      "step": 11984
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.023034653557287e-06,
+      "loss": 1.1421,
+      "step": 11985
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.021507634416055e-06,
+      "loss": 1.0532,
+      "step": 11986
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.019980663287015e-06,
+      "loss": 1.1018,
+      "step": 11987
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.01845374020722e-06,
+      "loss": 1.2081,
+      "step": 11988
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.016926865213726e-06,
+      "loss": 1.0179,
+      "step": 11989
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.015400038343582e-06,
+      "loss": 1.1256,
+      "step": 11990
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.013873259633846e-06,
+      "loss": 1.028,
+      "step": 11991
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.012346529121564e-06,
+      "loss": 1.19,
+      "step": 11992
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.010819846843788e-06,
+      "loss": 1.1243,
+      "step": 11993
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.00929321283756e-06,
+      "loss": 0.9557,
+      "step": 11994
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.007766627139935e-06,
+      "loss": 0.9872,
+      "step": 11995
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.006240089787953e-06,
+      "loss": 1.1563,
+      "step": 11996
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.004713600818662e-06,
+      "loss": 1.2176,
+      "step": 11997
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.003187160269104e-06,
+      "loss": 1.0386,
+      "step": 11998
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.001660768176322e-06,
+      "loss": 1.135,
+      "step": 11999
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.000134424577351e-06,
+      "loss": 1.0864,
+      "step": 12000
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.99860812950924e-06,
+      "loss": 1.3003,
+      "step": 12001
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.997081883009025e-06,
+      "loss": 1.0538,
+      "step": 12002
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.99555568511374e-06,
+      "loss": 1.08,
+      "step": 12003
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.994029535860426e-06,
+      "loss": 1.107,
+      "step": 12004
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.992503435286113e-06,
+      "loss": 1.1615,
+      "step": 12005
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.99097738342784e-06,
+      "loss": 1.2094,
+      "step": 12006
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.989451380322636e-06,
+      "loss": 1.1616,
+      "step": 12007
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.987925426007537e-06,
+      "loss": 1.0044,
+      "step": 12008
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.986399520519567e-06,
+      "loss": 1.1246,
+      "step": 12009
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.984873663895762e-06,
+      "loss": 1.057,
+      "step": 12010
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.983347856173146e-06,
+      "loss": 1.1047,
+      "step": 12011
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.981822097388745e-06,
+      "loss": 1.1332,
+      "step": 12012
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.98029638757959e-06,
+      "loss": 1.147,
+      "step": 12013
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.978770726782697e-06,
+      "loss": 1.0845,
+      "step": 12014
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.9772451150351e-06,
+      "loss": 1.1521,
+      "step": 12015
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.975719552373811e-06,
+      "loss": 1.0412,
+      "step": 12016
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.974194038835857e-06,
+      "loss": 1.1049,
+      "step": 12017
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.972668574458255e-06,
+      "loss": 1.138,
+      "step": 12018
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.971143159278025e-06,
+      "loss": 1.3757,
+      "step": 12019
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.969617793332184e-06,
+      "loss": 1.0239,
+      "step": 12020
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.96809247665775e-06,
+      "loss": 1.1289,
+      "step": 12021
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.966567209291732e-06,
+      "loss": 1.2066,
+      "step": 12022
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.96504199127115e-06,
+      "loss": 1.1412,
+      "step": 12023
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.963516822633014e-06,
+      "loss": 1.0716,
+      "step": 12024
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.961991703414337e-06,
+      "loss": 1.0763,
+      "step": 12025
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.960466633652125e-06,
+      "loss": 1.1025,
+      "step": 12026
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.958941613383391e-06,
+      "loss": 1.2113,
+      "step": 12027
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.957416642645142e-06,
+      "loss": 1.1531,
+      "step": 12028
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.955891721474385e-06,
+      "loss": 1.0163,
+      "step": 12029
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.954366849908121e-06,
+      "loss": 1.0592,
+      "step": 12030
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.952842027983362e-06,
+      "loss": 1.2441,
+      "step": 12031
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.951317255737103e-06,
+      "loss": 1.1949,
+      "step": 12032
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.949792533206352e-06,
+      "loss": 1.2259,
+      "step": 12033
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.948267860428104e-06,
+      "loss": 1.078,
+      "step": 12034
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.946743237439363e-06,
+      "loss": 1.0136,
+      "step": 12035
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.945218664277123e-06,
+      "loss": 0.9721,
+      "step": 12036
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.943694140978387e-06,
+      "loss": 1.1152,
+      "step": 12037
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.942169667580142e-06,
+      "loss": 1.1161,
+      "step": 12038
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.940645244119391e-06,
+      "loss": 1.0855,
+      "step": 12039
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.939120870633123e-06,
+      "loss": 1.3062,
+      "step": 12040
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.937596547158331e-06,
+      "loss": 1.0655,
+      "step": 12041
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.936072273732002e-06,
+      "loss": 1.0551,
+      "step": 12042
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.934548050391133e-06,
+      "loss": 0.9763,
+      "step": 12043
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.933023877172707e-06,
+      "loss": 1.1357,
+      "step": 12044
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.931499754113713e-06,
+      "loss": 1.3036,
+      "step": 12045
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.929975681251136e-06,
+      "loss": 0.974,
+      "step": 12046
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.928451658621964e-06,
+      "loss": 1.1213,
+      "step": 12047
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.926927686263176e-06,
+      "loss": 1.2127,
+      "step": 12048
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.925403764211757e-06,
+      "loss": 1.0201,
+      "step": 12049
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.923879892504686e-06,
+      "loss": 1.1935,
+      "step": 12050
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.922356071178945e-06,
+      "loss": 1.12,
+      "step": 12051
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.920832300271512e-06,
+      "loss": 1.1135,
+      "step": 12052
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.919308579819368e-06,
+      "loss": 1.054,
+      "step": 12053
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.91778490985948e-06,
+      "loss": 1.0662,
+      "step": 12054
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.916261290428833e-06,
+      "loss": 1.1893,
+      "step": 12055
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.914737721564393e-06,
+      "loss": 1.0296,
+      "step": 12056
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.91321420330314e-06,
+      "loss": 1.1854,
+      "step": 12057
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.911690735682037e-06,
+      "loss": 1.0111,
+      "step": 12058
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.910167318738058e-06,
+      "loss": 0.8989,
+      "step": 12059
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.908643952508177e-06,
+      "loss": 1.0563,
+      "step": 12060
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.907120637029353e-06,
+      "loss": 1.1459,
+      "step": 12061
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.905597372338558e-06,
+      "loss": 1.0912,
+      "step": 12062
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.904074158472754e-06,
+      "loss": 1.1299,
+      "step": 12063
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.902550995468908e-06,
+      "loss": 1.148,
+      "step": 12064
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.90102788336398e-06,
+      "loss": 1.2796,
+      "step": 12065
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.899504822194933e-06,
+      "loss": 1.0555,
+      "step": 12066
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.897981811998725e-06,
+      "loss": 1.1609,
+      "step": 12067
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.896458852812319e-06,
+      "loss": 0.9526,
+      "step": 12068
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.894935944672668e-06,
+      "loss": 1.2108,
+      "step": 12069
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.893413087616733e-06,
+      "loss": 1.1115,
+      "step": 12070
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.891890281681465e-06,
+      "loss": 1.06,
+      "step": 12071
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.890367526903822e-06,
+      "loss": 1.1131,
+      "step": 12072
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.888844823320753e-06,
+      "loss": 1.0601,
+      "step": 12073
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.887322170969215e-06,
+      "loss": 1.0317,
+      "step": 12074
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.88579956988615e-06,
+      "loss": 1.1234,
+      "step": 12075
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.884277020108515e-06,
+      "loss": 1.2215,
+      "step": 12076
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.882754521673253e-06,
+      "loss": 1.0031,
+      "step": 12077
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.881232074617313e-06,
+      "loss": 1.0819,
+      "step": 12078
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.879709678977638e-06,
+      "loss": 1.1221,
+      "step": 12079
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.878187334791176e-06,
+      "loss": 1.085,
+      "step": 12080
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.876665042094867e-06,
+      "loss": 1.075,
+      "step": 12081
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.875142800925652e-06,
+      "loss": 1.1361,
+      "step": 12082
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.87362061132047e-06,
+      "loss": 0.9546,
+      "step": 12083
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.872098473316263e-06,
+      "loss": 1.3589,
+      "step": 12084
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.870576386949971e-06,
+      "loss": 1.1244,
+      "step": 12085
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.869054352258526e-06,
+      "loss": 1.0562,
+      "step": 12086
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.867532369278867e-06,
+      "loss": 1.2019,
+      "step": 12087
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.866010438047921e-06,
+      "loss": 1.0332,
+      "step": 12088
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.864488558602633e-06,
+      "loss": 1.1024,
+      "step": 12089
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.862966730979922e-06,
+      "loss": 1.1249,
+      "step": 12090
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.861444955216726e-06,
+      "loss": 1.0411,
+      "step": 12091
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.85992323134997e-06,
+      "loss": 1.0465,
+      "step": 12092
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.858401559416586e-06,
+      "loss": 1.1155,
+      "step": 12093
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.856879939453495e-06,
+      "loss": 0.9463,
+      "step": 12094
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.855358371497629e-06,
+      "loss": 1.1535,
+      "step": 12095
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.853836855585904e-06,
+      "loss": 1.2392,
+      "step": 12096
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.852315391755249e-06,
+      "loss": 1.0421,
+      "step": 12097
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.850793980042582e-06,
+      "loss": 0.9465,
+      "step": 12098
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.849272620484825e-06,
+      "loss": 1.0768,
+      "step": 12099
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.847751313118898e-06,
+      "loss": 1.2128,
+      "step": 12100
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.846230057981714e-06,
+      "loss": 1.232,
+      "step": 12101
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.844708855110196e-06,
+      "loss": 1.0824,
+      "step": 12102
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.843187704541252e-06,
+      "loss": 1.116,
+      "step": 12103
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.841666606311805e-06,
+      "loss": 1.2405,
+      "step": 12104
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.840145560458756e-06,
+      "loss": 1.0391,
+      "step": 12105
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.838624567019026e-06,
+      "loss": 1.1642,
+      "step": 12106
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.83710362602952e-06,
+      "loss": 1.1765,
+      "step": 12107
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.83558273752715e-06,
+      "loss": 1.1632,
+      "step": 12108
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.834061901548818e-06,
+      "loss": 1.039,
+      "step": 12109
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.832541118131437e-06,
+      "loss": 1.1473,
+      "step": 12110
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.831020387311907e-06,
+      "loss": 1.0424,
+      "step": 12111
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.829499709127134e-06,
+      "loss": 0.9279,
+      "step": 12112
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.82797908361402e-06,
+      "loss": 1.1484,
+      "step": 12113
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.826458510809467e-06,
+      "loss": 1.1583,
+      "step": 12114
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.824937990750372e-06,
+      "loss": 1.1707,
+      "step": 12115
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.823417523473638e-06,
+      "loss": 1.0643,
+      "step": 12116
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.821897109016156e-06,
+      "loss": 1.207,
+      "step": 12117
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.820376747414829e-06,
+      "loss": 1.2227,
+      "step": 12118
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.818856438706546e-06,
+      "loss": 1.0569,
+      "step": 12119
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.817336182928204e-06,
+      "loss": 1.1053,
+      "step": 12120
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.815815980116692e-06,
+      "loss": 1.0772,
+      "step": 12121
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.814295830308904e-06,
+      "loss": 1.1265,
+      "step": 12122
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.812775733541727e-06,
+      "loss": 1.0465,
+      "step": 12123
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.811255689852053e-06,
+      "loss": 1.0798,
+      "step": 12124
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.809735699276765e-06,
+      "loss": 1.1829,
+      "step": 12125
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.808215761852751e-06,
+      "loss": 1.2376,
+      "step": 12126
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.806695877616894e-06,
+      "loss": 1.0667,
+      "step": 12127
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.805176046606079e-06,
+      "loss": 1.0607,
+      "step": 12128
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.803656268857184e-06,
+      "loss": 1.1813,
+      "step": 12129
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.802136544407096e-06,
+      "loss": 1.1217,
+      "step": 12130
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.800616873292686e-06,
+      "loss": 1.1303,
+      "step": 12131
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.79909725555084e-06,
+      "loss": 1.2134,
+      "step": 12132
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.797577691218428e-06,
+      "loss": 1.2098,
+      "step": 12133
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.796058180332332e-06,
+      "loss": 0.9473,
+      "step": 12134
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.794538722929417e-06,
+      "loss": 1.0519,
+      "step": 12135
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.793019319046566e-06,
+      "loss": 1.207,
+      "step": 12136
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.791499968720644e-06,
+      "loss": 1.2698,
+      "step": 12137
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.789980671988524e-06,
+      "loss": 1.0326,
+      "step": 12138
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.78846142888707e-06,
+      "loss": 1.1052,
+      "step": 12139
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.786942239453156e-06,
+      "loss": 0.9019,
+      "step": 12140
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.785423103723644e-06,
+      "loss": 1.1736,
+      "step": 12141
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.783904021735405e-06,
+      "loss": 1.1306,
+      "step": 12142
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.782384993525293e-06,
+      "loss": 1.0397,
+      "step": 12143
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.780866019130175e-06,
+      "loss": 1.1365,
+      "step": 12144
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.779347098586916e-06,
+      "loss": 1.2612,
+      "step": 12145
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.77782823193237e-06,
+      "loss": 1.2083,
+      "step": 12146
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.776309419203399e-06,
+      "loss": 1.0972,
+      "step": 12147
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.774790660436857e-06,
+      "loss": 1.1844,
+      "step": 12148
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.773271955669603e-06,
+      "loss": 1.2311,
+      "step": 12149
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.77175330493849e-06,
+      "loss": 1.1641,
+      "step": 12150
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.770234708280372e-06,
+      "loss": 1.1188,
+      "step": 12151
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.768716165732097e-06,
+      "loss": 1.1559,
+      "step": 12152
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.767197677330522e-06,
+      "loss": 1.1247,
+      "step": 12153
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.765679243112491e-06,
+      "loss": 1.1372,
+      "step": 12154
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.764160863114857e-06,
+      "loss": 1.2375,
+      "step": 12155
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.76264253737446e-06,
+      "loss": 1.4616,
+      "step": 12156
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.76112426592815e-06,
+      "loss": 1.2687,
+      "step": 12157
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.75960604881277e-06,
+      "loss": 1.0948,
+      "step": 12158
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.758087886065163e-06,
+      "loss": 1.1699,
+      "step": 12159
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.756569777722167e-06,
+      "loss": 1.1815,
+      "step": 12160
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.755051723820628e-06,
+      "loss": 1.1551,
+      "step": 12161
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.753533724397379e-06,
+      "loss": 1.15,
+      "step": 12162
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.752015779489262e-06,
+      "loss": 1.3193,
+      "step": 12163
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.750497889133109e-06,
+      "loss": 1.1053,
+      "step": 12164
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.74898005336576e-06,
+      "loss": 1.1219,
+      "step": 12165
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.747462272224039e-06,
+      "loss": 1.1157,
+      "step": 12166
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.745944545744787e-06,
+      "loss": 1.177,
+      "step": 12167
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.744426873964834e-06,
+      "loss": 1.1737,
+      "step": 12168
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.742909256921005e-06,
+      "loss": 1.0376,
+      "step": 12169
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.741391694650133e-06,
+      "loss": 1.1408,
+      "step": 12170
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.73987418718904e-06,
+      "loss": 1.1465,
+      "step": 12171
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.738356734574556e-06,
+      "loss": 1.2205,
+      "step": 12172
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.7368393368435e-06,
+      "loss": 1.2242,
+      "step": 12173
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.7353219940327e-06,
+      "loss": 1.0002,
+      "step": 12174
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.733804706178974e-06,
+      "loss": 1.0995,
+      "step": 12175
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.732287473319143e-06,
+      "loss": 1.1245,
+      "step": 12176
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.730770295490026e-06,
+      "loss": 1.1675,
+      "step": 12177
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.729253172728441e-06,
+      "loss": 1.1996,
+      "step": 12178
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.727736105071202e-06,
+      "loss": 1.1082,
+      "step": 12179
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.726219092555127e-06,
+      "loss": 1.1311,
+      "step": 12180
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.724702135217026e-06,
+      "loss": 1.0997,
+      "step": 12181
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.723185233093716e-06,
+      "loss": 1.008,
+      "step": 12182
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.721668386222001e-06,
+      "loss": 1.1246,
+      "step": 12183
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.720151594638695e-06,
+      "loss": 1.1512,
+      "step": 12184
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.718634858380604e-06,
+      "loss": 1.1917,
+      "step": 12185
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.717118177484538e-06,
+      "loss": 1.0753,
+      "step": 12186
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.715601551987298e-06,
+      "loss": 1.1794,
+      "step": 12187
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.714084981925691e-06,
+      "loss": 1.1339,
+      "step": 12188
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.712568467336519e-06,
+      "loss": 1.1669,
+      "step": 12189
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.711052008256583e-06,
+      "loss": 1.0643,
+      "step": 12190
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.709535604722685e-06,
+      "loss": 1.0932,
+      "step": 12191
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.708019256771617e-06,
+      "loss": 1.1721,
+      "step": 12192
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.706502964440188e-06,
+      "loss": 1.032,
+      "step": 12193
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.704986727765182e-06,
+      "loss": 1.2193,
+      "step": 12194
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.703470546783401e-06,
+      "loss": 1.2644,
+      "step": 12195
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.701954421531634e-06,
+      "loss": 1.1481,
+      "step": 12196
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.700438352046677e-06,
+      "loss": 1.0281,
+      "step": 12197
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.698922338365317e-06,
+      "loss": 0.9269,
+      "step": 12198
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.697406380524346e-06,
+      "loss": 1.1453,
+      "step": 12199
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.695890478560548e-06,
+      "loss": 1.022,
+      "step": 12200
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.694374632510716e-06,
+      "loss": 1.1285,
+      "step": 12201
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.692858842411628e-06,
+      "loss": 1.0413,
+      "step": 12202
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.691343108300072e-06,
+      "loss": 1.1674,
+      "step": 12203
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.689827430212827e-06,
+      "loss": 1.1828,
+      "step": 12204
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.688311808186679e-06,
+      "loss": 1.0781,
+      "step": 12205
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.686796242258404e-06,
+      "loss": 1.1442,
+      "step": 12206
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.685280732464781e-06,
+      "loss": 1.1773,
+      "step": 12207
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.683765278842587e-06,
+      "loss": 1.153,
+      "step": 12208
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6822498814286e-06,
+      "loss": 1.0245,
+      "step": 12209
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.68073454025959e-06,
+      "loss": 1.1898,
+      "step": 12210
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.679219255372334e-06,
+      "loss": 1.1503,
+      "step": 12211
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.677704026803599e-06,
+      "loss": 1.1194,
+      "step": 12212
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.67618885459016e-06,
+      "loss": 1.2271,
+      "step": 12213
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.674673738768781e-06,
+      "loss": 1.1268,
+      "step": 12214
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.673158679376233e-06,
+      "loss": 1.0906,
+      "step": 12215
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.67164367644928e-06,
+      "loss": 1.0615,
+      "step": 12216
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.670128730024691e-06,
+      "loss": 1.1618,
+      "step": 12217
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.66861384013922e-06,
+      "loss": 1.1429,
+      "step": 12218
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.66709900682964e-06,
+      "loss": 1.1046,
+      "step": 12219
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.665584230132703e-06,
+      "loss": 1.1246,
+      "step": 12220
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.664069510085174e-06,
+      "loss": 1.1986,
+      "step": 12221
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.662554846723807e-06,
+      "loss": 1.2187,
+      "step": 12222
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.661040240085361e-06,
+      "loss": 1.1953,
+      "step": 12223
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.659525690206586e-06,
+      "loss": 1.1332,
+      "step": 12224
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.658011197124244e-06,
+      "loss": 1.0496,
+      "step": 12225
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.656496760875078e-06,
+      "loss": 1.2261,
+      "step": 12226
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.654982381495848e-06,
+      "loss": 1.1997,
+      "step": 12227
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.653468059023295e-06,
+      "loss": 1.1604,
+      "step": 12228
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.651953793494175e-06,
+      "loss": 1.1061,
+      "step": 12229
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.65043958494523e-06,
+      "loss": 1.0907,
+      "step": 12230
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.648925433413204e-06,
+      "loss": 1.0717,
+      "step": 12231
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.647411338934847e-06,
+      "loss": 1.0645,
+      "step": 12232
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.645897301546896e-06,
+      "loss": 1.0772,
+      "step": 12233
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.644383321286095e-06,
+      "loss": 1.2217,
+      "step": 12234
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.642869398189183e-06,
+      "loss": 1.176,
+      "step": 12235
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6413555322929e-06,
+      "loss": 1.0851,
+      "step": 12236
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.639841723633979e-06,
+      "loss": 1.1693,
+      "step": 12237
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.63832797224916e-06,
+      "loss": 1.216,
+      "step": 12238
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.636814278175176e-06,
+      "loss": 1.1575,
+      "step": 12239
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.63530064144876e-06,
+      "loss": 1.1714,
+      "step": 12240
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.633787062106642e-06,
+      "loss": 0.8665,
+      "step": 12241
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.632273540185554e-06,
+      "loss": 1.2198,
+      "step": 12242
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.630760075722222e-06,
+      "loss": 1.0176,
+      "step": 12243
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.629246668753378e-06,
+      "loss": 1.2068,
+      "step": 12244
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6277333193157435e-06,
+      "loss": 1.0924,
+      "step": 12245
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.626220027446045e-06,
+      "loss": 1.1868,
+      "step": 12246
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.624706793181005e-06,
+      "loss": 0.9643,
+      "step": 12247
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6231936165573475e-06,
+      "loss": 1.1147,
+      "step": 12248
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.621680497611788e-06,
+      "loss": 1.1055,
+      "step": 12249
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6201674363810475e-06,
+      "loss": 1.0862,
+      "step": 12250
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.618654432901849e-06,
+      "loss": 1.2092,
+      "step": 12251
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.617141487210901e-06,
+      "loss": 1.2521,
+      "step": 12252
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.615628599344924e-06,
+      "loss": 1.0551,
+      "step": 12253
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.614115769340626e-06,
+      "loss": 1.1965,
+      "step": 12254
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.612602997234723e-06,
+      "loss": 1.1055,
+      "step": 12255
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.611090283063924e-06,
+      "loss": 1.2988,
+      "step": 12256
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.609577626864939e-06,
+      "loss": 1.0212,
+      "step": 12257
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6080650286744735e-06,
+      "loss": 1.0476,
+      "step": 12258
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.606552488529238e-06,
+      "loss": 0.8681,
+      "step": 12259
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.605040006465931e-06,
+      "loss": 1.0739,
+      "step": 12260
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.603527582521264e-06,
+      "loss": 1.034,
+      "step": 12261
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.602015216731932e-06,
+      "loss": 1.0188,
+      "step": 12262
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.600502909134639e-06,
+      "loss": 1.0994,
+      "step": 12263
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.598990659766083e-06,
+      "loss": 1.042,
+      "step": 12264
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.597478468662964e-06,
+      "loss": 1.0558,
+      "step": 12265
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5959663358619765e-06,
+      "loss": 1.1751,
+      "step": 12266
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.594454261399816e-06,
+      "loss": 1.0557,
+      "step": 12267
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.592942245313175e-06,
+      "loss": 1.1048,
+      "step": 12268
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.591430287638747e-06,
+      "loss": 1.1893,
+      "step": 12269
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.589918388413224e-06,
+      "loss": 1.0754,
+      "step": 12270
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.588406547673292e-06,
+      "loss": 1.041,
+      "step": 12271
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.586894765455643e-06,
+      "loss": 0.9813,
+      "step": 12272
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.58538304179696e-06,
+      "loss": 0.9074,
+      "step": 12273
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.583871376733931e-06,
+      "loss": 0.9755,
+      "step": 12274
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.582359770303236e-06,
+      "loss": 1.1681,
+      "step": 12275
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.580848222541562e-06,
+      "loss": 0.8817,
+      "step": 12276
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.579336733485584e-06,
+      "loss": 1.1896,
+      "step": 12277
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.577825303171987e-06,
+      "loss": 1.1213,
+      "step": 12278
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.576313931637444e-06,
+      "loss": 1.0836,
+      "step": 12279
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5748026189186376e-06,
+      "loss": 1.1812,
+      "step": 12280
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.573291365052236e-06,
+      "loss": 1.0724,
+      "step": 12281
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.571780170074919e-06,
+      "loss": 1.0778,
+      "step": 12282
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.570269034023354e-06,
+      "loss": 1.2382,
+      "step": 12283
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.568757956934216e-06,
+      "loss": 1.0693,
+      "step": 12284
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.567246938844171e-06,
+      "loss": 1.1958,
+      "step": 12285
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.565735979789889e-06,
+      "loss": 1.0437,
+      "step": 12286
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.564225079808034e-06,
+      "loss": 1.1845,
+      "step": 12287
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.562714238935276e-06,
+      "loss": 1.1529,
+      "step": 12288
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5612034572082726e-06,
+      "loss": 1.1864,
+      "step": 12289
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.559692734663692e-06,
+      "loss": 1.2077,
+      "step": 12290
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5581820713381895e-06,
+      "loss": 1.1141,
+      "step": 12291
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.556671467268428e-06,
+      "loss": 1.1352,
+      "step": 12292
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.555160922491065e-06,
+      "loss": 1.0638,
+      "step": 12293
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.553650437042756e-06,
+      "loss": 1.0801,
+      "step": 12294
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.552140010960155e-06,
+      "loss": 1.2252,
+      "step": 12295
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5506296442799195e-06,
+      "loss": 0.9323,
+      "step": 12296
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.549119337038697e-06,
+      "loss": 1.22,
+      "step": 12297
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5476090892731425e-06,
+      "loss": 1.0007,
+      "step": 12298
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.546098901019902e-06,
+      "loss": 1.163,
+      "step": 12299
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.544588772315625e-06,
+      "loss": 1.0389,
+      "step": 12300
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.543078703196956e-06,
+      "loss": 1.0689,
+      "step": 12301
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.541568693700543e-06,
+      "loss": 1.0938,
+      "step": 12302
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.540058743863026e-06,
+      "loss": 1.0214,
+      "step": 12303
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.538548853721052e-06,
+      "loss": 1.1388,
+      "step": 12304
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.537039023311255e-06,
+      "loss": 1.1559,
+      "step": 12305
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.535529252670279e-06,
+      "loss": 1.0451,
+      "step": 12306
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.534019541834761e-06,
+      "loss": 1.0773,
+      "step": 12307
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.532509890841336e-06,
+      "loss": 1.1894,
+      "step": 12308
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.531000299726638e-06,
+      "loss": 0.9993,
+      "step": 12309
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5294907685273035e-06,
+      "loss": 1.1862,
+      "step": 12310
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.527981297279961e-06,
+      "loss": 1.2079,
+      "step": 12311
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.526471886021245e-06,
+      "loss": 1.1279,
+      "step": 12312
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5249625347877785e-06,
+      "loss": 1.0678,
+      "step": 12313
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.523453243616195e-06,
+      "loss": 1.2808,
+      "step": 12314
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.521944012543117e-06,
+      "loss": 1.0356,
+      "step": 12315
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.520434841605168e-06,
+      "loss": 1.1157,
+      "step": 12316
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.518925730838978e-06,
+      "loss": 1.0164,
+      "step": 12317
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.51741668028116e-06,
+      "loss": 1.1995,
+      "step": 12318
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.515907689968341e-06,
+      "loss": 1.138,
+      "step": 12319
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.514398759937135e-06,
+      "loss": 1.1264,
+      "step": 12320
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.512889890224163e-06,
+      "loss": 1.1355,
+      "step": 12321
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.511381080866036e-06,
+      "loss": 1.1072,
+      "step": 12322
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.509872331899375e-06,
+      "loss": 1.265,
+      "step": 12323
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.508363643360786e-06,
+      "loss": 0.9872,
+      "step": 12324
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.506855015286887e-06,
+      "loss": 1.136,
+      "step": 12325
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.505346447714281e-06,
+      "loss": 1.2118,
+      "step": 12326
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.503837940679584e-06,
+      "loss": 1.0548,
+      "step": 12327
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.502329494219396e-06,
+      "loss": 1.115,
+      "step": 12328
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.500821108370329e-06,
+      "loss": 1.0872,
+      "step": 12329
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.499312783168981e-06,
+      "loss": 1.1072,
+      "step": 12330
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.497804518651959e-06,
+      "loss": 1.2791,
+      "step": 12331
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.496296314855862e-06,
+      "loss": 1.1483,
+      "step": 12332
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.494788171817286e-06,
+      "loss": 1.1364,
+      "step": 12333
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.493280089572841e-06,
+      "loss": 1.2081,
+      "step": 12334
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.491772068159111e-06,
+      "loss": 1.0746,
+      "step": 12335
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4902641076127e-06,
+      "loss": 1.1258,
+      "step": 12336
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.488756207970196e-06,
+      "loss": 1.105,
+      "step": 12337
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4872483692681966e-06,
+      "loss": 1.1402,
+      "step": 12338
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.485740591543286e-06,
+      "loss": 1.184,
+      "step": 12339
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.484232874832059e-06,
+      "loss": 0.9681,
+      "step": 12340
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.482725219171101e-06,
+      "loss": 1.2375,
+      "step": 12341
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.481217624596999e-06,
+      "loss": 1.0716,
+      "step": 12342
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.479710091146337e-06,
+      "loss": 1.1568,
+      "step": 12343
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.478202618855701e-06,
+      "loss": 1.1129,
+      "step": 12344
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.476695207761668e-06,
+      "loss": 1.0959,
+      "step": 12345
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.475187857900825e-06,
+      "loss": 1.1202,
+      "step": 12346
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.473680569309746e-06,
+      "loss": 1.2235,
+      "step": 12347
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.472173342025011e-06,
+      "loss": 1.0454,
+      "step": 12348
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.470666176083193e-06,
+      "loss": 1.0905,
+      "step": 12349
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.469159071520871e-06,
+      "loss": 1.0363,
+      "step": 12350
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.467652028374613e-06,
+      "loss": 1.0032,
+      "step": 12351
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.466145046680994e-06,
+      "loss": 1.1927,
+      "step": 12352
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.464638126476581e-06,
+      "loss": 1.0452,
+      "step": 12353
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.463131267797947e-06,
+      "loss": 1.2138,
+      "step": 12354
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.461624470681654e-06,
+      "loss": 1.1421,
+      "step": 12355
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.460117735164271e-06,
+      "loss": 1.0115,
+      "step": 12356
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.458611061282363e-06,
+      "loss": 1.0749,
+      "step": 12357
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.457104449072488e-06,
+      "loss": 1.056,
+      "step": 12358
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.455597898571212e-06,
+      "loss": 1.0378,
+      "step": 12359
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.45409140981509e-06,
+      "loss": 0.9848,
+      "step": 12360
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.452584982840682e-06,
+      "loss": 1.1159,
+      "step": 12361
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.451078617684546e-06,
+      "loss": 1.3552,
+      "step": 12362
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.449572314383237e-06,
+      "loss": 1.166,
+      "step": 12363
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.448066072973306e-06,
+      "loss": 1.0618,
+      "step": 12364
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.446559893491306e-06,
+      "loss": 1.1417,
+      "step": 12365
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.445053775973787e-06,
+      "loss": 1.1219,
+      "step": 12366
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.443547720457302e-06,
+      "loss": 1.1005,
+      "step": 12367
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.442041726978394e-06,
+      "loss": 1.251,
+      "step": 12368
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.440535795573611e-06,
+      "loss": 1.1281,
+      "step": 12369
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.439029926279498e-06,
+      "loss": 1.2087,
+      "step": 12370
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.437524119132596e-06,
+      "loss": 1.128,
+      "step": 12371
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.436018374169447e-06,
+      "loss": 1.0399,
+      "step": 12372
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.434512691426593e-06,
+      "loss": 1.1417,
+      "step": 12373
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.43300707094057e-06,
+      "loss": 1.0503,
+      "step": 12374
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.431501512747916e-06,
+      "loss": 1.1433,
+      "step": 12375
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.429996016885167e-06,
+      "loss": 1.0269,
+      "step": 12376
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.428490583388859e-06,
+      "loss": 1.181,
+      "step": 12377
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4269852122955174e-06,
+      "loss": 1.2935,
+      "step": 12378
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.425479903641682e-06,
+      "loss": 1.0815,
+      "step": 12379
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.423974657463875e-06,
+      "loss": 1.1499,
+      "step": 12380
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4224694737986295e-06,
+      "loss": 1.1386,
+      "step": 12381
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.420964352682467e-06,
+      "loss": 1.0374,
+      "step": 12382
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4194592941519185e-06,
+      "loss": 1.3176,
+      "step": 12383
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.417954298243499e-06,
+      "loss": 0.9536,
+      "step": 12384
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.41644936499374e-06,
+      "loss": 1.0778,
+      "step": 12385
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.414944494439153e-06,
+      "loss": 1.1033,
+      "step": 12386
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.413439686616264e-06,
+      "loss": 1.0626,
+      "step": 12387
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.411934941561585e-06,
+      "loss": 1.3206,
+      "step": 12388
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.410430259311636e-06,
+      "loss": 1.1446,
+      "step": 12389
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.408925639902926e-06,
+      "loss": 1.0531,
+      "step": 12390
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.407421083371973e-06,
+      "loss": 1.1135,
+      "step": 12391
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4059165897552845e-06,
+      "loss": 1.1557,
+      "step": 12392
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.404412159089372e-06,
+      "loss": 1.0259,
+      "step": 12393
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.402907791410741e-06,
+      "loss": 1.2355,
+      "step": 12394
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.401403486755903e-06,
+      "loss": 1.1473,
+      "step": 12395
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3998992451613596e-06,
+      "loss": 1.1356,
+      "step": 12396
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3983950666636154e-06,
+      "loss": 1.1697,
+      "step": 12397
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.39689095129917e-06,
+      "loss": 1.2222,
+      "step": 12398
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.395386899104526e-06,
+      "loss": 1.1992,
+      "step": 12399
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.393882910116182e-06,
+      "loss": 1.1267,
+      "step": 12400
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.392378984370636e-06,
+      "loss": 1.1449,
+      "step": 12401
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3908751219043835e-06,
+      "loss": 1.1815,
+      "step": 12402
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.389371322753918e-06,
+      "loss": 1.103,
+      "step": 12403
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.387867586955735e-06,
+      "loss": 1.0916,
+      "step": 12404
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.386363914546322e-06,
+      "loss": 1.0885,
+      "step": 12405
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.384860305562171e-06,
+      "loss": 1.0839,
+      "step": 12406
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.383356760039768e-06,
+      "loss": 1.0604,
+      "step": 12407
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.381853278015603e-06,
+      "loss": 1.2853,
+      "step": 12408
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3803498595261565e-06,
+      "loss": 1.2341,
+      "step": 12409
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3788465046079195e-06,
+      "loss": 1.1921,
+      "step": 12410
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.377343213297366e-06,
+      "loss": 1.1391,
+      "step": 12411
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3758399856309805e-06,
+      "loss": 1.067,
+      "step": 12412
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.37433682164524e-06,
+      "loss": 1.2527,
+      "step": 12413
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3728337213766245e-06,
+      "loss": 1.0113,
+      "step": 12414
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.371330684861607e-06,
+      "loss": 1.1998,
+      "step": 12415
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.369827712136661e-06,
+      "loss": 1.1559,
+      "step": 12416
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.368324803238266e-06,
+      "loss": 1.2559,
+      "step": 12417
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.366821958202885e-06,
+      "loss": 1.0083,
+      "step": 12418
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.365319177066994e-06,
+      "loss": 1.3046,
+      "step": 12419
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.363816459867057e-06,
+      "loss": 1.102,
+      "step": 12420
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.362313806639543e-06,
+      "loss": 0.8889,
+      "step": 12421
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.360811217420915e-06,
+      "loss": 1.1651,
+      "step": 12422
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3593086922476375e-06,
+      "loss": 1.155,
+      "step": 12423
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.35780623115617e-06,
+      "loss": 1.0213,
+      "step": 12424
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.356303834182978e-06,
+      "loss": 1.2287,
+      "step": 12425
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.354801501364514e-06,
+      "loss": 0.9693,
+      "step": 12426
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.353299232737242e-06,
+      "loss": 1.107,
+      "step": 12427
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.351797028337611e-06,
+      "loss": 1.1301,
+      "step": 12428
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.35029488820208e-06,
+      "loss": 1.0015,
+      "step": 12429
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.348792812367097e-06,
+      "loss": 1.1387,
+      "step": 12430
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.347290800869118e-06,
+      "loss": 1.1928,
+      "step": 12431
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.345788853744589e-06,
+      "loss": 1.1109,
+      "step": 12432
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.344286971029959e-06,
+      "loss": 1.0471,
+      "step": 12433
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.342785152761673e-06,
+      "loss": 1.1468,
+      "step": 12434
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.341283398976178e-06,
+      "loss": 1.1689,
+      "step": 12435
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3397817097099145e-06,
+      "loss": 1.1135,
+      "step": 12436
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.338280084999328e-06,
+      "loss": 1.1476,
+      "step": 12437
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3367785248808535e-06,
+      "loss": 0.9784,
+      "step": 12438
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.335277029390934e-06,
+      "loss": 1.1804,
+      "step": 12439
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.333775598566002e-06,
+      "loss": 1.1095,
+      "step": 12440
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.332274232442496e-06,
+      "loss": 0.9838,
+      "step": 12441
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.330772931056849e-06,
+      "loss": 0.9212,
+      "step": 12442
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.329271694445492e-06,
+      "loss": 1.0953,
+      "step": 12443
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.327770522644859e-06,
+      "loss": 1.2056,
+      "step": 12444
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.326269415691375e-06,
+      "loss": 1.123,
+      "step": 12445
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.324768373621471e-06,
+      "loss": 1.2172,
+      "step": 12446
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.323267396471568e-06,
+      "loss": 1.1356,
+      "step": 12447
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.321766484278097e-06,
+      "loss": 0.9811,
+      "step": 12448
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.320265637077473e-06,
+      "loss": 1.2129,
+      "step": 12449
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.318764854906124e-06,
+      "loss": 1.1681,
+      "step": 12450
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.317264137800466e-06,
+      "loss": 1.1461,
+      "step": 12451
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.315763485796918e-06,
+      "loss": 1.0775,
+      "step": 12452
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.314262898931894e-06,
+      "loss": 1.0573,
+      "step": 12453
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.312762377241815e-06,
+      "loss": 1.2641,
+      "step": 12454
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.311261920763087e-06,
+      "loss": 1.1723,
+      "step": 12455
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.309761529532127e-06,
+      "loss": 1.2131,
+      "step": 12456
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.30826120358534e-06,
+      "loss": 1.0627,
+      "step": 12457
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.306760942959141e-06,
+      "loss": 1.1843,
+      "step": 12458
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3052607476899305e-06,
+      "loss": 0.9827,
+      "step": 12459
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.30376061781412e-06,
+      "loss": 1.0591,
+      "step": 12460
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.302260553368106e-06,
+      "loss": 1.1243,
+      "step": 12461
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.300760554388297e-06,
+      "loss": 1.1438,
+      "step": 12462
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.29926062091109e-06,
+      "loss": 0.9864,
+      "step": 12463
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2977607529728865e-06,
+      "loss": 1.25,
+      "step": 12464
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.296260950610081e-06,
+      "loss": 1.0919,
+      "step": 12465
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2947612138590725e-06,
+      "loss": 1.103,
+      "step": 12466
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.293261542756251e-06,
+      "loss": 1.0043,
+      "step": 12467
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.291761937338016e-06,
+      "loss": 1.1014,
+      "step": 12468
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.290262397640749e-06,
+      "loss": 0.964,
+      "step": 12469
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2887629237008485e-06,
+      "loss": 1.059,
+      "step": 12470
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.287263515554697e-06,
+      "loss": 0.9751,
+      "step": 12471
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.285764173238682e-06,
+      "loss": 1.0465,
+      "step": 12472
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.284264896789187e-06,
+      "loss": 1.1814,
+      "step": 12473
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.282765686242599e-06,
+      "loss": 1.0263,
+      "step": 12474
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.281266541635293e-06,
+      "loss": 1.2004,
+      "step": 12475
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.279767463003656e-06,
+      "loss": 1.1404,
+      "step": 12476
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.278268450384061e-06,
+      "loss": 1.2371,
+      "step": 12477
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.276769503812889e-06,
+      "loss": 1.0513,
+      "step": 12478
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.275270623326508e-06,
+      "loss": 1.1739,
+      "step": 12479
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2737718089612985e-06,
+      "loss": 1.1007,
+      "step": 12480
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.272273060753629e-06,
+      "loss": 1.1765,
+      "step": 12481
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.270774378739871e-06,
+      "loss": 1.1633,
+      "step": 12482
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.269275762956392e-06,
+      "loss": 1.2424,
+      "step": 12483
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.26777721343956e-06,
+      "loss": 1.0247,
+      "step": 12484
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.266278730225738e-06,
+      "loss": 1.069,
+      "step": 12485
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.264780313351294e-06,
+      "loss": 1.1359,
+      "step": 12486
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.263281962852586e-06,
+      "loss": 1.1724,
+      "step": 12487
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.261783678765977e-06,
+      "loss": 1.1311,
+      "step": 12488
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.260285461127825e-06,
+      "loss": 1.0275,
+      "step": 12489
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.258787309974488e-06,
+      "loss": 1.1726,
+      "step": 12490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.257289225342323e-06,
+      "loss": 1.1669,
+      "step": 12491
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.255791207267679e-06,
+      "loss": 1.1988,
+      "step": 12492
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.254293255786915e-06,
+      "loss": 1.1603,
+      "step": 12493
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2527953709363765e-06,
+      "loss": 1.1787,
+      "step": 12494
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.251297552752418e-06,
+      "loss": 1.137,
+      "step": 12495
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.249799801271381e-06,
+      "loss": 1.0706,
+      "step": 12496
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.248302116529618e-06,
+      "loss": 1.1602,
+      "step": 12497
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2468044985634665e-06,
+      "loss": 1.2534,
+      "step": 12498
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.245306947409276e-06,
+      "loss": 1.0907,
+      "step": 12499
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.243809463103381e-06,
+      "loss": 1.0914,
+      "step": 12500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.242312045682127e-06,
+      "loss": 1.2153,
+      "step": 12501
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.24081469518185e-06,
+      "loss": 1.2597,
+      "step": 12502
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.239317411638886e-06,
+      "loss": 1.0911,
+      "step": 12503
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2378201950895696e-06,
+      "loss": 1.1139,
+      "step": 12504
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.236323045570233e-06,
+      "loss": 1.1608,
+      "step": 12505
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.234825963117209e-06,
+      "loss": 1.2052,
+      "step": 12506
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2333289477668265e-06,
+      "loss": 1.1674,
+      "step": 12507
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.231831999555415e-06,
+      "loss": 1.1539,
+      "step": 12508
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.230335118519298e-06,
+      "loss": 1.1242,
+      "step": 12509
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.228838304694805e-06,
+      "loss": 1.247,
+      "step": 12510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.227341558118254e-06,
+      "loss": 1.0868,
+      "step": 12511
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.225844878825971e-06,
+      "loss": 0.9768,
+      "step": 12512
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.224348266854273e-06,
+      "loss": 1.2487,
+      "step": 12513
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2228517222394815e-06,
+      "loss": 1.0812,
+      "step": 12514
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.221355245017909e-06,
+      "loss": 1.1765,
+      "step": 12515
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.219858835225876e-06,
+      "loss": 1.1347,
+      "step": 12516
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2183624928996905e-06,
+      "loss": 1.1112,
+      "step": 12517
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.216866218075668e-06,
+      "loss": 1.0188,
+      "step": 12518
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.215370010790117e-06,
+      "loss": 1.1445,
+      "step": 12519
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.213873871079347e-06,
+      "loss": 1.088,
+      "step": 12520
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.212377798979664e-06,
+      "loss": 1.082,
+      "step": 12521
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.210881794527373e-06,
+      "loss": 0.9728,
+      "step": 12522
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2093858577587795e-06,
+      "loss": 1.1827,
+      "step": 12523
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.207889988710185e-06,
+      "loss": 1.217,
+      "step": 12524
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.206394187417887e-06,
+      "loss": 1.1788,
+      "step": 12525
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2048984539181876e-06,
+      "loss": 1.0986,
+      "step": 12526
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.203402788247383e-06,
+      "loss": 1.0506,
+      "step": 12527
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.201907190441766e-06,
+      "loss": 1.0466,
+      "step": 12528
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2004116605376364e-06,
+      "loss": 0.9944,
+      "step": 12529
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.198916198571278e-06,
+      "loss": 1.104,
+      "step": 12530
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.19742080457899e-06,
+      "loss": 0.9867,
+      "step": 12531
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.195925478597054e-06,
+      "loss": 1.0984,
+      "step": 12532
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.194430220661762e-06,
+      "loss": 1.1694,
+      "step": 12533
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.192935030809396e-06,
+      "loss": 1.2381,
+      "step": 12534
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.191439909076243e-06,
+      "loss": 1.0815,
+      "step": 12535
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.189944855498583e-06,
+      "loss": 1.1873,
+      "step": 12536
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.188449870112699e-06,
+      "loss": 1.0425,
+      "step": 12537
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.186954952954865e-06,
+      "loss": 1.1445,
+      "step": 12538
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.185460104061364e-06,
+      "loss": 1.2005,
+      "step": 12539
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.183965323468468e-06,
+      "loss": 1.2904,
+      "step": 12540
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.182470611212453e-06,
+      "loss": 1.1011,
+      "step": 12541
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.180975967329589e-06,
+      "loss": 1.0757,
+      "step": 12542
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.179481391856149e-06,
+      "loss": 1.1872,
+      "step": 12543
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.177986884828398e-06,
+      "loss": 1.1182,
+      "step": 12544
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.176492446282611e-06,
+      "loss": 1.0113,
+      "step": 12545
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1749980762550434e-06,
+      "loss": 1.2043,
+      "step": 12546
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.173503774781967e-06,
+      "loss": 1.1614,
+      "step": 12547
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.17200954189964e-06,
+      "loss": 1.1601,
+      "step": 12548
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.170515377644327e-06,
+      "loss": 1.1334,
+      "step": 12549
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.169021282052283e-06,
+      "loss": 1.1876,
+      "step": 12550
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.167527255159769e-06,
+      "loss": 1.0194,
+      "step": 12551
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.166033297003035e-06,
+      "loss": 1.0861,
+      "step": 12552
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.164539407618342e-06,
+      "loss": 1.035,
+      "step": 12553
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.163045587041936e-06,
+      "loss": 1.0372,
+      "step": 12554
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1615518353100726e-06,
+      "loss": 1.1854,
+      "step": 12555
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.160058152458996e-06,
+      "loss": 1.2154,
+      "step": 12556
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.158564538524959e-06,
+      "loss": 1.1071,
+      "step": 12557
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1570709935442015e-06,
+      "loss": 1.1091,
+      "step": 12558
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.155577517552973e-06,
+      "loss": 1.2575,
+      "step": 12559
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1540841105875105e-06,
+      "loss": 1.0838,
+      "step": 12560
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1525907726840595e-06,
+      "loss": 1.0983,
+      "step": 12561
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.151097503878854e-06,
+      "loss": 1.0994,
+      "step": 12562
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.149604304208135e-06,
+      "loss": 1.1225,
+      "step": 12563
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1481111737081365e-06,
+      "loss": 1.0703,
+      "step": 12564
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1466181124150935e-06,
+      "loss": 1.1236,
+      "step": 12565
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.145125120365237e-06,
+      "loss": 1.1018,
+      "step": 12566
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.143632197594799e-06,
+      "loss": 1.1381,
+      "step": 12567
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1421393441400046e-06,
+      "loss": 1.1398,
+      "step": 12568
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.140646560037087e-06,
+      "loss": 1.0466,
+      "step": 12569
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.139153845322264e-06,
+      "loss": 1.0059,
+      "step": 12570
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.137661200031769e-06,
+      "loss": 1.1771,
+      "step": 12571
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.136168624201815e-06,
+      "loss": 1.2147,
+      "step": 12572
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1346761178686265e-06,
+      "loss": 1.192,
+      "step": 12573
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.133183681068423e-06,
+      "loss": 1.0355,
+      "step": 12574
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.131691313837419e-06,
+      "loss": 1.1007,
+      "step": 12575
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.130199016211834e-06,
+      "loss": 1.3258,
+      "step": 12576
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.128706788227877e-06,
+      "loss": 1.1474,
+      "step": 12577
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.127214629921765e-06,
+      "loss": 1.332,
+      "step": 12578
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.125722541329702e-06,
+      "loss": 1.1277,
+      "step": 12579
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.124230522487903e-06,
+      "loss": 1.2339,
+      "step": 12580
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.122738573432568e-06,
+      "loss": 1.0916,
+      "step": 12581
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.121246694199911e-06,
+      "loss": 1.0631,
+      "step": 12582
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.119754884826124e-06,
+      "loss": 1.2135,
+      "step": 12583
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.118263145347418e-06,
+      "loss": 1.088,
+      "step": 12584
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.116771475799993e-06,
+      "loss": 1.1287,
+      "step": 12585
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.115279876220044e-06,
+      "loss": 1.1143,
+      "step": 12586
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.11378834664377e-06,
+      "loss": 1.0743,
+      "step": 12587
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.112296887107361e-06,
+      "loss": 1.1217,
+      "step": 12588
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1108054976470185e-06,
+      "loss": 1.2119,
+      "step": 12589
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1093141782989275e-06,
+      "loss": 1.0496,
+      "step": 12590
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.107822929099282e-06,
+      "loss": 1.2408,
+      "step": 12591
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.106331750084266e-06,
+      "loss": 1.104,
+      "step": 12592
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1048406412900695e-06,
+      "loss": 1.1466,
+      "step": 12593
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1033496027528755e-06,
+      "loss": 1.197,
+      "step": 12594
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.101858634508869e-06,
+      "loss": 1.1306,
+      "step": 12595
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.100367736594229e-06,
+      "loss": 0.9973,
+      "step": 12596
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.098876909045139e-06,
+      "loss": 0.935,
+      "step": 12597
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.097386151897773e-06,
+      "loss": 0.9151,
+      "step": 12598
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.095895465188309e-06,
+      "loss": 1.1983,
+      "step": 12599
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0944048489529205e-06,
+      "loss": 1.2573,
+      "step": 12600
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.092914303227783e-06,
+      "loss": 1.0871,
+      "step": 12601
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.091423828049064e-06,
+      "loss": 1.0105,
+      "step": 12602
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.089933423452937e-06,
+      "loss": 0.9867,
+      "step": 12603
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0884430894755655e-06,
+      "loss": 1.019,
+      "step": 12604
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.08695282615312e-06,
+      "loss": 1.042,
+      "step": 12605
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.08546263352176e-06,
+      "loss": 1.04,
+      "step": 12606
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0839725116176515e-06,
+      "loss": 1.0967,
+      "step": 12607
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.082482460476954e-06,
+      "loss": 0.9497,
+      "step": 12608
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.080992480135829e-06,
+      "loss": 1.1658,
+      "step": 12609
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.07950257063043e-06,
+      "loss": 1.1473,
+      "step": 12610
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.078012731996915e-06,
+      "loss": 1.0869,
+      "step": 12611
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0765229642714374e-06,
+      "loss": 1.1497,
+      "step": 12612
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.07503326749015e-06,
+      "loss": 1.0864,
+      "step": 12613
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.073543641689205e-06,
+      "loss": 1.0569,
+      "step": 12614
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.072054086904747e-06,
+      "loss": 0.9823,
+      "step": 12615
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.070564603172927e-06,
+      "loss": 1.1504,
+      "step": 12616
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.069075190529888e-06,
+      "loss": 1.1402,
+      "step": 12617
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.067585849011775e-06,
+      "loss": 1.0907,
+      "step": 12618
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.066096578654729e-06,
+      "loss": 1.0554,
+      "step": 12619
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.064607379494892e-06,
+      "loss": 1.1709,
+      "step": 12620
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0631182515684e-06,
+      "loss": 1.1298,
+      "step": 12621
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.061629194911392e-06,
+      "loss": 1.0596,
+      "step": 12622
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0601402095600015e-06,
+      "loss": 1.1702,
+      "step": 12623
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.058651295550363e-06,
+      "loss": 1.2096,
+      "step": 12624
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.057162452918607e-06,
+      "loss": 1.1961,
+      "step": 12625
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.055673681700863e-06,
+      "loss": 1.2121,
+      "step": 12626
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.05418498193326e-06,
+      "loss": 1.1535,
+      "step": 12627
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.052696353651926e-06,
+      "loss": 1.0184,
+      "step": 12628
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0512077968929805e-06,
+      "loss": 1.0414,
+      "step": 12629
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.049719311692552e-06,
+      "loss": 1.1301,
+      "step": 12630
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.048230898086759e-06,
+      "loss": 1.1052,
+      "step": 12631
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.046742556111721e-06,
+      "loss": 1.0957,
+      "step": 12632
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.045254285803556e-06,
+      "loss": 1.231,
+      "step": 12633
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.043766087198381e-06,
+      "loss": 1.2157,
+      "step": 12634
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.042277960332305e-06,
+      "loss": 1.2018,
+      "step": 12635
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.04078990524145e-06,
+      "loss": 1.087,
+      "step": 12636
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.039301921961916e-06,
+      "loss": 1.3475,
+      "step": 12637
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.037814010529822e-06,
+      "loss": 1.117,
+      "step": 12638
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.036326170981267e-06,
+      "loss": 1.1866,
+      "step": 12639
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.034838403352362e-06,
+      "loss": 1.1158,
+      "step": 12640
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.033350707679207e-06,
+      "loss": 1.1778,
+      "step": 12641
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0318630839979086e-06,
+      "loss": 0.9578,
+      "step": 12642
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.030375532344562e-06,
+      "loss": 1.1343,
+      "step": 12643
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.02888805275527e-06,
+      "loss": 1.2033,
+      "step": 12644
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.027400645266123e-06,
+      "loss": 1.1317,
+      "step": 12645
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.025913309913223e-06,
+      "loss": 1.049,
+      "step": 12646
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.02442604673266e-06,
+      "loss": 1.1853,
+      "step": 12647
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.022938855760528e-06,
+      "loss": 1.0517,
+      "step": 12648
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.021451737032912e-06,
+      "loss": 0.9926,
+      "step": 12649
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0199646905859055e-06,
+      "loss": 1.0977,
+      "step": 12650
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.018477716455588e-06,
+      "loss": 1.1174,
+      "step": 12651
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.016990814678052e-06,
+      "loss": 1.0814,
+      "step": 12652
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.015503985289374e-06,
+      "loss": 1.1783,
+      "step": 12653
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.014017228325638e-06,
+      "loss": 0.997,
+      "step": 12654
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.012530543822921e-06,
+      "loss": 1.1137,
+      "step": 12655
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.011043931817304e-06,
+      "loss": 0.9851,
+      "step": 12656
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.009557392344859e-06,
+      "loss": 1.1841,
+      "step": 12657
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.008070925441663e-06,
+      "loss": 1.3409,
+      "step": 12658
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.006584531143784e-06,
+      "loss": 1.1803,
+      "step": 12659
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.005098209487295e-06,
+      "loss": 1.0425,
+      "step": 12660
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.003611960508268e-06,
+      "loss": 1.2059,
+      "step": 12661
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0021257842427645e-06,
+      "loss": 1.1368,
+      "step": 12662
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.000639680726854e-06,
+      "loss": 1.0172,
+      "step": 12663
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.999153649996595e-06,
+      "loss": 1.2633,
+      "step": 12664
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.997667692088053e-06,
+      "loss": 1.0015,
+      "step": 12665
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9961818070372845e-06,
+      "loss": 1.0684,
+      "step": 12666
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.994695994880349e-06,
+      "loss": 1.2129,
+      "step": 12667
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.993210255653309e-06,
+      "loss": 1.0583,
+      "step": 12668
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.991724589392208e-06,
+      "loss": 1.0359,
+      "step": 12669
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.990238996133109e-06,
+      "loss": 1.074,
+      "step": 12670
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.988753475912054e-06,
+      "loss": 1.0485,
+      "step": 12671
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9872680287651e-06,
+      "loss": 1.0741,
+      "step": 12672
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.985782654728288e-06,
+      "loss": 1.2428,
+      "step": 12673
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9842973538376675e-06,
+      "loss": 1.0859,
+      "step": 12674
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.982812126129281e-06,
+      "loss": 1.1666,
+      "step": 12675
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.981326971639173e-06,
+      "loss": 1.1155,
+      "step": 12676
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.979841890403378e-06,
+      "loss": 1.1825,
+      "step": 12677
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.978356882457942e-06,
+      "loss": 1.0994,
+      "step": 12678
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9768719478388956e-06,
+      "loss": 1.1176,
+      "step": 12679
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.975387086582279e-06,
+      "loss": 1.0838,
+      "step": 12680
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.97390229872412e-06,
+      "loss": 1.1977,
+      "step": 12681
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.972417584300455e-06,
+      "loss": 1.1926,
+      "step": 12682
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.970932943347308e-06,
+      "loss": 0.9338,
+      "step": 12683
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.969448375900715e-06,
+      "loss": 1.0794,
+      "step": 12684
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.967963881996693e-06,
+      "loss": 1.1467,
+      "step": 12685
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.966479461671274e-06,
+      "loss": 1.0653,
+      "step": 12686
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.964995114960475e-06,
+      "loss": 1.0601,
+      "step": 12687
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.963510841900321e-06,
+      "loss": 1.0191,
+      "step": 12688
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.962026642526827e-06,
+      "loss": 1.0289,
+      "step": 12689
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.960542516876014e-06,
+      "loss": 0.9522,
+      "step": 12690
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.959058464983892e-06,
+      "loss": 1.0994,
+      "step": 12691
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.957574486886483e-06,
+      "loss": 1.0092,
+      "step": 12692
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9560905826197885e-06,
+      "loss": 1.0505,
+      "step": 12693
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.954606752219828e-06,
+      "loss": 1.042,
+      "step": 12694
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.953122995722601e-06,
+      "loss": 1.1317,
+      "step": 12695
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.951639313164122e-06,
+      "loss": 1.1695,
+      "step": 12696
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.950155704580391e-06,
+      "loss": 1.1853,
+      "step": 12697
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.94867217000741e-06,
+      "loss": 1.1098,
+      "step": 12698
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.947188709481185e-06,
+      "loss": 1.1315,
+      "step": 12699
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.945705323037709e-06,
+      "loss": 1.0627,
+      "step": 12700
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9442220107129844e-06,
+      "loss": 1.1529,
+      "step": 12701
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.942738772543004e-06,
+      "loss": 1.1668,
+      "step": 12702
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.941255608563764e-06,
+      "loss": 1.2159,
+      "step": 12703
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.939772518811252e-06,
+      "loss": 1.225,
+      "step": 12704
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.938289503321465e-06,
+      "loss": 0.979,
+      "step": 12705
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.936806562130385e-06,
+      "loss": 1.0703,
+      "step": 12706
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.935323695274002e-06,
+      "loss": 1.1729,
+      "step": 12707
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.933840902788298e-06,
+      "loss": 1.1274,
+      "step": 12708
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.932358184709263e-06,
+      "loss": 0.9628,
+      "step": 12709
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.930875541072868e-06,
+      "loss": 1.1544,
+      "step": 12710
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.929392971915101e-06,
+      "loss": 1.1409,
+      "step": 12711
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.927910477271933e-06,
+      "loss": 0.9937,
+      "step": 12712
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.926428057179346e-06,
+      "loss": 1.1755,
+      "step": 12713
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.92494571167331e-06,
+      "loss": 1.1801,
+      "step": 12714
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.923463440789799e-06,
+      "loss": 1.1591,
+      "step": 12715
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9219812445647805e-06,
+      "loss": 1.2593,
+      "step": 12716
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.920499123034228e-06,
+      "loss": 1.0608,
+      "step": 12717
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.919017076234103e-06,
+      "loss": 1.0632,
+      "step": 12718
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.917535104200372e-06,
+      "loss": 1.0571,
+      "step": 12719
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.916053206969e-06,
+      "loss": 1.1823,
+      "step": 12720
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.914571384575949e-06,
+      "loss": 1.0691,
+      "step": 12721
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.913089637057173e-06,
+      "loss": 0.9735,
+      "step": 12722
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.911607964448637e-06,
+      "loss": 1.143,
+      "step": 12723
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.910126366786289e-06,
+      "loss": 1.1537,
+      "step": 12724
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.908644844106091e-06,
+      "loss": 0.9808,
+      "step": 12725
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.907163396443988e-06,
+      "loss": 1.1864,
+      "step": 12726
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9056820238359355e-06,
+      "loss": 0.8999,
+      "step": 12727
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.90420072631788e-06,
+      "loss": 1.0205,
+      "step": 12728
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.902719503925769e-06,
+      "loss": 1.0917,
+      "step": 12729
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.901238356695545e-06,
+      "loss": 1.1948,
+      "step": 12730
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.899757284663156e-06,
+      "loss": 1.0543,
+      "step": 12731
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.898276287864536e-06,
+      "loss": 1.1357,
+      "step": 12732
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.896795366335633e-06,
+      "loss": 1.1297,
+      "step": 12733
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8953145201123776e-06,
+      "loss": 1.2717,
+      "step": 12734
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.893833749230709e-06,
+      "loss": 1.1251,
+      "step": 12735
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.89235305372656e-06,
+      "loss": 1.1556,
+      "step": 12736
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.890872433635863e-06,
+      "loss": 1.1383,
+      "step": 12737
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.889391888994548e-06,
+      "loss": 1.1589,
+      "step": 12738
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.887911419838546e-06,
+      "loss": 1.0849,
+      "step": 12739
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8864310262037795e-06,
+      "loss": 1.1283,
+      "step": 12740
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.884950708126177e-06,
+      "loss": 1.1698,
+      "step": 12741
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.883470465641657e-06,
+      "loss": 1.1407,
+      "step": 12742
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.881990298786148e-06,
+      "loss": 1.0919,
+      "step": 12743
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.88051020759556e-06,
+      "loss": 1.1364,
+      "step": 12744
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8790301921058155e-06,
+      "loss": 1.0457,
+      "step": 12745
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.877550252352834e-06,
+      "loss": 1.0805,
+      "step": 12746
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.87607038837252e-06,
+      "loss": 1.1871,
+      "step": 12747
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.874590600200794e-06,
+      "loss": 1.1065,
+      "step": 12748
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.873110887873562e-06,
+      "loss": 1.0845,
+      "step": 12749
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.871631251426729e-06,
+      "loss": 0.9378,
+      "step": 12750
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.87015169089621e-06,
+      "loss": 1.1091,
+      "step": 12751
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.868672206317903e-06,
+      "loss": 1.1234,
+      "step": 12752
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.867192797727714e-06,
+      "loss": 1.0175,
+      "step": 12753
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.865713465161541e-06,
+      "loss": 1.1817,
+      "step": 12754
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.864234208655287e-06,
+      "loss": 0.9572,
+      "step": 12755
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.862755028244844e-06,
+      "loss": 1.0316,
+      "step": 12756
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.861275923966113e-06,
+      "loss": 1.102,
+      "step": 12757
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.859796895854982e-06,
+      "loss": 1.0664,
+      "step": 12758
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.858317943947348e-06,
+      "loss": 1.0503,
+      "step": 12759
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.856839068279096e-06,
+      "loss": 1.1803,
+      "step": 12760
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8553602688861175e-06,
+      "loss": 1.0454,
+      "step": 12761
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.853881545804295e-06,
+      "loss": 1.1418,
+      "step": 12762
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.852402899069517e-06,
+      "loss": 1.0628,
+      "step": 12763
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.850924328717663e-06,
+      "loss": 1.1255,
+      "step": 12764
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.849445834784616e-06,
+      "loss": 1.1335,
+      "step": 12765
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.847967417306251e-06,
+      "loss": 1.0898,
+      "step": 12766
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.846489076318449e-06,
+      "loss": 1.1959,
+      "step": 12767
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8450108118570805e-06,
+      "loss": 0.989,
+      "step": 12768
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.843532623958023e-06,
+      "loss": 0.9621,
+      "step": 12769
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.842054512657146e-06,
+      "loss": 1.0401,
+      "step": 12770
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.840576477990318e-06,
+      "loss": 1.0947,
+      "step": 12771
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.839098519993407e-06,
+      "loss": 0.965,
+      "step": 12772
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.83762063870228e-06,
+      "loss": 1.0996,
+      "step": 12773
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8361428341527995e-06,
+      "loss": 1.0899,
+      "step": 12774
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.834665106380829e-06,
+      "loss": 1.0679,
+      "step": 12775
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.833187455422226e-06,
+      "loss": 1.0976,
+      "step": 12776
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.83170988131285e-06,
+      "loss": 1.1979,
+      "step": 12777
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.830232384088557e-06,
+      "loss": 1.0784,
+      "step": 12778
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8287549637852046e-06,
+      "loss": 1.1891,
+      "step": 12779
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.827277620438641e-06,
+      "loss": 1.1954,
+      "step": 12780
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8258003540847196e-06,
+      "loss": 1.2691,
+      "step": 12781
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.824323164759288e-06,
+      "loss": 1.0838,
+      "step": 12782
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.822846052498195e-06,
+      "loss": 1.1492,
+      "step": 12783
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.821369017337283e-06,
+      "loss": 1.1453,
+      "step": 12784
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8198920593123966e-06,
+      "loss": 1.1238,
+      "step": 12785
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.81841517845938e-06,
+      "loss": 0.9891,
+      "step": 12786
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.8169383748140685e-06,
+      "loss": 1.1825,
+      "step": 12787
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.815461648412303e-06,
+      "loss": 1.2241,
+      "step": 12788
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.8139849992899155e-06,
+      "loss": 0.9348,
+      "step": 12789
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.812508427482745e-06,
+      "loss": 1.0715,
+      "step": 12790
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.811031933026619e-06,
+      "loss": 1.1505,
+      "step": 12791
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.809555515957371e-06,
+      "loss": 0.982,
+      "step": 12792
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.808079176310826e-06,
+      "loss": 0.9715,
+      "step": 12793
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.806602914122815e-06,
+      "loss": 1.1419,
+      "step": 12794
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.805126729429158e-06,
+      "loss": 1.1421,
+      "step": 12795
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.803650622265681e-06,
+      "loss": 1.2688,
+      "step": 12796
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.802174592668201e-06,
+      "loss": 1.2218,
+      "step": 12797
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.800698640672541e-06,
+      "loss": 1.0544,
+      "step": 12798
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.799222766314515e-06,
+      "loss": 1.0576,
+      "step": 12799
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.797746969629941e-06,
+      "loss": 1.0768,
+      "step": 12800
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.796271250654628e-06,
+      "loss": 1.0432,
+      "step": 12801
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.794795609424392e-06,
+      "loss": 1.0754,
+      "step": 12802
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.793320045975038e-06,
+      "loss": 1.0338,
+      "step": 12803
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.79184456034238e-06,
+      "loss": 1.1098,
+      "step": 12804
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7903691525622166e-06,
+      "loss": 1.1257,
+      "step": 12805
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7888938226703564e-06,
+      "loss": 1.1071,
+      "step": 12806
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7874185707025976e-06,
+      "loss": 0.9918,
+      "step": 12807
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.785943396694743e-06,
+      "loss": 1.1534,
+      "step": 12808
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.78446830068259e-06,
+      "loss": 1.1808,
+      "step": 12809
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.782993282701936e-06,
+      "loss": 1.1238,
+      "step": 12810
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7815183427885725e-06,
+      "loss": 1.1813,
+      "step": 12811
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.780043480978296e-06,
+      "loss": 0.9505,
+      "step": 12812
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.778568697306891e-06,
+      "loss": 0.989,
+      "step": 12813
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.777093991810153e-06,
+      "loss": 1.1524,
+      "step": 12814
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.775619364523863e-06,
+      "loss": 1.0553,
+      "step": 12815
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7741448154838105e-06,
+      "loss": 0.9225,
+      "step": 12816
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7726703447257755e-06,
+      "loss": 1.1725,
+      "step": 12817
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.771195952285541e-06,
+      "loss": 1.1577,
+      "step": 12818
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.769721638198883e-06,
+      "loss": 1.1362,
+      "step": 12819
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.768247402501585e-06,
+      "loss": 1.1087,
+      "step": 12820
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.766773245229413e-06,
+      "loss": 1.1289,
+      "step": 12821
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.76529916641815e-06,
+      "loss": 1.1746,
+      "step": 12822
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.76382516610356e-06,
+      "loss": 1.2013,
+      "step": 12823
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.762351244321419e-06,
+      "loss": 1.0629,
+      "step": 12824
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.76087740110749e-06,
+      "loss": 1.0458,
+      "step": 12825
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.759403636497541e-06,
+      "loss": 1.0399,
+      "step": 12826
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.757929950527335e-06,
+      "loss": 1.0208,
+      "step": 12827
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.756456343232636e-06,
+      "loss": 1.1668,
+      "step": 12828
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7549828146492e-06,
+      "loss": 1.0846,
+      "step": 12829
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.753509364812789e-06,
+      "loss": 1.227,
+      "step": 12830
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7520359937591595e-06,
+      "loss": 1.045,
+      "step": 12831
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.750562701524063e-06,
+      "loss": 1.1944,
+      "step": 12832
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.74908948814325e-06,
+      "loss": 0.9233,
+      "step": 12833
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7476163536524796e-06,
+      "loss": 1.0328,
+      "step": 12834
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.746143298087494e-06,
+      "loss": 1.1113,
+      "step": 12835
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.744670321484044e-06,
+      "loss": 1.112,
+      "step": 12836
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.743197423877867e-06,
+      "loss": 0.9927,
+      "step": 12837
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.741724605304714e-06,
+      "loss": 1.1832,
+      "step": 12838
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.740251865800321e-06,
+      "loss": 1.018,
+      "step": 12839
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.738779205400429e-06,
+      "loss": 1.1106,
+      "step": 12840
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.737306624140776e-06,
+      "loss": 0.8987,
+      "step": 12841
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.735834122057096e-06,
+      "loss": 1.0515,
+      "step": 12842
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.734361699185119e-06,
+      "loss": 1.0703,
+      "step": 12843
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.732889355560583e-06,
+      "loss": 1.1587,
+      "step": 12844
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.731417091219212e-06,
+      "loss": 1.2246,
+      "step": 12845
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.729944906196739e-06,
+      "loss": 1.179,
+      "step": 12846
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.728472800528883e-06,
+      "loss": 1.0578,
+      "step": 12847
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.727000774251373e-06,
+      "loss": 1.0855,
+      "step": 12848
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.725528827399926e-06,
+      "loss": 1.1423,
+      "step": 12849
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.724056960010268e-06,
+      "loss": 0.9925,
+      "step": 12850
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.72258517211811e-06,
+      "loss": 1.13,
+      "step": 12851
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.721113463759174e-06,
+      "loss": 0.9543,
+      "step": 12852
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.719641834969168e-06,
+      "loss": 1.1315,
+      "step": 12853
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.718170285783811e-06,
+      "loss": 1.125,
+      "step": 12854
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.716698816238807e-06,
+      "loss": 1.0645,
+      "step": 12855
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.71522742636987e-06,
+      "loss": 1.11,
+      "step": 12856
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7137561162127e-06,
+      "loss": 0.9593,
+      "step": 12857
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.712284885803007e-06,
+      "loss": 1.0463,
+      "step": 12858
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.710813735176489e-06,
+      "loss": 1.2041,
+      "step": 12859
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.70934266436885e-06,
+      "loss": 1.0281,
+      "step": 12860
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.707871673415785e-06,
+      "loss": 1.0541,
+      "step": 12861
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.706400762352996e-06,
+      "loss": 1.1691,
+      "step": 12862
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.704929931216171e-06,
+      "loss": 1.1355,
+      "step": 12863
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.70345918004101e-06,
+      "loss": 0.952,
+      "step": 12864
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.701988508863195e-06,
+      "loss": 1.0476,
+      "step": 12865
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.700517917718423e-06,
+      "loss": 1.1013,
+      "step": 12866
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.699047406642377e-06,
+      "loss": 1.1518,
+      "step": 12867
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6975769756707434e-06,
+      "loss": 0.956,
+      "step": 12868
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.696106624839203e-06,
+      "loss": 1.1307,
+      "step": 12869
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.694636354183439e-06,
+      "loss": 1.2374,
+      "step": 12870
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.693166163739131e-06,
+      "loss": 1.1096,
+      "step": 12871
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.691696053541954e-06,
+      "loss": 1.1859,
+      "step": 12872
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.690226023627585e-06,
+      "loss": 1.2217,
+      "step": 12873
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.688756074031695e-06,
+      "loss": 1.1643,
+      "step": 12874
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.68728620478996e-06,
+      "loss": 0.9068,
+      "step": 12875
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6858164159380445e-06,
+      "loss": 1.0257,
+      "step": 12876
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6843467075116196e-06,
+      "loss": 1.2508,
+      "step": 12877
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.682877079546346e-06,
+      "loss": 1.0423,
+      "step": 12878
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.681407532077895e-06,
+      "loss": 1.0252,
+      "step": 12879
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.679938065141918e-06,
+      "loss": 1.2523,
+      "step": 12880
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6784686787740845e-06,
+      "loss": 1.1013,
+      "step": 12881
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.676999373010047e-06,
+      "loss": 1.1226,
+      "step": 12882
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.675530147885462e-06,
+      "loss": 1.2141,
+      "step": 12883
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.674061003435982e-06,
+      "loss": 1.1693,
+      "step": 12884
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.672591939697261e-06,
+      "loss": 1.0901,
+      "step": 12885
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6711229567049485e-06,
+      "loss": 1.1906,
+      "step": 12886
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.669654054494693e-06,
+      "loss": 1.2203,
+      "step": 12887
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.668185233102137e-06,
+      "loss": 0.9694,
+      "step": 12888
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.666716492562929e-06,
+      "loss": 1.1264,
+      "step": 12889
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.665247832912707e-06,
+      "loss": 0.8932,
+      "step": 12890
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.663779254187115e-06,
+      "loss": 1.1134,
+      "step": 12891
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.662310756421787e-06,
+      "loss": 1.1315,
+      "step": 12892
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.660842339652362e-06,
+      "loss": 1.0856,
+      "step": 12893
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.659374003914474e-06,
+      "loss": 1.0248,
+      "step": 12894
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.657905749243754e-06,
+      "loss": 1.0672,
+      "step": 12895
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6564375756758314e-06,
+      "loss": 1.092,
+      "step": 12896
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6549694832463385e-06,
+      "loss": 1.1754,
+      "step": 12897
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.653501471990896e-06,
+      "loss": 1.2342,
+      "step": 12898
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.652033541945134e-06,
+      "loss": 1.1171,
+      "step": 12899
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.650565693144669e-06,
+      "loss": 1.1858,
+      "step": 12900
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.649097925625128e-06,
+      "loss": 1.2958,
+      "step": 12901
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.647630239422121e-06,
+      "loss": 1.1346,
+      "step": 12902
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6461626345712735e-06,
+      "loss": 1.0654,
+      "step": 12903
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.644695111108192e-06,
+      "loss": 1.044,
+      "step": 12904
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.643227669068495e-06,
+      "loss": 1.0738,
+      "step": 12905
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6417603084877894e-06,
+      "loss": 1.1771,
+      "step": 12906
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.640293029401686e-06,
+      "loss": 1.2331,
+      "step": 12907
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.638825831845788e-06,
+      "loss": 1.139,
+      "step": 12908
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.637358715855704e-06,
+      "loss": 1.1486,
+      "step": 12909
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.635891681467034e-06,
+      "loss": 1.1194,
+      "step": 12910
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.634424728715381e-06,
+      "loss": 0.9789,
+      "step": 12911
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.632957857636339e-06,
+      "loss": 1.1912,
+      "step": 12912
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.631491068265511e-06,
+      "loss": 1.1945,
+      "step": 12913
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.630024360638486e-06,
+      "loss": 1.0751,
+      "step": 12914
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.628557734790862e-06,
+      "loss": 1.1615,
+      "step": 12915
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.627091190758218e-06,
+      "loss": 1.3184,
+      "step": 12916
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.625624728576159e-06,
+      "loss": 1.0477,
+      "step": 12917
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.624158348280263e-06,
+      "loss": 1.0306,
+      "step": 12918
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6226920499061164e-06,
+      "loss": 1.0612,
+      "step": 12919
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6212258334893e-06,
+      "loss": 1.1916,
+      "step": 12920
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.619759699065398e-06,
+      "loss": 1.0807,
+      "step": 12921
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.618293646669987e-06,
+      "loss": 0.9386,
+      "step": 12922
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.616827676338643e-06,
+      "loss": 1.172,
+      "step": 12923
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6153617881069405e-06,
+      "loss": 1.0443,
+      "step": 12924
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.613895982010456e-06,
+      "loss": 1.1916,
+      "step": 12925
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6124302580847565e-06,
+      "loss": 0.9663,
+      "step": 12926
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.610964616365414e-06,
+      "loss": 0.953,
+      "step": 12927
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.60949905688799e-06,
+      "loss": 1.0742,
+      "step": 12928
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.608033579688056e-06,
+      "loss": 1.1413,
+      "step": 12929
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.60656818480117e-06,
+      "loss": 1.0511,
+      "step": 12930
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.605102872262896e-06,
+      "loss": 1.0515,
+      "step": 12931
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.603637642108789e-06,
+      "loss": 1.009,
+      "step": 12932
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.60217249437441e-06,
+      "loss": 1.2242,
+      "step": 12933
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.60070742909531e-06,
+      "loss": 1.1726,
+      "step": 12934
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5992424463070455e-06,
+      "loss": 1.0402,
+      "step": 12935
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.597777546045163e-06,
+      "loss": 0.9567,
+      "step": 12936
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.596312728345218e-06,
+      "loss": 1.1469,
+      "step": 12937
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.59484799324275e-06,
+      "loss": 1.046,
+      "step": 12938
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.593383340773309e-06,
+      "loss": 1.1841,
+      "step": 12939
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.591918770972434e-06,
+      "loss": 0.9951,
+      "step": 12940
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.590454283875669e-06,
+      "loss": 1.0878,
+      "step": 12941
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.58898987951855e-06,
+      "loss": 1.0479,
+      "step": 12942
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.587525557936616e-06,
+      "loss": 1.0533,
+      "step": 12943
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5860613191654e-06,
+      "loss": 1.1601,
+      "step": 12944
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.584597163240438e-06,
+      "loss": 1.0495,
+      "step": 12945
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.583133090197257e-06,
+      "loss": 1.1399,
+      "step": 12946
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.581669100071388e-06,
+      "loss": 1.1695,
+      "step": 12947
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.580205192898356e-06,
+      "loss": 1.1249,
+      "step": 12948
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.578741368713688e-06,
+      "loss": 1.0518,
+      "step": 12949
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.577277627552904e-06,
+      "loss": 1.1436,
+      "step": 12950
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.575813969451528e-06,
+      "loss": 1.1968,
+      "step": 12951
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.574350394445074e-06,
+      "loss": 1.1823,
+      "step": 12952
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.572886902569064e-06,
+      "loss": 1.0693,
+      "step": 12953
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.571423493859008e-06,
+      "loss": 1.0097,
+      "step": 12954
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.569960168350422e-06,
+      "loss": 1.0946,
+      "step": 12955
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5684969260788155e-06,
+      "loss": 1.09,
+      "step": 12956
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.567033767079696e-06,
+      "loss": 1.1008,
+      "step": 12957
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.565570691388572e-06,
+      "loss": 1.0649,
+      "step": 12958
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.564107699040944e-06,
+      "loss": 1.1225,
+      "step": 12959
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.562644790072322e-06,
+      "loss": 1.095,
+      "step": 12960
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5611819645182e-06,
+      "loss": 0.9362,
+      "step": 12961
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.559719222414079e-06,
+      "loss": 1.1544,
+      "step": 12962
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.558256563795453e-06,
+      "loss": 1.1422,
+      "step": 12963
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5567939886978224e-06,
+      "loss": 1.2279,
+      "step": 12964
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.555331497156671e-06,
+      "loss": 1.1538,
+      "step": 12965
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5538690892074986e-06,
+      "loss": 1.0679,
+      "step": 12966
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5524067648857836e-06,
+      "loss": 1.194,
+      "step": 12967
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.550944524227022e-06,
+      "loss": 1.0822,
+      "step": 12968
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.549482367266691e-06,
+      "loss": 1.1541,
+      "step": 12969
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.548020294040276e-06,
+      "loss": 1.2079,
+      "step": 12970
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.546558304583255e-06,
+      "loss": 0.9146,
+      "step": 12971
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.54509639893111e-06,
+      "loss": 1.0384,
+      "step": 12972
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.543634577119313e-06,
+      "loss": 1.2661,
+      "step": 12973
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.542172839183343e-06,
+      "loss": 1.0538,
+      "step": 12974
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5407111851586655e-06,
+      "loss": 1.1419,
+      "step": 12975
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.539249615080756e-06,
+      "loss": 1.0148,
+      "step": 12976
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5377881289850794e-06,
+      "loss": 1.1738,
+      "step": 12977
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5363267269071055e-06,
+      "loss": 1.0887,
+      "step": 12978
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.534865408882293e-06,
+      "loss": 1.1708,
+      "step": 12979
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.533404174946107e-06,
+      "loss": 1.1023,
+      "step": 12980
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.531943025134006e-06,
+      "loss": 1.1171,
+      "step": 12981
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.530481959481452e-06,
+      "loss": 0.9965,
+      "step": 12982
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.529020978023894e-06,
+      "loss": 1.0134,
+      "step": 12983
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.527560080796791e-06,
+      "loss": 1.1955,
+      "step": 12984
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5260992678355904e-06,
+      "loss": 1.1806,
+      "step": 12985
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.524638539175747e-06,
+      "loss": 1.0721,
+      "step": 12986
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.523177894852704e-06,
+      "loss": 1.0579,
+      "step": 12987
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.521717334901909e-06,
+      "loss": 1.0006,
+      "step": 12988
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.520256859358804e-06,
+      "loss": 1.1712,
+      "step": 12989
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.518796468258834e-06,
+      "loss": 1.2477,
+      "step": 12990
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.517336161637433e-06,
+      "loss": 1.2008,
+      "step": 12991
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.515875939530043e-06,
+      "loss": 1.2932,
+      "step": 12992
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.514415801972096e-06,
+      "loss": 0.9347,
+      "step": 12993
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.512955748999029e-06,
+      "loss": 1.0146,
+      "step": 12994
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.511495780646269e-06,
+      "loss": 1.1482,
+      "step": 12995
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.510035896949248e-06,
+      "loss": 1.1746,
+      "step": 12996
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.508576097943391e-06,
+      "loss": 1.1684,
+      "step": 12997
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.507116383664127e-06,
+      "loss": 1.3522,
+      "step": 12998
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.505656754146873e-06,
+      "loss": 1.2027,
+      "step": 12999
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.504197209427051e-06,
+      "loss": 1.0607,
+      "step": 13000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.502737749540084e-06,
+      "loss": 1.1627,
+      "step": 13001
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.501278374521389e-06,
+      "loss": 0.9704,
+      "step": 13002
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.499819084406377e-06,
+      "loss": 1.1322,
+      "step": 13003
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.498359879230463e-06,
+      "loss": 1.1677,
+      "step": 13004
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.496900759029055e-06,
+      "loss": 1.1166,
+      "step": 13005
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.495441723837567e-06,
+      "loss": 1.0938,
+      "step": 13006
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.493982773691398e-06,
+      "loss": 0.9478,
+      "step": 13007
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.492523908625958e-06,
+      "loss": 1.1729,
+      "step": 13008
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.491065128676648e-06,
+      "loss": 1.1696,
+      "step": 13009
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4896064338788674e-06,
+      "loss": 1.1787,
+      "step": 13010
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.488147824268014e-06,
+      "loss": 1.0614,
+      "step": 13011
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.486689299879487e-06,
+      "loss": 1.2161,
+      "step": 13012
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.485230860748677e-06,
+      "loss": 1.058,
+      "step": 13013
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.483772506910978e-06,
+      "loss": 1.068,
+      "step": 13014
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.482314238401779e-06,
+      "loss": 1.2043,
+      "step": 13015
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4808560552564685e-06,
+      "loss": 1.0204,
+      "step": 13016
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.479397957510432e-06,
+      "loss": 1.1445,
+      "step": 13017
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.477939945199054e-06,
+      "loss": 1.2808,
+      "step": 13018
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.476482018357713e-06,
+      "loss": 1.0869,
+      "step": 13019
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.475024177021795e-06,
+      "loss": 1.1361,
+      "step": 13020
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4735664212266705e-06,
+      "loss": 1.1557,
+      "step": 13021
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.472108751007719e-06,
+      "loss": 1.0348,
+      "step": 13022
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.470651166400313e-06,
+      "loss": 0.9076,
+      "step": 13023
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.469193667439826e-06,
+      "loss": 1.0886,
+      "step": 13024
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.467736254161621e-06,
+      "loss": 1.0751,
+      "step": 13025
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.466278926601073e-06,
+      "loss": 0.9653,
+      "step": 13026
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.464821684793538e-06,
+      "loss": 1.2737,
+      "step": 13027
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.46336452877439e-06,
+      "loss": 1.1088,
+      "step": 13028
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.46190745857898e-06,
+      "loss": 1.1211,
+      "step": 13029
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.460450474242672e-06,
+      "loss": 1.0773,
+      "step": 13030
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.458993575800822e-06,
+      "loss": 1.1102,
+      "step": 13031
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.457536763288785e-06,
+      "loss": 0.9027,
+      "step": 13032
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.456080036741909e-06,
+      "loss": 1.038,
+      "step": 13033
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4546233961955525e-06,
+      "loss": 1.0987,
+      "step": 13034
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.453166841685057e-06,
+      "loss": 1.1774,
+      "step": 13035
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.451710373245774e-06,
+      "loss": 1.1091,
+      "step": 13036
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.450253990913042e-06,
+      "loss": 1.2167,
+      "step": 13037
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4487976947222085e-06,
+      "loss": 1.0285,
+      "step": 13038
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.447341484708609e-06,
+      "loss": 1.2017,
+      "step": 13039
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.445885360907586e-06,
+      "loss": 1.025,
+      "step": 13040
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.444429323354472e-06,
+      "loss": 0.9651,
+      "step": 13041
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4429733720846e-06,
+      "loss": 1.2695,
+      "step": 13042
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.441517507133307e-06,
+      "loss": 1.103,
+      "step": 13043
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.440061728535916e-06,
+      "loss": 0.9593,
+      "step": 13044
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.438606036327759e-06,
+      "loss": 1.153,
+      "step": 13045
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.437150430544159e-06,
+      "loss": 1.1315,
+      "step": 13046
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.435694911220441e-06,
+      "loss": 1.136,
+      "step": 13047
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.434239478391923e-06,
+      "loss": 0.9962,
+      "step": 13048
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.43278413209393e-06,
+      "loss": 1.0557,
+      "step": 13049
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.431328872361773e-06,
+      "loss": 1.13,
+      "step": 13050
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.429873699230771e-06,
+      "loss": 0.9384,
+      "step": 13051
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.428418612736232e-06,
+      "loss": 1.1062,
+      "step": 13052
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.426963612913473e-06,
+      "loss": 1.0824,
+      "step": 13053
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.425508699797796e-06,
+      "loss": 1.0331,
+      "step": 13054
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.424053873424514e-06,
+      "loss": 1.2717,
+      "step": 13055
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.422599133828924e-06,
+      "loss": 1.0366,
+      "step": 13056
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.421144481046335e-06,
+      "loss": 1.1494,
+      "step": 13057
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.419689915112044e-06,
+      "loss": 1.1569,
+      "step": 13058
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.41823543606135e-06,
+      "loss": 0.9228,
+      "step": 13059
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.416781043929547e-06,
+      "loss": 1.0576,
+      "step": 13060
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.41532673875193e-06,
+      "loss": 1.1204,
+      "step": 13061
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.413872520563792e-06,
+      "loss": 1.1993,
+      "step": 13062
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4124183894004224e-06,
+      "loss": 1.0564,
+      "step": 13063
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.410964345297105e-06,
+      "loss": 1.1213,
+      "step": 13064
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.409510388289131e-06,
+      "loss": 1.1663,
+      "step": 13065
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.408056518411778e-06,
+      "loss": 1.2849,
+      "step": 13066
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.406602735700331e-06,
+      "loss": 1.1132,
+      "step": 13067
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.405149040190067e-06,
+      "loss": 1.288,
+      "step": 13068
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.403695431916266e-06,
+      "loss": 1.1817,
+      "step": 13069
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.402241910914198e-06,
+      "loss": 1.0774,
+      "step": 13070
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4007884772191395e-06,
+      "loss": 1.1865,
+      "step": 13071
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.39933513086636e-06,
+      "loss": 1.0447,
+      "step": 13072
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.397881871891127e-06,
+      "loss": 1.1696,
+      "step": 13073
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.396428700328707e-06,
+      "loss": 1.236,
+      "step": 13074
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3949756162143665e-06,
+      "loss": 1.0184,
+      "step": 13075
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.393522619583363e-06,
+      "loss": 1.0995,
+      "step": 13076
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.392069710470963e-06,
+      "loss": 1.2717,
+      "step": 13077
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.390616888912418e-06,
+      "loss": 1.1648,
+      "step": 13078
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3891641549429865e-06,
+      "loss": 1.0945,
+      "step": 13079
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.387711508597921e-06,
+      "loss": 1.0567,
+      "step": 13080
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.386258949912476e-06,
+      "loss": 1.1388,
+      "step": 13081
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3848064789218965e-06,
+      "loss": 1.1198,
+      "step": 13082
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.38335409566143e-06,
+      "loss": 0.9143,
+      "step": 13083
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3819018001663255e-06,
+      "loss": 1.0406,
+      "step": 13084
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.380449592471826e-06,
+      "loss": 1.0991,
+      "step": 13085
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.378997472613168e-06,
+      "loss": 1.02,
+      "step": 13086
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.377545440625594e-06,
+      "loss": 1.0983,
+      "step": 13087
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.376093496544338e-06,
+      "loss": 1.1251,
+      "step": 13088
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.374641640404636e-06,
+      "loss": 1.0845,
+      "step": 13089
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3731898722417186e-06,
+      "loss": 1.0748,
+      "step": 13090
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.371738192090818e-06,
+      "loss": 1.1264,
+      "step": 13091
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.37028659998716e-06,
+      "loss": 1.0974,
+      "step": 13092
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.368835095965973e-06,
+      "loss": 1.1361,
+      "step": 13093
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3673836800624775e-06,
+      "loss": 1.0389,
+      "step": 13094
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.365932352311899e-06,
+      "loss": 1.1668,
+      "step": 13095
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.364481112749453e-06,
+      "loss": 1.0918,
+      "step": 13096
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.363029961410361e-06,
+      "loss": 1.1249,
+      "step": 13097
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.361578898329834e-06,
+      "loss": 1.3869,
+      "step": 13098
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.360127923543088e-06,
+      "loss": 1.0309,
+      "step": 13099
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.358677037085331e-06,
+      "loss": 1.1653,
+      "step": 13100
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.357226238991776e-06,
+      "loss": 1.0971,
+      "step": 13101
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.355775529297625e-06,
+      "loss": 1.0845,
+      "step": 13102
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.354324908038087e-06,
+      "loss": 1.0438,
+      "step": 13103
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.352874375248359e-06,
+      "loss": 1.1818,
+      "step": 13104
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3514239309636474e-06,
+      "loss": 1.0153,
+      "step": 13105
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.349973575219144e-06,
+      "loss": 1.261,
+      "step": 13106
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.34852330805005e-06,
+      "loss": 1.0958,
+      "step": 13107
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.347073129491554e-06,
+      "loss": 1.1133,
+      "step": 13108
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.345623039578853e-06,
+      "loss": 1.0427,
+      "step": 13109
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.344173038347132e-06,
+      "loss": 1.2058,
+      "step": 13110
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3427231258315816e-06,
+      "loss": 1.1535,
+      "step": 13111
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.341273302067384e-06,
+      "loss": 1.2017,
+      "step": 13112
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.339823567089722e-06,
+      "loss": 1.0619,
+      "step": 13113
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3383739209337806e-06,
+      "loss": 1.0729,
+      "step": 13114
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.336924363634734e-06,
+      "loss": 1.1613,
+      "step": 13115
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.33547489522776e-06,
+      "loss": 1.3427,
+      "step": 13116
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3340255157480344e-06,
+      "loss": 1.2121,
+      "step": 13117
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.332576225230727e-06,
+      "loss": 0.8771,
+      "step": 13118
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.331127023711012e-06,
+      "loss": 1.0346,
+      "step": 13119
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.329677911224052e-06,
+      "loss": 0.9965,
+      "step": 13120
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.328228887805017e-06,
+      "loss": 1.0896,
+      "step": 13121
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.326779953489065e-06,
+      "loss": 1.1376,
+      "step": 13122
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.325331108311364e-06,
+      "loss": 1.1718,
+      "step": 13123
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.323882352307069e-06,
+      "loss": 1.2569,
+      "step": 13124
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.32243368551134e-06,
+      "loss": 1.1517,
+      "step": 13125
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.320985107959328e-06,
+      "loss": 1.2094,
+      "step": 13126
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.319536619686187e-06,
+      "loss": 1.1738,
+      "step": 13127
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.31808822072707e-06,
+      "loss": 1.0703,
+      "step": 13128
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.316639911117124e-06,
+      "loss": 0.8444,
+      "step": 13129
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.315191690891496e-06,
+      "loss": 1.1396,
+      "step": 13130
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.313743560085326e-06,
+      "loss": 1.1641,
+      "step": 13131
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.312295518733761e-06,
+      "loss": 1.0791,
+      "step": 13132
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.310847566871937e-06,
+      "loss": 1.0996,
+      "step": 13133
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.309399704534995e-06,
+      "loss": 1.0786,
+      "step": 13134
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.307951931758067e-06,
+      "loss": 1.0194,
+      "step": 13135
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3065042485762906e-06,
+      "loss": 1.1205,
+      "step": 13136
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.30505665502479e-06,
+      "loss": 1.0823,
+      "step": 13137
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.303609151138701e-06,
+      "loss": 1.1334,
+      "step": 13138
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3021617369531455e-06,
+      "loss": 1.199,
+      "step": 13139
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.30071441250325e-06,
+      "loss": 1.1834,
+      "step": 13140
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.299267177824137e-06,
+      "loss": 1.2691,
+      "step": 13141
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2978200329509265e-06,
+      "loss": 1.0276,
+      "step": 13142
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.296372977918734e-06,
+      "loss": 0.9581,
+      "step": 13143
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.294926012762681e-06,
+      "loss": 1.045,
+      "step": 13144
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2934791375178725e-06,
+      "loss": 1.0575,
+      "step": 13145
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2920323522194295e-06,
+      "loss": 1.1333,
+      "step": 13146
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2905856569024535e-06,
+      "loss": 1.1418,
+      "step": 13147
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2891390516020566e-06,
+      "loss": 1.229,
+      "step": 13148
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.287692536353341e-06,
+      "loss": 1.0138,
+      "step": 13149
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.28624611119141e-06,
+      "loss": 1.1602,
+      "step": 13150
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.284799776151363e-06,
+      "loss": 1.1316,
+      "step": 13151
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.283353531268301e-06,
+      "loss": 1.1769,
+      "step": 13152
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.281907376577316e-06,
+      "loss": 1.1462,
+      "step": 13153
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.280461312113508e-06,
+      "loss": 1.0488,
+      "step": 13154
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.279015337911963e-06,
+      "loss": 1.1026,
+      "step": 13155
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.277569454007775e-06,
+      "loss": 1.1231,
+      "step": 13156
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.276123660436026e-06,
+      "loss": 1.089,
+      "step": 13157
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2746779572318065e-06,
+      "loss": 0.9922,
+      "step": 13158
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.273232344430195e-06,
+      "loss": 1.1909,
+      "step": 13159
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.271786822066277e-06,
+      "loss": 1.1749,
+      "step": 13160
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.270341390175126e-06,
+      "loss": 1.08,
+      "step": 13161
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.268896048791824e-06,
+      "loss": 0.9932,
+      "step": 13162
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.267450797951439e-06,
+      "loss": 1.2689,
+      "step": 13163
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2660056376890476e-06,
+      "loss": 1.0747,
+      "step": 13164
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.264560568039717e-06,
+      "loss": 1.2041,
+      "step": 13165
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2631155890385145e-06,
+      "loss": 1.0175,
+      "step": 13166
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.261670700720508e-06,
+      "loss": 1.0584,
+      "step": 13167
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.260225903120763e-06,
+      "loss": 1.0136,
+      "step": 13168
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.258781196274333e-06,
+      "loss": 1.0028,
+      "step": 13169
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2573365802162834e-06,
+      "loss": 1.1546,
+      "step": 13170
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.255892054981667e-06,
+      "loss": 1.0725,
+      "step": 13171
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.254447620605541e-06,
+      "loss": 1.1175,
+      "step": 13172
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.253003277122953e-06,
+      "loss": 1.2256,
+      "step": 13173
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.251559024568958e-06,
+      "loss": 1.2247,
+      "step": 13174
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.250114862978601e-06,
+      "loss": 1.1269,
+      "step": 13175
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2486707923869305e-06,
+      "loss": 1.0145,
+      "step": 13176
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.247226812828985e-06,
+      "loss": 0.9044,
+      "step": 13177
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.245782924339809e-06,
+      "loss": 1.1156,
+      "step": 13178
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2443391269544405e-06,
+      "loss": 0.9931,
+      "step": 13179
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.242895420707917e-06,
+      "loss": 1.1586,
+      "step": 13180
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.241451805635272e-06,
+      "loss": 1.1039,
+      "step": 13181
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.240008281771539e-06,
+      "loss": 1.1608,
+      "step": 13182
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.238564849151745e-06,
+      "loss": 1.1286,
+      "step": 13183
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.237121507810924e-06,
+      "loss": 1.0459,
+      "step": 13184
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.235678257784094e-06,
+      "loss": 1.1437,
+      "step": 13185
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.234235099106284e-06,
+      "loss": 1.1985,
+      "step": 13186
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.232792031812512e-06,
+      "loss": 1.1544,
+      "step": 13187
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.231349055937801e-06,
+      "loss": 1.0297,
+      "step": 13188
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.229906171517162e-06,
+      "loss": 1.1168,
+      "step": 13189
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.228463378585616e-06,
+      "loss": 1.2093,
+      "step": 13190
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.227020677178169e-06,
+      "loss": 1.1795,
+      "step": 13191
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2255780673298375e-06,
+      "loss": 1.1035,
+      "step": 13192
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.224135549075623e-06,
+      "loss": 1.0061,
+      "step": 13193
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.222693122450537e-06,
+      "loss": 1.1262,
+      "step": 13194
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.221250787489578e-06,
+      "loss": 1.2009,
+      "step": 13195
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.219808544227751e-06,
+      "loss": 1.1804,
+      "step": 13196
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.218366392700053e-06,
+      "loss": 1.1892,
+      "step": 13197
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.216924332941481e-06,
+      "loss": 1.108,
+      "step": 13198
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.215482364987029e-06,
+      "loss": 1.1816,
+      "step": 13199
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.214040488871693e-06,
+      "loss": 1.0242,
+      "step": 13200
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.212598704630457e-06,
+      "loss": 1.1817,
+      "step": 13201
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.211157012298314e-06,
+      "loss": 0.9377,
+      "step": 13202
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.209715411910246e-06,
+      "loss": 1.1836,
+      "step": 13203
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.2082739035012404e-06,
+      "loss": 1.1474,
+      "step": 13204
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.206832487106276e-06,
+      "loss": 1.0937,
+      "step": 13205
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.2053911627603305e-06,
+      "loss": 1.1407,
+      "step": 13206
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.203949930498382e-06,
+      "loss": 0.9,
+      "step": 13207
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.202508790355406e-06,
+      "loss": 1.1683,
+      "step": 13208
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.201067742366371e-06,
+      "loss": 1.0152,
+      "step": 13209
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.199626786566253e-06,
+      "loss": 1.0695,
+      "step": 13210
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.198185922990014e-06,
+      "loss": 1.02,
+      "step": 13211
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.196745151672624e-06,
+      "loss": 1.1621,
+      "step": 13212
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.195304472649042e-06,
+      "loss": 1.1243,
+      "step": 13213
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.193863885954231e-06,
+      "loss": 1.2121,
+      "step": 13214
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.192423391623152e-06,
+      "loss": 1.1418,
+      "step": 13215
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.190982989690759e-06,
+      "loss": 1.1495,
+      "step": 13216
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.189542680192007e-06,
+      "loss": 0.9914,
+      "step": 13217
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.188102463161847e-06,
+      "loss": 1.1098,
+      "step": 13218
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.186662338635232e-06,
+      "loss": 1.0463,
+      "step": 13219
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.185222306647105e-06,
+      "loss": 1.0535,
+      "step": 13220
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.183782367232416e-06,
+      "loss": 1.2173,
+      "step": 13221
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1823425204261035e-06,
+      "loss": 1.0372,
+      "step": 13222
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.180902766263113e-06,
+      "loss": 1.2226,
+      "step": 13223
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.179463104778378e-06,
+      "loss": 1.1747,
+      "step": 13224
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1780235360068404e-06,
+      "loss": 1.0751,
+      "step": 13225
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.17658405998343e-06,
+      "loss": 1.1566,
+      "step": 13226
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.175144676743081e-06,
+      "loss": 0.9354,
+      "step": 13227
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.173705386320721e-06,
+      "loss": 1.0689,
+      "step": 13228
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.172266188751282e-06,
+      "loss": 1.0681,
+      "step": 13229
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.170827084069682e-06,
+      "loss": 1.1305,
+      "step": 13230
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.169388072310849e-06,
+      "loss": 1.1957,
+      "step": 13231
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1679491535097e-06,
+      "loss": 1.1864,
+      "step": 13232
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.166510327701159e-06,
+      "loss": 1.1983,
+      "step": 13233
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.165071594920134e-06,
+      "loss": 1.0129,
+      "step": 13234
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.163632955201547e-06,
+      "loss": 1.0599,
+      "step": 13235
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.162194408580304e-06,
+      "loss": 1.2196,
+      "step": 13236
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.160755955091318e-06,
+      "loss": 1.129,
+      "step": 13237
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.159317594769492e-06,
+      "loss": 1.0501,
+      "step": 13238
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.157879327649734e-06,
+      "loss": 0.9515,
+      "step": 13239
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.156441153766944e-06,
+      "loss": 1.0746,
+      "step": 13240
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.155003073156026e-06,
+      "loss": 1.1343,
+      "step": 13241
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.153565085851874e-06,
+      "loss": 1.1706,
+      "step": 13242
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.152127191889388e-06,
+      "loss": 1.1498,
+      "step": 13243
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.150689391303455e-06,
+      "loss": 1.0907,
+      "step": 13244
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.149251684128973e-06,
+      "loss": 1.1237,
+      "step": 13245
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.147814070400826e-06,
+      "loss": 1.1185,
+      "step": 13246
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.146376550153904e-06,
+      "loss": 1.1241,
+      "step": 13247
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.144939123423088e-06,
+      "loss": 1.2445,
+      "step": 13248
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.143501790243266e-06,
+      "loss": 0.9591,
+      "step": 13249
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.142064550649308e-06,
+      "loss": 1.1661,
+      "step": 13250
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.140627404676101e-06,
+      "loss": 1.0726,
+      "step": 13251
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.139190352358517e-06,
+      "loss": 1.0435,
+      "step": 13252
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.13775339373143e-06,
+      "loss": 0.9917,
+      "step": 13253
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.136316528829707e-06,
+      "loss": 1.1363,
+      "step": 13254
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1348797576882215e-06,
+      "loss": 1.2535,
+      "step": 13255
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.133443080341836e-06,
+      "loss": 1.0909,
+      "step": 13256
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.132006496825418e-06,
+      "loss": 1.0452,
+      "step": 13257
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.130570007173824e-06,
+      "loss": 1.0881,
+      "step": 13258
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.129133611421919e-06,
+      "loss": 1.0983,
+      "step": 13259
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1276973096045555e-06,
+      "loss": 0.9112,
+      "step": 13260
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.126261101756592e-06,
+      "loss": 1.1533,
+      "step": 13261
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.124824987912877e-06,
+      "loss": 1.0808,
+      "step": 13262
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1233889681082655e-06,
+      "loss": 1.1189,
+      "step": 13263
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.121953042377602e-06,
+      "loss": 1.0429,
+      "step": 13264
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.120517210755734e-06,
+      "loss": 1.1557,
+      "step": 13265
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.119081473277502e-06,
+      "loss": 1.0456,
+      "step": 13266
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.117645829977752e-06,
+      "loss": 1.051,
+      "step": 13267
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.116210280891317e-06,
+      "loss": 1.0032,
+      "step": 13268
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.114774826053041e-06,
+      "loss": 0.9443,
+      "step": 13269
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1133394654977495e-06,
+      "loss": 1.1377,
+      "step": 13270
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.111904199260283e-06,
+      "loss": 1.183,
+      "step": 13271
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.110469027375463e-06,
+      "loss": 1.1701,
+      "step": 13272
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.109033949878124e-06,
+      "loss": 1.0991,
+      "step": 13273
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.107598966803085e-06,
+      "loss": 1.1286,
+      "step": 13274
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.106164078185176e-06,
+      "loss": 1.2429,
+      "step": 13275
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.10472928405921e-06,
+      "loss": 1.1249,
+      "step": 13276
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.103294584460011e-06,
+      "loss": 1.2083,
+      "step": 13277
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.101859979422391e-06,
+      "loss": 0.9644,
+      "step": 13278
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.100425468981168e-06,
+      "loss": 1.1551,
+      "step": 13279
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.098991053171147e-06,
+      "loss": 1.1268,
+      "step": 13280
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.097556732027143e-06,
+      "loss": 1.008,
+      "step": 13281
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.096122505583959e-06,
+      "loss": 1.0267,
+      "step": 13282
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0946883738764025e-06,
+      "loss": 1.0759,
+      "step": 13283
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.093254336939272e-06,
+      "loss": 1.1263,
+      "step": 13284
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.091820394807373e-06,
+      "loss": 1.0139,
+      "step": 13285
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0903865475154955e-06,
+      "loss": 1.2205,
+      "step": 13286
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.088952795098442e-06,
+      "loss": 1.1015,
+      "step": 13287
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0875191375909995e-06,
+      "loss": 1.0698,
+      "step": 13288
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.086085575027962e-06,
+      "loss": 1.0852,
+      "step": 13289
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0846521074441176e-06,
+      "loss": 1.1168,
+      "step": 13290
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.083218734874253e-06,
+      "loss": 1.0707,
+      "step": 13291
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.081785457353148e-06,
+      "loss": 1.1009,
+      "step": 13292
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.080352274915591e-06,
+      "loss": 1.1087,
+      "step": 13293
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0789191875963525e-06,
+      "loss": 1.046,
+      "step": 13294
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.077486195430218e-06,
+      "loss": 1.1713,
+      "step": 13295
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.076053298451955e-06,
+      "loss": 0.9378,
+      "step": 13296
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.07462049669634e-06,
+      "loss": 1.0213,
+      "step": 13297
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.073187790198141e-06,
+      "loss": 1.1647,
+      "step": 13298
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.071755178992124e-06,
+      "loss": 1.1143,
+      "step": 13299
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.070322663113059e-06,
+      "loss": 1.065,
+      "step": 13300
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.068890242595704e-06,
+      "loss": 1.1,
+      "step": 13301
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0674579174748234e-06,
+      "loss": 0.991,
+      "step": 13302
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.066025687785173e-06,
+      "loss": 1.1977,
+      "step": 13303
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.064593553561512e-06,
+      "loss": 1.1131,
+      "step": 13304
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0631615148385895e-06,
+      "loss": 1.0398,
+      "step": 13305
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0617295716511605e-06,
+      "loss": 1.1133,
+      "step": 13306
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.060297724033972e-06,
+      "loss": 1.1044,
+      "step": 13307
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0588659720217725e-06,
+      "loss": 1.1178,
+      "step": 13308
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.057434315649304e-06,
+      "loss": 1.0896,
+      "step": 13309
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.056002754951311e-06,
+      "loss": 1.1824,
+      "step": 13310
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.054571289962531e-06,
+      "loss": 1.15,
+      "step": 13311
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.053139920717704e-06,
+      "loss": 1.1118,
+      "step": 13312
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.051708647251563e-06,
+      "loss": 1.0009,
+      "step": 13313
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.050277469598843e-06,
+      "loss": 1.1776,
+      "step": 13314
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0488463877942715e-06,
+      "loss": 1.1114,
+      "step": 13315
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.04741540187258e-06,
+      "loss": 1.2786,
+      "step": 13316
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0459845118684905e-06,
+      "loss": 1.0522,
+      "step": 13317
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.04455371781673e-06,
+      "loss": 1.1202,
+      "step": 13318
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.043123019752017e-06,
+      "loss": 1.2288,
+      "step": 13319
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.041692417709073e-06,
+      "loss": 1.1415,
+      "step": 13320
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0402619117226114e-06,
+      "loss": 1.1231,
+      "step": 13321
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.038831501827349e-06,
+      "loss": 1.0836,
+      "step": 13322
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.037401188057996e-06,
+      "loss": 1.2251,
+      "step": 13323
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.035970970449265e-06,
+      "loss": 1.2951,
+      "step": 13324
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.034540849035857e-06,
+      "loss": 1.0167,
+      "step": 13325
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0331108238524835e-06,
+      "loss": 1.2073,
+      "step": 13326
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0316808949338425e-06,
+      "loss": 1.1911,
+      "step": 13327
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.030251062314638e-06,
+      "loss": 0.9435,
+      "step": 13328
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.028821326029563e-06,
+      "loss": 1.0804,
+      "step": 13329
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0273916861133176e-06,
+      "loss": 1.1695,
+      "step": 13330
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.025962142600591e-06,
+      "loss": 1.1018,
+      "step": 13331
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.024532695526077e-06,
+      "loss": 1.1712,
+      "step": 13332
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.023103344924461e-06,
+      "loss": 1.0551,
+      "step": 13333
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.021674090830431e-06,
+      "loss": 1.1754,
+      "step": 13334
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.020244933278676e-06,
+      "loss": 1.067,
+      "step": 13335
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.018815872303868e-06,
+      "loss": 1.1788,
+      "step": 13336
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.017386907940693e-06,
+      "loss": 1.1083,
+      "step": 13337
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.015958040223824e-06,
+      "loss": 1.0919,
+      "step": 13338
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.014529269187936e-06,
+      "loss": 1.0326,
+      "step": 13339
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.013100594867704e-06,
+      "loss": 1.1237,
+      "step": 13340
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.011672017297794e-06,
+      "loss": 0.9725,
+      "step": 13341
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.010243536512877e-06,
+      "loss": 1.1588,
+      "step": 13342
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.008815152547611e-06,
+      "loss": 1.1827,
+      "step": 13343
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.007386865436669e-06,
+      "loss": 1.1363,
+      "step": 13344
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.005958675214702e-06,
+      "loss": 1.0108,
+      "step": 13345
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.004530581916375e-06,
+      "loss": 1.0238,
+      "step": 13346
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0031025855763396e-06,
+      "loss": 1.1181,
+      "step": 13347
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.00167468622925e-06,
+      "loss": 1.0244,
+      "step": 13348
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0002468839097546e-06,
+      "loss": 1.1577,
+      "step": 13349
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.998819178652508e-06,
+      "loss": 1.0705,
+      "step": 13350
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.997391570492151e-06,
+      "loss": 1.1433,
+      "step": 13351
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9959640594633304e-06,
+      "loss": 1.1356,
+      "step": 13352
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.994536645600686e-06,
+      "loss": 1.1979,
+      "step": 13353
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9931093289388575e-06,
+      "loss": 1.2344,
+      "step": 13354
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.991682109512481e-06,
+      "loss": 1.0057,
+      "step": 13355
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.990254987356195e-06,
+      "loss": 1.0354,
+      "step": 13356
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.988827962504624e-06,
+      "loss": 1.0856,
+      "step": 13357
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.987401034992404e-06,
+      "loss": 1.1007,
+      "step": 13358
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.985974204854159e-06,
+      "loss": 1.2556,
+      "step": 13359
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9845474721245175e-06,
+      "loss": 1.0639,
+      "step": 13360
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.983120836838097e-06,
+      "loss": 1.2801,
+      "step": 13361
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.981694299029524e-06,
+      "loss": 1.2213,
+      "step": 13362
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.98026785873341e-06,
+      "loss": 1.196,
+      "step": 13363
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.978841515984375e-06,
+      "loss": 1.1317,
+      "step": 13364
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.977415270817029e-06,
+      "loss": 1.194,
+      "step": 13365
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.975989123265986e-06,
+      "loss": 1.0247,
+      "step": 13366
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.974563073365851e-06,
+      "loss": 1.1648,
+      "step": 13367
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.973137121151234e-06,
+      "loss": 0.9809,
+      "step": 13368
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.971711266656736e-06,
+      "loss": 1.1807,
+      "step": 13369
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.970285509916959e-06,
+      "loss": 1.2564,
+      "step": 13370
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9688598509665e-06,
+      "loss": 1.2224,
+      "step": 13371
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.967434289839959e-06,
+      "loss": 1.063,
+      "step": 13372
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.966008826571927e-06,
+      "loss": 1.2341,
+      "step": 13373
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.964583461196999e-06,
+      "loss": 1.023,
+      "step": 13374
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.963158193749761e-06,
+      "loss": 1.1569,
+      "step": 13375
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.961733024264803e-06,
+      "loss": 0.934,
+      "step": 13376
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.960307952776707e-06,
+      "loss": 1.1737,
+      "step": 13377
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.958882979320057e-06,
+      "loss": 1.0917,
+      "step": 13378
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9574581039294335e-06,
+      "loss": 1.2118,
+      "step": 13379
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9560333266394135e-06,
+      "loss": 1.1353,
+      "step": 13380
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.95460864748457e-06,
+      "loss": 1.1604,
+      "step": 13381
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.95318406649948e-06,
+      "loss": 1.0565,
+      "step": 13382
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.951759583718708e-06,
+      "loss": 1.1705,
+      "step": 13383
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.950335199176829e-06,
+      "loss": 1.0596,
+      "step": 13384
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.948910912908402e-06,
+      "loss": 1.222,
+      "step": 13385
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.947486724947994e-06,
+      "loss": 1.1197,
+      "step": 13386
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9460626353301665e-06,
+      "loss": 0.9343,
+      "step": 13387
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.944638644089475e-06,
+      "loss": 1.1081,
+      "step": 13388
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9432147512604795e-06,
+      "loss": 1.1156,
+      "step": 13389
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.941790956877729e-06,
+      "loss": 1.105,
+      "step": 13390
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.94036726097578e-06,
+      "loss": 1.0188,
+      "step": 13391
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.938943663589176e-06,
+      "loss": 0.9381,
+      "step": 13392
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.937520164752469e-06,
+      "loss": 1.1905,
+      "step": 13393
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.936096764500196e-06,
+      "loss": 1.1047,
+      "step": 13394
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.934673462866907e-06,
+      "loss": 1.1176,
+      "step": 13395
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.933250259887135e-06,
+      "loss": 1.0668,
+      "step": 13396
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.931827155595422e-06,
+      "loss": 1.204,
+      "step": 13397
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.930404150026296e-06,
+      "loss": 1.0565,
+      "step": 13398
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.928981243214296e-06,
+      "loss": 1.1593,
+      "step": 13399
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9275584351939476e-06,
+      "loss": 1.1767,
+      "step": 13400
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.92613572599978e-06,
+      "loss": 1.1627,
+      "step": 13401
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.924713115666316e-06,
+      "loss": 1.0886,
+      "step": 13402
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.923290604228082e-06,
+      "loss": 0.9902,
+      "step": 13403
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.921868191719594e-06,
+      "loss": 1.2445,
+      "step": 13404
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.920445878175374e-06,
+      "loss": 1.0574,
+      "step": 13405
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.919023663629933e-06,
+      "loss": 1.0251,
+      "step": 13406
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.917601548117786e-06,
+      "loss": 1.0496,
+      "step": 13407
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.916179531673443e-06,
+      "loss": 1.057,
+      "step": 13408
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.914757614331415e-06,
+      "loss": 1.1109,
+      "step": 13409
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9133357961262024e-06,
+      "loss": 1.0472,
+      "step": 13410
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.911914077092314e-06,
+      "loss": 0.9979,
+      "step": 13411
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.910492457264246e-06,
+      "loss": 1.1649,
+      "step": 13412
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.909070936676502e-06,
+      "loss": 1.1303,
+      "step": 13413
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.907649515363574e-06,
+      "loss": 0.9128,
+      "step": 13414
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.906228193359957e-06,
+      "loss": 1.0313,
+      "step": 13415
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.904806970700141e-06,
+      "loss": 1.188,
+      "step": 13416
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9033858474186165e-06,
+      "loss": 1.2269,
+      "step": 13417
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.901964823549871e-06,
+      "loss": 1.212,
+      "step": 13418
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.900543899128387e-06,
+      "loss": 1.1713,
+      "step": 13419
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.899123074188648e-06,
+      "loss": 1.2082,
+      "step": 13420
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.89770234876513e-06,
+      "loss": 1.1167,
+      "step": 13421
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.896281722892313e-06,
+      "loss": 1.1193,
+      "step": 13422
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.894861196604668e-06,
+      "loss": 1.1397,
+      "step": 13423
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8934407699366695e-06,
+      "loss": 1.0174,
+      "step": 13424
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.892020442922788e-06,
+      "loss": 1.1635,
+      "step": 13425
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.890600215597488e-06,
+      "loss": 1.1447,
+      "step": 13426
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.889180087995236e-06,
+      "loss": 1.1213,
+      "step": 13427
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.887760060150493e-06,
+      "loss": 0.9386,
+      "step": 13428
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.886340132097723e-06,
+      "loss": 1.175,
+      "step": 13429
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.884920303871376e-06,
+      "loss": 1.2014,
+      "step": 13430
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.883500575505914e-06,
+      "loss": 1.0676,
+      "step": 13431
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.882080947035784e-06,
+      "loss": 1.1196,
+      "step": 13432
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.880661418495441e-06,
+      "loss": 1.1438,
+      "step": 13433
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.879241989919331e-06,
+      "loss": 1.1769,
+      "step": 13434
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.877822661341899e-06,
+      "loss": 1.2417,
+      "step": 13435
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.876403432797587e-06,
+      "loss": 1.0082,
+      "step": 13436
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.874984304320839e-06,
+      "loss": 1.0909,
+      "step": 13437
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.873565275946088e-06,
+      "loss": 1.0417,
+      "step": 13438
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8721463477077745e-06,
+      "loss": 1.2057,
+      "step": 13439
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.870727519640327e-06,
+      "loss": 1.1488,
+      "step": 13440
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.869308791778182e-06,
+      "loss": 1.0789,
+      "step": 13441
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.867890164155763e-06,
+      "loss": 0.9816,
+      "step": 13442
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.866471636807498e-06,
+      "loss": 1.0476,
+      "step": 13443
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.865053209767808e-06,
+      "loss": 1.166,
+      "step": 13444
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.86363488307112e-06,
+      "loss": 0.9912,
+      "step": 13445
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.862216656751846e-06,
+      "loss": 1.1203,
+      "step": 13446
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.860798530844407e-06,
+      "loss": 1.0135,
+      "step": 13447
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.859380505383212e-06,
+      "loss": 0.9878,
+      "step": 13448
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.857962580402678e-06,
+      "loss": 1.0987,
+      "step": 13449
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.856544755937208e-06,
+      "loss": 1.1695,
+      "step": 13450
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.855127032021213e-06,
+      "loss": 1.1953,
+      "step": 13451
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.853709408689094e-06,
+      "loss": 1.1557,
+      "step": 13452
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.852291885975257e-06,
+      "loss": 1.0439,
+      "step": 13453
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.850874463914094e-06,
+      "loss": 1.179,
+      "step": 13454
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.849457142540008e-06,
+      "loss": 0.9578,
+      "step": 13455
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.84803992188739e-06,
+      "loss": 1.1628,
+      "step": 13456
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.846622801990633e-06,
+      "loss": 1.1553,
+      "step": 13457
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.845205782884125e-06,
+      "loss": 1.0751,
+      "step": 13458
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.843788864602256e-06,
+      "loss": 1.0923,
+      "step": 13459
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.842372047179405e-06,
+      "loss": 1.1246,
+      "step": 13460
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.84095533064996e-06,
+      "loss": 1.2004,
+      "step": 13461
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.839538715048297e-06,
+      "loss": 1.0403,
+      "step": 13462
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.838122200408794e-06,
+      "loss": 1.0382,
+      "step": 13463
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.836705786765827e-06,
+      "loss": 1.1025,
+      "step": 13464
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.835289474153762e-06,
+      "loss": 1.1425,
+      "step": 13465
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.833873262606977e-06,
+      "loss": 1.0524,
+      "step": 13466
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.832457152159837e-06,
+      "loss": 0.9733,
+      "step": 13467
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.831041142846705e-06,
+      "loss": 1.0089,
+      "step": 13468
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.829625234701941e-06,
+      "loss": 1.1591,
+      "step": 13469
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.82820942775991e-06,
+      "loss": 1.0619,
+      "step": 13470
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.826793722054969e-06,
+      "loss": 1.0928,
+      "step": 13471
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8253781176214675e-06,
+      "loss": 1.2065,
+      "step": 13472
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.823962614493764e-06,
+      "loss": 0.9377,
+      "step": 13473
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.822547212706209e-06,
+      "loss": 1.1393,
+      "step": 13474
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.821131912293145e-06,
+      "loss": 1.0945,
+      "step": 13475
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.819716713288919e-06,
+      "loss": 1.07,
+      "step": 13476
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.818301615727876e-06,
+      "loss": 1.2403,
+      "step": 13477
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.816886619644357e-06,
+      "loss": 1.1542,
+      "step": 13478
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.815471725072695e-06,
+      "loss": 1.0947,
+      "step": 13479
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.814056932047226e-06,
+      "loss": 1.0927,
+      "step": 13480
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.812642240602289e-06,
+      "loss": 1.0925,
+      "step": 13481
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.811227650772209e-06,
+      "loss": 1.1062,
+      "step": 13482
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.809813162591315e-06,
+      "loss": 1.1006,
+      "step": 13483
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.80839877609393e-06,
+      "loss": 1.2681,
+      "step": 13484
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8069844913143824e-06,
+      "loss": 0.9305,
+      "step": 13485
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.805570308286992e-06,
+      "loss": 1.1591,
+      "step": 13486
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.804156227046073e-06,
+      "loss": 1.1366,
+      "step": 13487
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.802742247625939e-06,
+      "loss": 1.1165,
+      "step": 13488
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.801328370060911e-06,
+      "loss": 1.2751,
+      "step": 13489
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.799914594385296e-06,
+      "loss": 1.1034,
+      "step": 13490
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.798500920633402e-06,
+      "loss": 1.038,
+      "step": 13491
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.797087348839529e-06,
+      "loss": 1.126,
+      "step": 13492
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.795673879037992e-06,
+      "loss": 1.1296,
+      "step": 13493
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.794260511263084e-06,
+      "loss": 1.0547,
+      "step": 13494
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.792847245549104e-06,
+      "loss": 1.1209,
+      "step": 13495
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7914340819303475e-06,
+      "loss": 1.0622,
+      "step": 13496
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.79002102044111e-06,
+      "loss": 1.1589,
+      "step": 13497
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.788608061115684e-06,
+      "loss": 1.1175,
+      "step": 13498
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.787195203988352e-06,
+      "loss": 1.1501,
+      "step": 13499
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7857824490934e-06,
+      "loss": 0.9031,
+      "step": 13500
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7843697964651144e-06,
+      "loss": 1.2002,
+      "step": 13501
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.782957246137781e-06,
+      "loss": 1.1487,
+      "step": 13502
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.781544798145673e-06,
+      "loss": 1.0579,
+      "step": 13503
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.780132452523066e-06,
+      "loss": 1.1657,
+      "step": 13504
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7787202093042316e-06,
+      "loss": 1.1282,
+      "step": 13505
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.777308068523446e-06,
+      "loss": 1.1681,
+      "step": 13506
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7758960302149755e-06,
+      "loss": 1.1287,
+      "step": 13507
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.774484094413087e-06,
+      "loss": 1.1333,
+      "step": 13508
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.773072261152037e-06,
+      "loss": 1.1672,
+      "step": 13509
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.771660530466097e-06,
+      "loss": 1.0777,
+      "step": 13510
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.77024890238952e-06,
+      "loss": 1.1972,
+      "step": 13511
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.76883737695656e-06,
+      "loss": 1.0563,
+      "step": 13512
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.767425954201478e-06,
+      "loss": 0.9908,
+      "step": 13513
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.766014634158519e-06,
+      "loss": 0.9543,
+      "step": 13514
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7646034168619335e-06,
+      "loss": 1.1573,
+      "step": 13515
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.763192302345963e-06,
+      "loss": 1.1003,
+      "step": 13516
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.76178129064486e-06,
+      "loss": 0.9593,
+      "step": 13517
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.76037038179286e-06,
+      "loss": 1.1087,
+      "step": 13518
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7589595758242036e-06,
+      "loss": 1.0998,
+      "step": 13519
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.757548872773122e-06,
+      "loss": 1.1747,
+      "step": 13520
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.756138272673854e-06,
+      "loss": 1.0082,
+      "step": 13521
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.754727775560633e-06,
+      "loss": 1.0901,
+      "step": 13522
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.753317381467682e-06,
+      "loss": 1.0588,
+      "step": 13523
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7519070904292255e-06,
+      "loss": 1.124,
+      "step": 13524
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.750496902479495e-06,
+      "loss": 1.0908,
+      "step": 13525
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.749086817652707e-06,
+      "loss": 1.1384,
+      "step": 13526
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.74767683598308e-06,
+      "loss": 1.0648,
+      "step": 13527
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.746266957504828e-06,
+      "loss": 1.1609,
+      "step": 13528
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.744857182252171e-06,
+      "loss": 1.3104,
+      "step": 13529
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.743447510259315e-06,
+      "loss": 1.2562,
+      "step": 13530
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7420379415604696e-06,
+      "loss": 1.0137,
+      "step": 13531
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.740628476189838e-06,
+      "loss": 1.2491,
+      "step": 13532
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7392191141816315e-06,
+      "loss": 1.1534,
+      "step": 13533
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.737809855570046e-06,
+      "loss": 1.1253,
+      "step": 13534
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.73640070038928e-06,
+      "loss": 1.0771,
+      "step": 13535
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.734991648673527e-06,
+      "loss": 1.0775,
+      "step": 13536
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.733582700456988e-06,
+      "loss": 1.0998,
+      "step": 13537
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.732173855773848e-06,
+      "loss": 1.181,
+      "step": 13538
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7307651146582985e-06,
+      "loss": 1.0228,
+      "step": 13539
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.729356477144521e-06,
+      "loss": 1.147,
+      "step": 13540
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.727947943266706e-06,
+      "loss": 1.0509,
+      "step": 13541
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.72653951305903e-06,
+      "loss": 1.2255,
+      "step": 13542
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.725131186555672e-06,
+      "loss": 1.1394,
+      "step": 13543
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7237229637908055e-06,
+      "loss": 1.1273,
+      "step": 13544
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.722314844798611e-06,
+      "loss": 1.1742,
+      "step": 13545
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.720906829613253e-06,
+      "loss": 0.9595,
+      "step": 13546
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.719498918268903e-06,
+      "loss": 0.898,
+      "step": 13547
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.718091110799722e-06,
+      "loss": 1.1291,
+      "step": 13548
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.71668340723988e-06,
+      "loss": 1.0863,
+      "step": 13549
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.715275807623536e-06,
+      "loss": 1.0689,
+      "step": 13550
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.713868311984847e-06,
+      "loss": 1.1368,
+      "step": 13551
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.712460920357965e-06,
+      "loss": 0.9924,
+      "step": 13552
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.711053632777051e-06,
+      "loss": 1.1779,
+      "step": 13553
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.709646449276252e-06,
+      "loss": 1.1925,
+      "step": 13554
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7082393698897164e-06,
+      "loss": 1.2336,
+      "step": 13555
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.706832394651585e-06,
+      "loss": 1.0659,
+      "step": 13556
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.705425523596011e-06,
+      "loss": 1.1768,
+      "step": 13557
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.704018756757128e-06,
+      "loss": 1.1957,
+      "step": 13558
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.702612094169073e-06,
+      "loss": 1.1468,
+      "step": 13559
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.701205535865988e-06,
+      "loss": 1.0233,
+      "step": 13560
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.699799081882002e-06,
+      "loss": 1.2551,
+      "step": 13561
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.698392732251246e-06,
+      "loss": 1.1016,
+      "step": 13562
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.696986487007844e-06,
+      "loss": 1.134,
+      "step": 13563
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6955803461859295e-06,
+      "loss": 1.059,
+      "step": 13564
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.69417430981962e-06,
+      "loss": 1.14,
+      "step": 13565
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.692768377943039e-06,
+      "loss": 1.084,
+      "step": 13566
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6913625505902966e-06,
+      "loss": 1.1086,
+      "step": 13567
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.689956827795519e-06,
+      "loss": 1.1043,
+      "step": 13568
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.688551209592813e-06,
+      "loss": 1.1692,
+      "step": 13569
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.68714569601629e-06,
+      "loss": 0.9798,
+      "step": 13570
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.685740287100052e-06,
+      "loss": 1.1877,
+      "step": 13571
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.684334982878216e-06,
+      "loss": 1.1206,
+      "step": 13572
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.682929783384876e-06,
+      "loss": 0.9886,
+      "step": 13573
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.681524688654135e-06,
+      "loss": 1.1946,
+      "step": 13574
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.680119698720086e-06,
+      "loss": 1.3629,
+      "step": 13575
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.67871481361683e-06,
+      "loss": 1.0683,
+      "step": 13576
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.677310033378458e-06,
+      "loss": 1.263,
+      "step": 13577
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6759053580390596e-06,
+      "loss": 1.1064,
+      "step": 13578
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6745007876327176e-06,
+      "loss": 1.2027,
+      "step": 13579
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.673096322193523e-06,
+      "loss": 1.0502,
+      "step": 13580
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.671691961755557e-06,
+      "loss": 1.2972,
+      "step": 13581
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6702877063528975e-06,
+      "loss": 1.2395,
+      "step": 13582
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.668883556019617e-06,
+      "loss": 1.1239,
+      "step": 13583
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.667479510789797e-06,
+      "loss": 0.9858,
+      "step": 13584
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.666075570697511e-06,
+      "loss": 1.0656,
+      "step": 13585
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6646717357768264e-06,
+      "loss": 1.1531,
+      "step": 13586
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.663268006061809e-06,
+      "loss": 1.0787,
+      "step": 13587
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.661864381586517e-06,
+      "loss": 0.8427,
+      "step": 13588
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.660460862385025e-06,
+      "loss": 1.1269,
+      "step": 13589
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.659057448491384e-06,
+      "loss": 1.0784,
+      "step": 13590
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6576541399396524e-06,
+      "loss": 1.0505,
+      "step": 13591
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.656250936763882e-06,
+      "loss": 1.0028,
+      "step": 13592
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.65484783899813e-06,
+      "loss": 1.0025,
+      "step": 13593
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6534448466764416e-06,
+      "loss": 1.1503,
+      "step": 13594
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.652041959832866e-06,
+      "loss": 1.0888,
+      "step": 13595
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.650639178501438e-06,
+      "loss": 1.1,
+      "step": 13596
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6492365027162135e-06,
+      "loss": 1.1466,
+      "step": 13597
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.647833932511222e-06,
+      "loss": 1.0069,
+      "step": 13598
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6464314679204976e-06,
+      "loss": 0.9832,
+      "step": 13599
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.645029108978083e-06,
+      "loss": 1.0932,
+      "step": 13600
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.643626855718004e-06,
+      "loss": 1.1932,
+      "step": 13601
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.642224708174289e-06,
+      "loss": 1.0105,
+      "step": 13602
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.640822666380962e-06,
+      "loss": 1.051,
+      "step": 13603
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.639420730372052e-06,
+      "loss": 1.0324,
+      "step": 13604
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.638018900181576e-06,
+      "loss": 1.1473,
+      "step": 13605
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.636617175843555e-06,
+      "loss": 1.2277,
+      "step": 13606
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.635215557391998e-06,
+      "loss": 1.0457,
+      "step": 13607
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.633814044860928e-06,
+      "loss": 1.1281,
+      "step": 13608
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.632412638284349e-06,
+      "loss": 1.1785,
+      "step": 13609
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.631011337696272e-06,
+      "loss": 1.2305,
+      "step": 13610
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.629610143130696e-06,
+      "loss": 1.135,
+      "step": 13611
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.628209054621634e-06,
+      "loss": 1.0716,
+      "step": 13612
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.62680807220308e-06,
+      "loss": 1.2336,
+      "step": 13613
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6254071959090335e-06,
+      "loss": 1.1053,
+      "step": 13614
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.624006425773486e-06,
+      "loss": 1.1311,
+      "step": 13615
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.622605761830437e-06,
+      "loss": 1.1264,
+      "step": 13616
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.621205204113872e-06,
+      "loss": 1.2159,
+      "step": 13617
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.619804752657781e-06,
+      "loss": 1.1001,
+      "step": 13618
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.618404407496141e-06,
+      "loss": 0.9857,
+      "step": 13619
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.617004168662944e-06,
+      "loss": 1.1248,
+      "step": 13620
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.615604036192166e-06,
+      "loss": 1.0972,
+      "step": 13621
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.614204010117785e-06,
+      "loss": 0.8927,
+      "step": 13622
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.61280409047377e-06,
+      "loss": 1.1932,
+      "step": 13623
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.611404277294101e-06,
+      "loss": 1.0789,
+      "step": 13624
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.610004570612745e-06,
+      "loss": 1.2548,
+      "step": 13625
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.608604970463667e-06,
+      "loss": 1.0926,
+      "step": 13626
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.607205476880827e-06,
+      "loss": 1.1003,
+      "step": 13627
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6058060898981956e-06,
+      "loss": 1.1196,
+      "step": 13628
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.604406809549728e-06,
+      "loss": 0.9559,
+      "step": 13629
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.60300763586938e-06,
+      "loss": 1.2431,
+      "step": 13630
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.601608568891101e-06,
+      "loss": 1.1028,
+      "step": 13631
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.600209608648852e-06,
+      "loss": 1.1251,
+      "step": 13632
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.598810755176576e-06,
+      "loss": 1.0692,
+      "step": 13633
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.597412008508221e-06,
+      "loss": 1.2164,
+      "step": 13634
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.596013368677722e-06,
+      "loss": 1.1883,
+      "step": 13635
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.594614835719032e-06,
+      "loss": 1.1459,
+      "step": 13636
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5932164096660845e-06,
+      "loss": 1.1637,
+      "step": 13637
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.591818090552817e-06,
+      "loss": 1.1324,
+      "step": 13638
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.590419878413153e-06,
+      "loss": 1.1631,
+      "step": 13639
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.589021773281036e-06,
+      "loss": 1.039,
+      "step": 13640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.587623775190388e-06,
+      "loss": 1.0042,
+      "step": 13641
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.58622588417513e-06,
+      "loss": 1.2704,
+      "step": 13642
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.584828100269195e-06,
+      "loss": 1.0518,
+      "step": 13643
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.583430423506497e-06,
+      "loss": 1.0388,
+      "step": 13644
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.582032853920953e-06,
+      "loss": 1.054,
+      "step": 13645
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.580635391546475e-06,
+      "loss": 1.0887,
+      "step": 13646
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.579238036416983e-06,
+      "loss": 1.1202,
+      "step": 13647
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.577840788566384e-06,
+      "loss": 1.1859,
+      "step": 13648
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.576443648028584e-06,
+      "loss": 1.2657,
+      "step": 13649
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5750466148374825e-06,
+      "loss": 1.0668,
+      "step": 13650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5736496890269895e-06,
+      "loss": 1.0208,
+      "step": 13651
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5722528706310035e-06,
+      "loss": 1.0733,
+      "step": 13652
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.570856159683418e-06,
+      "loss": 1.1365,
+      "step": 13653
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.569459556218122e-06,
+      "loss": 1.1788,
+      "step": 13654
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.568063060269019e-06,
+      "loss": 1.016,
+      "step": 13655
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.566666671869991e-06,
+      "loss": 1.1729,
+      "step": 13656
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.565270391054925e-06,
+      "loss": 1.1403,
+      "step": 13657
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5638742178577e-06,
+      "loss": 1.0249,
+      "step": 13658
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.562478152312205e-06,
+      "loss": 1.2027,
+      "step": 13659
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.561082194452316e-06,
+      "loss": 1.2326,
+      "step": 13660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.559686344311908e-06,
+      "loss": 1.1003,
+      "step": 13661
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.55829060192485e-06,
+      "loss": 1.1345,
+      "step": 13662
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.556894967325019e-06,
+      "loss": 1.1606,
+      "step": 13663
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.555499440546282e-06,
+      "loss": 1.0509,
+      "step": 13664
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.554104021622503e-06,
+      "loss": 1.1646,
+      "step": 13665
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.55270871058754e-06,
+      "loss": 1.1021,
+      "step": 13666
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.551313507475256e-06,
+      "loss": 0.863,
+      "step": 13667
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.549918412319516e-06,
+      "loss": 1.1467,
+      "step": 13668
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5485234251541685e-06,
+      "loss": 1.0459,
+      "step": 13669
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.547128546013065e-06,
+      "loss": 1.153,
+      "step": 13670
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.545733774930053e-06,
+      "loss": 1.0207,
+      "step": 13671
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.544339111938987e-06,
+      "loss": 1.0364,
+      "step": 13672
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5429445570737085e-06,
+      "loss": 1.1896,
+      "step": 13673
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.541550110368055e-06,
+      "loss": 1.2167,
+      "step": 13674
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.540155771855866e-06,
+      "loss": 1.177,
+      "step": 13675
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.538761541570985e-06,
+      "loss": 1.1485,
+      "step": 13676
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.537367419547242e-06,
+      "loss": 1.1298,
+      "step": 13677
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.535973405818466e-06,
+      "loss": 1.0491,
+      "step": 13678
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5345795004184835e-06,
+      "loss": 1.1357,
+      "step": 13679
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.533185703381128e-06,
+      "loss": 1.131,
+      "step": 13680
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.53179201474022e-06,
+      "loss": 1.0621,
+      "step": 13681
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.530398434529574e-06,
+      "loss": 1.1141,
+      "step": 13682
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5290049627830175e-06,
+      "loss": 1.1291,
+      "step": 13683
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.527611599534363e-06,
+      "loss": 1.186,
+      "step": 13684
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.526218344817421e-06,
+      "loss": 1.1008,
+      "step": 13685
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.524825198665999e-06,
+      "loss": 1.136,
+      "step": 13686
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.523432161113913e-06,
+      "loss": 1.198,
+      "step": 13687
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.522039232194963e-06,
+      "loss": 0.9778,
+      "step": 13688
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.520646411942952e-06,
+      "loss": 1.0017,
+      "step": 13689
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.519253700391676e-06,
+      "loss": 0.9505,
+      "step": 13690
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.517861097574937e-06,
+      "loss": 1.126,
+      "step": 13691
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.516468603526529e-06,
+      "loss": 1.1226,
+      "step": 13692
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.515076218280242e-06,
+      "loss": 1.258,
+      "step": 13693
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5136839418698625e-06,
+      "loss": 1.2439,
+      "step": 13694
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.512291774329184e-06,
+      "loss": 1.1612,
+      "step": 13695
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.510899715691984e-06,
+      "loss": 1.0353,
+      "step": 13696
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.509507765992048e-06,
+      "loss": 1.1074,
+      "step": 13697
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5081159252631485e-06,
+      "loss": 1.1892,
+      "step": 13698
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.506724193539068e-06,
+      "loss": 1.186,
+      "step": 13699
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.505332570853579e-06,
+      "loss": 1.1547,
+      "step": 13700
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.503941057240448e-06,
+      "loss": 1.2107,
+      "step": 13701
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.502549652733442e-06,
+      "loss": 1.0621,
+      "step": 13702
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.501158357366334e-06,
+      "loss": 1.0862,
+      "step": 13703
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.499767171172882e-06,
+      "loss": 1.1052,
+      "step": 13704
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.498376094186846e-06,
+      "loss": 1.0881,
+      "step": 13705
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.496985126441978e-06,
+      "loss": 1.1548,
+      "step": 13706
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.495594267972043e-06,
+      "loss": 1.1697,
+      "step": 13707
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.494203518810788e-06,
+      "loss": 1.298,
+      "step": 13708
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.492812878991963e-06,
+      "loss": 1.066,
+      "step": 13709
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.49142234854931e-06,
+      "loss": 0.9908,
+      "step": 13710
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.490031927516581e-06,
+      "loss": 1.1724,
+      "step": 13711
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.488641615927515e-06,
+      "loss": 1.0088,
+      "step": 13712
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.487251413815848e-06,
+      "loss": 1.0395,
+      "step": 13713
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.485861321215316e-06,
+      "loss": 1.137,
+      "step": 13714
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.484471338159656e-06,
+      "loss": 1.1238,
+      "step": 13715
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.483081464682597e-06,
+      "loss": 1.022,
+      "step": 13716
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.481691700817867e-06,
+      "loss": 1.1015,
+      "step": 13717
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4803020465991884e-06,
+      "loss": 1.0552,
+      "step": 13718
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.478912502060291e-06,
+      "loss": 1.0158,
+      "step": 13719
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.477523067234891e-06,
+      "loss": 0.912,
+      "step": 13720
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.476133742156706e-06,
+      "loss": 1.1396,
+      "step": 13721
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4747445268594464e-06,
+      "loss": 1.0628,
+      "step": 13722
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.473355421376834e-06,
+      "loss": 1.0687,
+      "step": 13723
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.471966425742572e-06,
+      "loss": 1.0152,
+      "step": 13724
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.470577539990369e-06,
+      "loss": 0.9545,
+      "step": 13725
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.469188764153924e-06,
+      "loss": 0.9773,
+      "step": 13726
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.467800098266948e-06,
+      "loss": 1.0578,
+      "step": 13727
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4664115423631345e-06,
+      "loss": 1.1495,
+      "step": 13728
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.465023096476175e-06,
+      "loss": 1.0352,
+      "step": 13729
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.463634760639773e-06,
+      "loss": 0.9978,
+      "step": 13730
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.462246534887614e-06,
+      "loss": 1.1028,
+      "step": 13731
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.460858419253388e-06,
+      "loss": 1.1287,
+      "step": 13732
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.459470413770773e-06,
+      "loss": 1.1251,
+      "step": 13733
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.458082518473463e-06,
+      "loss": 0.9633,
+      "step": 13734
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.456694733395133e-06,
+      "loss": 1.1499,
+      "step": 13735
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.45530705856946e-06,
+      "loss": 1.1143,
+      "step": 13736
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.453919494030116e-06,
+      "loss": 1.1555,
+      "step": 13737
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.45253203981078e-06,
+      "loss": 0.8865,
+      "step": 13738
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.451144695945116e-06,
+      "loss": 1.027,
+      "step": 13739
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.449757462466795e-06,
+      "loss": 1.1562,
+      "step": 13740
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.448370339409473e-06,
+      "loss": 1.0665,
+      "step": 13741
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.446983326806823e-06,
+      "loss": 1.0414,
+      "step": 13742
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.445596424692497e-06,
+      "loss": 1.0718,
+      "step": 13743
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.444209633100151e-06,
+      "loss": 1.2211,
+      "step": 13744
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4428229520634335e-06,
+      "loss": 1.0075,
+      "step": 13745
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.441436381616008e-06,
+      "loss": 1.0036,
+      "step": 13746
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.440049921791512e-06,
+      "loss": 1.0014,
+      "step": 13747
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.438663572623597e-06,
+      "loss": 1.1288,
+      "step": 13748
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.437277334145897e-06,
+      "loss": 1.1389,
+      "step": 13749
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.435891206392059e-06,
+      "loss": 0.998,
+      "step": 13750
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.434505189395721e-06,
+      "loss": 0.9537,
+      "step": 13751
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.433119283190517e-06,
+      "loss": 1.0284,
+      "step": 13752
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.431733487810076e-06,
+      "loss": 1.1231,
+      "step": 13753
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.430347803288025e-06,
+      "loss": 1.0387,
+      "step": 13754
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.428962229657999e-06,
+      "loss": 0.9944,
+      "step": 13755
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.427576766953615e-06,
+      "loss": 0.9423,
+      "step": 13756
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.426191415208497e-06,
+      "loss": 1.0681,
+      "step": 13757
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.424806174456258e-06,
+      "loss": 1.2338,
+      "step": 13758
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.423421044730521e-06,
+      "loss": 0.9233,
+      "step": 13759
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.422036026064899e-06,
+      "loss": 1.0928,
+      "step": 13760
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4206511184929964e-06,
+      "loss": 1.1645,
+      "step": 13761
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.41926632204842e-06,
+      "loss": 1.2325,
+      "step": 13762
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.417881636764784e-06,
+      "loss": 1.0067,
+      "step": 13763
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.416497062675684e-06,
+      "loss": 1.1591,
+      "step": 13764
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.41511259981472e-06,
+      "loss": 1.1638,
+      "step": 13765
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.413728248215488e-06,
+      "loss": 1.1585,
+      "step": 13766
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.412344007911585e-06,
+      "loss": 1.1673,
+      "step": 13767
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.410959878936603e-06,
+      "loss": 1.0766,
+      "step": 13768
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.409575861324124e-06,
+      "loss": 1.1793,
+      "step": 13769
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.408191955107743e-06,
+      "loss": 1.2509,
+      "step": 13770
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4068081603210395e-06,
+      "loss": 1.1478,
+      "step": 13771
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.405424476997594e-06,
+      "loss": 1.1125,
+      "step": 13772
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.404040905170979e-06,
+      "loss": 1.3578,
+      "step": 13773
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.402657444874782e-06,
+      "loss": 0.9522,
+      "step": 13774
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.401274096142566e-06,
+      "loss": 1.1512,
+      "step": 13775
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.399890859007903e-06,
+      "loss": 1.135,
+      "step": 13776
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3985077335043565e-06,
+      "loss": 1.129,
+      "step": 13777
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.397124719665501e-06,
+      "loss": 1.2503,
+      "step": 13778
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.39574181752489e-06,
+      "loss": 1.0314,
+      "step": 13779
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.394359027116085e-06,
+      "loss": 0.9151,
+      "step": 13780
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3929763484726385e-06,
+      "loss": 1.1881,
+      "step": 13781
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.391593781628109e-06,
+      "loss": 0.934,
+      "step": 13782
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.390211326616048e-06,
+      "loss": 1.0038,
+      "step": 13783
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3888289834700006e-06,
+      "loss": 1.1214,
+      "step": 13784
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.387446752223507e-06,
+      "loss": 1.2659,
+      "step": 13785
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.386064632910122e-06,
+      "loss": 0.934,
+      "step": 13786
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.384682625563378e-06,
+      "loss": 1.1972,
+      "step": 13787
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.383300730216814e-06,
+      "loss": 1.075,
+      "step": 13788
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.38191894690396e-06,
+      "loss": 1.1569,
+      "step": 13789
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.380537275658355e-06,
+      "loss": 1.0333,
+      "step": 13790
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.379155716513526e-06,
+      "loss": 1.0546,
+      "step": 13791
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.377774269502999e-06,
+      "loss": 1.2527,
+      "step": 13792
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.376392934660292e-06,
+      "loss": 1.1366,
+      "step": 13793
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.375011712018935e-06,
+      "loss": 1.1649,
+      "step": 13794
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.373630601612443e-06,
+      "loss": 1.1075,
+      "step": 13795
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.372249603474329e-06,
+      "loss": 1.0251,
+      "step": 13796
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.370868717638105e-06,
+      "loss": 1.0897,
+      "step": 13797
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.369487944137286e-06,
+      "loss": 1.1134,
+      "step": 13798
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.368107283005377e-06,
+      "loss": 1.093,
+      "step": 13799
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.366726734275883e-06,
+      "loss": 0.9826,
+      "step": 13800
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.365346297982301e-06,
+      "loss": 1.1995,
+      "step": 13801
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.363965974158139e-06,
+      "loss": 0.9184,
+      "step": 13802
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.362585762836887e-06,
+      "loss": 1.2125,
+      "step": 13803
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3612056640520415e-06,
+      "loss": 1.144,
+      "step": 13804
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3598256778370886e-06,
+      "loss": 1.112,
+      "step": 13805
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.358445804225524e-06,
+      "loss": 1.0304,
+      "step": 13806
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.357066043250829e-06,
+      "loss": 1.1183,
+      "step": 13807
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.355686394946486e-06,
+      "loss": 1.0583,
+      "step": 13808
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.354306859345975e-06,
+      "loss": 1.152,
+      "step": 13809
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.352927436482776e-06,
+      "loss": 1.1167,
+      "step": 13810
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.35154812639036e-06,
+      "loss": 1.0753,
+      "step": 13811
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.350168929102202e-06,
+      "loss": 0.9833,
+      "step": 13812
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.348789844651765e-06,
+      "loss": 1.0732,
+      "step": 13813
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.347410873072524e-06,
+      "loss": 1.1919,
+      "step": 13814
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.346032014397938e-06,
+      "loss": 1.0069,
+      "step": 13815
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.344653268661465e-06,
+      "loss": 1.1631,
+      "step": 13816
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.343274635896569e-06,
+      "loss": 1.1539,
+      "step": 13817
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.341896116136705e-06,
+      "loss": 1.1631,
+      "step": 13818
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.34051770941532e-06,
+      "loss": 1.102,
+      "step": 13819
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.339139415765866e-06,
+      "loss": 1.0483,
+      "step": 13820
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.337761235221793e-06,
+      "loss": 1.2548,
+      "step": 13821
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3363831678165456e-06,
+      "loss": 1.1076,
+      "step": 13822
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3350052135835616e-06,
+      "loss": 1.1118,
+      "step": 13823
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.333627372556279e-06,
+      "loss": 1.0386,
+      "step": 13824
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.332249644768142e-06,
+      "loss": 1.0399,
+      "step": 13825
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.330872030252576e-06,
+      "loss": 1.1472,
+      "step": 13826
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.329494529043015e-06,
+      "loss": 1.1807,
+      "step": 13827
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.328117141172884e-06,
+      "loss": 1.1207,
+      "step": 13828
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.326739866675614e-06,
+      "loss": 0.9673,
+      "step": 13829
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.325362705584623e-06,
+      "loss": 1.2008,
+      "step": 13830
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.323985657933333e-06,
+      "loss": 1.2029,
+      "step": 13831
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.322608723755154e-06,
+      "loss": 1.0012,
+      "step": 13832
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.321231903083506e-06,
+      "loss": 1.1887,
+      "step": 13833
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.319855195951804e-06,
+      "loss": 1.1029,
+      "step": 13834
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.318478602393453e-06,
+      "loss": 1.162,
+      "step": 13835
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3171021224418574e-06,
+      "loss": 1.0202,
+      "step": 13836
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.315725756130419e-06,
+      "loss": 1.0311,
+      "step": 13837
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.314349503492542e-06,
+      "loss": 1.1311,
+      "step": 13838
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3129733645616245e-06,
+      "loss": 1.196,
+      "step": 13839
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.31159733937106e-06,
+      "loss": 1.0593,
+      "step": 13840
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.310221427954234e-06,
+      "loss": 1.2155,
+      "step": 13841
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.308845630344547e-06,
+      "loss": 1.2046,
+      "step": 13842
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.30746994657538e-06,
+      "loss": 1.0785,
+      "step": 13843
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.306094376680117e-06,
+      "loss": 1.1328,
+      "step": 13844
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3047189206921354e-06,
+      "loss": 1.2525,
+      "step": 13845
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.303343578644822e-06,
+      "loss": 1.0763,
+      "step": 13846
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.301968350571547e-06,
+      "loss": 1.1905,
+      "step": 13847
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.300593236505682e-06,
+      "loss": 1.0561,
+      "step": 13848
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.299218236480597e-06,
+      "loss": 1.229,
+      "step": 13849
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.297843350529664e-06,
+      "loss": 1.1765,
+      "step": 13850
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.296468578686244e-06,
+      "loss": 1.1205,
+      "step": 13851
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2950939209837e-06,
+      "loss": 1.0738,
+      "step": 13852
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.293719377455386e-06,
+      "loss": 1.2281,
+      "step": 13853
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.292344948134665e-06,
+      "loss": 1.0323,
+      "step": 13854
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.290970633054888e-06,
+      "loss": 1.1349,
+      "step": 13855
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.289596432249401e-06,
+      "loss": 1.0961,
+      "step": 13856
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.288222345751559e-06,
+      "loss": 1.0634,
+      "step": 13857
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.286848373594705e-06,
+      "loss": 1.1138,
+      "step": 13858
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.285474515812179e-06,
+      "loss": 1.173,
+      "step": 13859
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2841007724373195e-06,
+      "loss": 1.1255,
+      "step": 13860
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.282727143503468e-06,
+      "loss": 1.2189,
+      "step": 13861
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.281353629043955e-06,
+      "loss": 1.0366,
+      "step": 13862
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.279980229092112e-06,
+      "loss": 1.1216,
+      "step": 13863
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.278606943681266e-06,
+      "loss": 0.9313,
+      "step": 13864
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.277233772844748e-06,
+      "loss": 1.028,
+      "step": 13865
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.275860716615875e-06,
+      "loss": 1.2435,
+      "step": 13866
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.27448777502797e-06,
+      "loss": 1.0554,
+      "step": 13867
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.273114948114346e-06,
+      "loss": 1.1184,
+      "step": 13868
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.271742235908326e-06,
+      "loss": 1.1306,
+      "step": 13869
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.270369638443216e-06,
+      "loss": 1.1242,
+      "step": 13870
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.268997155752324e-06,
+      "loss": 1.1283,
+      "step": 13871
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.267624787868955e-06,
+      "loss": 1.1856,
+      "step": 13872
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.266252534826419e-06,
+      "loss": 1.2622,
+      "step": 13873
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.264880396658012e-06,
+      "loss": 1.2252,
+      "step": 13874
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.263508373397033e-06,
+      "loss": 1.1155,
+      "step": 13875
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.262136465076771e-06,
+      "loss": 1.1,
+      "step": 13876
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2607646717305294e-06,
+      "loss": 0.8727,
+      "step": 13877
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2593929933915896e-06,
+      "loss": 1.1256,
+      "step": 13878
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.258021430093241e-06,
+      "loss": 0.9189,
+      "step": 13879
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.256649981868763e-06,
+      "loss": 1.1015,
+      "step": 13880
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.255278648751444e-06,
+      "loss": 1.1304,
+      "step": 13881
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.253907430774559e-06,
+      "loss": 1.1033,
+      "step": 13882
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.252536327971383e-06,
+      "loss": 0.9859,
+      "step": 13883
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.251165340375183e-06,
+      "loss": 1.0691,
+      "step": 13884
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2497944680192395e-06,
+      "loss": 1.1261,
+      "step": 13885
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2484237109368145e-06,
+      "loss": 1.1022,
+      "step": 13886
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.247053069161171e-06,
+      "loss": 1.0043,
+      "step": 13887
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2456825427255684e-06,
+      "loss": 1.1384,
+      "step": 13888
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2443121316632725e-06,
+      "loss": 1.17,
+      "step": 13889
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.242941836007536e-06,
+      "loss": 1.0537,
+      "step": 13890
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.241571655791611e-06,
+      "loss": 1.1689,
+      "step": 13891
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.240201591048742e-06,
+      "loss": 1.1399,
+      "step": 13892
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.238831641812187e-06,
+      "loss": 1.2082,
+      "step": 13893
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.237461808115187e-06,
+      "loss": 1.0832,
+      "step": 13894
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.236092089990983e-06,
+      "loss": 1.0696,
+      "step": 13895
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2347224874728095e-06,
+      "loss": 1.2162,
+      "step": 13896
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.23335300059391e-06,
+      "loss": 1.2369,
+      "step": 13897
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.231983629387516e-06,
+      "loss": 1.0315,
+      "step": 13898
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.230614373886856e-06,
+      "loss": 1.0893,
+      "step": 13899
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.229245234125156e-06,
+      "loss": 1.1134,
+      "step": 13900
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2278762101356475e-06,
+      "loss": 1.2445,
+      "step": 13901
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.226507301951549e-06,
+      "loss": 1.1784,
+      "step": 13902
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.225138509606076e-06,
+      "loss": 1.082,
+      "step": 13903
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.223769833132454e-06,
+      "loss": 1.2599,
+      "step": 13904
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.222401272563891e-06,
+      "loss": 1.1456,
+      "step": 13905
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.221032827933598e-06,
+      "loss": 0.981,
+      "step": 13906
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.21966449927478e-06,
+      "loss": 1.2763,
+      "step": 13907
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.218296286620651e-06,
+      "loss": 1.2193,
+      "step": 13908
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.216928190004408e-06,
+      "loss": 1.1079,
+      "step": 13909
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.215560209459252e-06,
+      "loss": 0.9564,
+      "step": 13910
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.214192345018374e-06,
+      "loss": 1.1912,
+      "step": 13911
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.212824596714978e-06,
+      "loss": 1.068,
+      "step": 13912
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2114569645822515e-06,
+      "loss": 1.1594,
+      "step": 13913
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.210089448653379e-06,
+      "loss": 1.1584,
+      "step": 13914
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.208722048961548e-06,
+      "loss": 1.109,
+      "step": 13915
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.207354765539941e-06,
+      "loss": 0.9898,
+      "step": 13916
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.205987598421743e-06,
+      "loss": 1.0387,
+      "step": 13917
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.204620547640127e-06,
+      "loss": 1.0686,
+      "step": 13918
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.203253613228268e-06,
+      "loss": 1.0654,
+      "step": 13919
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.201886795219333e-06,
+      "loss": 1.1071,
+      "step": 13920
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.200520093646498e-06,
+      "loss": 1.2077,
+      "step": 13921
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1991535085429266e-06,
+      "loss": 1.145,
+      "step": 13922
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.197787039941781e-06,
+      "loss": 1.0277,
+      "step": 13923
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.196420687876217e-06,
+      "loss": 1.1452,
+      "step": 13924
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1950544523794e-06,
+      "loss": 1.1257,
+      "step": 13925
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.193688333484479e-06,
+      "loss": 1.187,
+      "step": 13926
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1923223312246095e-06,
+      "loss": 1.0745,
+      "step": 13927
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.190956445632933e-06,
+      "loss": 1.1518,
+      "step": 13928
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1895906767426055e-06,
+      "loss": 1.0481,
+      "step": 13929
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.188225024586764e-06,
+      "loss": 1.0777,
+      "step": 13930
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.186859489198552e-06,
+      "loss": 1.1832,
+      "step": 13931
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1854940706111e-06,
+      "loss": 1.0638,
+      "step": 13932
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1841287688575525e-06,
+      "loss": 1.0945,
+      "step": 13933
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.182763583971036e-06,
+      "loss": 1.1128,
+      "step": 13934
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.181398515984682e-06,
+      "loss": 1.0274,
+      "step": 13935
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.180033564931609e-06,
+      "loss": 1.1627,
+      "step": 13936
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.178668730844951e-06,
+      "loss": 1.048,
+      "step": 13937
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.177304013757825e-06,
+      "loss": 1.1521,
+      "step": 13938
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.175939413703343e-06,
+      "loss": 1.0758,
+      "step": 13939
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.174574930714627e-06,
+      "loss": 1.1054,
+      "step": 13940
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.173210564824789e-06,
+      "loss": 1.1765,
+      "step": 13941
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.171846316066933e-06,
+      "loss": 1.0593,
+      "step": 13942
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.170482184474166e-06,
+      "loss": 1.1667,
+      "step": 13943
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1691181700795955e-06,
+      "loss": 1.3592,
+      "step": 13944
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.167754272916321e-06,
+      "loss": 1.0089,
+      "step": 13945
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.166390493017438e-06,
+      "loss": 1.0422,
+      "step": 13946
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.16502683041604e-06,
+      "loss": 1.1004,
+      "step": 13947
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.163663285145224e-06,
+      "loss": 1.008,
+      "step": 13948
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1622998572380776e-06,
+      "loss": 0.9576,
+      "step": 13949
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.160936546727687e-06,
+      "loss": 1.3454,
+      "step": 13950
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.159573353647131e-06,
+      "loss": 1.1674,
+      "step": 13951
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.158210278029498e-06,
+      "loss": 1.249,
+      "step": 13952
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.156847319907864e-06,
+      "loss": 1.1373,
+      "step": 13953
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1554844793153005e-06,
+      "loss": 0.9932,
+      "step": 13954
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.154121756284879e-06,
+      "loss": 1.0622,
+      "step": 13955
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.152759150849674e-06,
+      "loss": 1.1373,
+      "step": 13956
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1513966630427495e-06,
+      "loss": 0.9932,
+      "step": 13957
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.150034292897169e-06,
+      "loss": 1.0601,
+      "step": 13958
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.14867204044599e-06,
+      "loss": 1.1571,
+      "step": 13959
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.147309905722276e-06,
+      "loss": 1.1795,
+      "step": 13960
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.14594788875908e-06,
+      "loss": 1.0846,
+      "step": 13961
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1445859895894525e-06,
+      "loss": 1.1453,
+      "step": 13962
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.14322420824644e-06,
+      "loss": 1.0689,
+      "step": 13963
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1418625447630966e-06,
+      "loss": 1.0599,
+      "step": 13964
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.140500999172463e-06,
+      "loss": 1.0706,
+      "step": 13965
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.139139571507579e-06,
+      "loss": 1.1703,
+      "step": 13966
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.137778261801477e-06,
+      "loss": 1.0897,
+      "step": 13967
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.136417070087202e-06,
+      "loss": 1.0647,
+      "step": 13968
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.135055996397782e-06,
+      "loss": 0.9604,
+      "step": 13969
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.133695040766244e-06,
+      "loss": 1.1376,
+      "step": 13970
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.132334203225613e-06,
+      "loss": 1.0913,
+      "step": 13971
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.13097348380892e-06,
+      "loss": 1.2392,
+      "step": 13972
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1296128825491805e-06,
+      "loss": 0.9309,
+      "step": 13973
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.128252399479414e-06,
+      "loss": 1.0308,
+      "step": 13974
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.12689203463263e-06,
+      "loss": 1.0923,
+      "step": 13975
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1255317880418485e-06,
+      "loss": 1.1177,
+      "step": 13976
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.124171659740076e-06,
+      "loss": 1.1086,
+      "step": 13977
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.122811649760318e-06,
+      "loss": 1.1423,
+      "step": 13978
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.121451758135574e-06,
+      "loss": 1.0426,
+      "step": 13979
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.120091984898852e-06,
+      "loss": 1.1121,
+      "step": 13980
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.118732330083147e-06,
+      "loss": 1.0616,
+      "step": 13981
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.117372793721451e-06,
+      "loss": 1.0773,
+      "step": 13982
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.116013375846756e-06,
+      "loss": 1.1714,
+      "step": 13983
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.114654076492057e-06,
+      "loss": 1.035,
+      "step": 13984
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.113294895690335e-06,
+      "loss": 0.9822,
+      "step": 13985
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.111935833474573e-06,
+      "loss": 1.0943,
+      "step": 13986
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.110576889877754e-06,
+      "loss": 1.2251,
+      "step": 13987
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.109218064932857e-06,
+      "loss": 1.071,
+      "step": 13988
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.107859358672853e-06,
+      "loss": 1.1782,
+      "step": 13989
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.106500771130711e-06,
+      "loss": 0.8927,
+      "step": 13990
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1051423023394095e-06,
+      "loss": 1.2104,
+      "step": 13991
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1037839523319086e-06,
+      "loss": 1.27,
+      "step": 13992
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.102425721141171e-06,
+      "loss": 1.1118,
+      "step": 13993
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.101067608800154e-06,
+      "loss": 1.1216,
+      "step": 13994
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.099709615341824e-06,
+      "loss": 1.1096,
+      "step": 13995
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.098351740799129e-06,
+      "loss": 1.1958,
+      "step": 13996
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.096993985205023e-06,
+      "loss": 1.2194,
+      "step": 13997
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.09563634859245e-06,
+      "loss": 0.8454,
+      "step": 13998
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.094278830994362e-06,
+      "loss": 1.1816,
+      "step": 13999
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.092921432443696e-06,
+      "loss": 1.1501,
+      "step": 14000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0915641529734e-06,
+      "loss": 1.0479,
+      "step": 14001
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.090206992616408e-06,
+      "loss": 1.1243,
+      "step": 14002
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.088849951405647e-06,
+      "loss": 1.1633,
+      "step": 14003
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.087493029374061e-06,
+      "loss": 1.0403,
+      "step": 14004
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.08613622655457e-06,
+      "loss": 1.0601,
+      "step": 14005
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.084779542980102e-06,
+      "loss": 1.1723,
+      "step": 14006
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.083422978683576e-06,
+      "loss": 1.1778,
+      "step": 14007
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.082066533697919e-06,
+      "loss": 1.0116,
+      "step": 14008
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.080710208056043e-06,
+      "loss": 1.224,
+      "step": 14009
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.079354001790865e-06,
+      "loss": 1.2351,
+      "step": 14010
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.077997914935289e-06,
+      "loss": 0.9988,
+      "step": 14011
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0766419475222335e-06,
+      "loss": 0.9684,
+      "step": 14012
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.075286099584598e-06,
+      "loss": 1.1118,
+      "step": 14013
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.073930371155286e-06,
+      "loss": 1.046,
+      "step": 14014
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.072574762267193e-06,
+      "loss": 1.1265,
+      "step": 14015
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.071219272953223e-06,
+      "loss": 1.0444,
+      "step": 14016
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.069863903246266e-06,
+      "loss": 1.1767,
+      "step": 14017
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.068508653179214e-06,
+      "loss": 1.0484,
+      "step": 14018
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.06715352278495e-06,
+      "loss": 1.1123,
+      "step": 14019
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.065798512096367e-06,
+      "loss": 1.0941,
+      "step": 14020
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.064443621146342e-06,
+      "loss": 1.2701,
+      "step": 14021
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.063088849967757e-06,
+      "loss": 1.0427,
+      "step": 14022
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.061734198593481e-06,
+      "loss": 1.0991,
+      "step": 14023
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.060379667056399e-06,
+      "loss": 0.9645,
+      "step": 14024
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.059025255389373e-06,
+      "loss": 1.2672,
+      "step": 14025
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.057670963625272e-06,
+      "loss": 1.115,
+      "step": 14026
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.056316791796964e-06,
+      "loss": 1.1595,
+      "step": 14027
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.054962739937308e-06,
+      "loss": 1.0242,
+      "step": 14028
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0536088080791645e-06,
+      "loss": 0.9922,
+      "step": 14029
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.052254996255386e-06,
+      "loss": 1.0214,
+      "step": 14030
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.050901304498829e-06,
+      "loss": 1.0887,
+      "step": 14031
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.049547732842343e-06,
+      "loss": 1.105,
+      "step": 14032
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.048194281318777e-06,
+      "loss": 1.1278,
+      "step": 14033
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.046840949960968e-06,
+      "loss": 1.1749,
+      "step": 14034
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.045487738801766e-06,
+      "loss": 1.1151,
+      "step": 14035
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.044134647874006e-06,
+      "loss": 1.2764,
+      "step": 14036
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.042781677210524e-06,
+      "loss": 1.126,
+      "step": 14037
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.041428826844148e-06,
+      "loss": 1.0953,
+      "step": 14038
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.040076096807715e-06,
+      "loss": 0.9621,
+      "step": 14039
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.038723487134049e-06,
+      "loss": 1.0736,
+      "step": 14040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0373709978559725e-06,
+      "loss": 0.9726,
+      "step": 14041
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.036018629006305e-06,
+      "loss": 1.12,
+      "step": 14042
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.03466638061787e-06,
+      "loss": 0.9505,
+      "step": 14043
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.033314252723478e-06,
+      "loss": 1.0077,
+      "step": 14044
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.031962245355943e-06,
+      "loss": 1.2188,
+      "step": 14045
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.030610358548071e-06,
+      "loss": 1.2228,
+      "step": 14046
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.029258592332676e-06,
+      "loss": 1.2067,
+      "step": 14047
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.027906946742556e-06,
+      "loss": 1.1982,
+      "step": 14048
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.026555421810511e-06,
+      "loss": 1.0652,
+      "step": 14049
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.025204017569336e-06,
+      "loss": 0.9766,
+      "step": 14050
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.023852734051833e-06,
+      "loss": 1.1744,
+      "step": 14051
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0225015712907894e-06,
+      "loss": 1.0943,
+      "step": 14052
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.021150529318994e-06,
+      "loss": 1.0785,
+      "step": 14053
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.019799608169231e-06,
+      "loss": 1.1261,
+      "step": 14054
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.018448807874288e-06,
+      "loss": 1.1962,
+      "step": 14055
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.017098128466942e-06,
+      "loss": 1.0376,
+      "step": 14056
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.015747569979971e-06,
+      "loss": 1.1073,
+      "step": 14057
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.014397132446144e-06,
+      "loss": 0.9702,
+      "step": 14058
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0130468158982405e-06,
+      "loss": 1.2009,
+      "step": 14059
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.011696620369024e-06,
+      "loss": 1.1672,
+      "step": 14060
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.010346545891261e-06,
+      "loss": 1.2499,
+      "step": 14061
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.00899659249771e-06,
+      "loss": 1.2475,
+      "step": 14062
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.007646760221136e-06,
+      "loss": 1.0381,
+      "step": 14063
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.006297049094294e-06,
+      "loss": 1.1453,
+      "step": 14064
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.004947459149937e-06,
+      "loss": 1.2096,
+      "step": 14065
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.003597990420812e-06,
+      "loss": 1.1396,
+      "step": 14066
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.002248642939672e-06,
+      "loss": 1.2206,
+      "step": 14067
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.00089941673926e-06,
+      "loss": 1.117,
+      "step": 14068
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.999550311852318e-06,
+      "loss": 1.1214,
+      "step": 14069
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.998201328311581e-06,
+      "loss": 1.1877,
+      "step": 14070
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.99685246614979e-06,
+      "loss": 1.2706,
+      "step": 14071
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.995503725399677e-06,
+      "loss": 1.2082,
+      "step": 14072
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.994155106093966e-06,
+      "loss": 1.0677,
+      "step": 14073
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.992806608265395e-06,
+      "loss": 1.2127,
+      "step": 14074
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.99145823194668e-06,
+      "loss": 1.1428,
+      "step": 14075
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.990109977170545e-06,
+      "loss": 1.0545,
+      "step": 14076
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.988761843969703e-06,
+      "loss": 1.1665,
+      "step": 14077
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.987413832376877e-06,
+      "loss": 1.1816,
+      "step": 14078
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.986065942424777e-06,
+      "loss": 1.1116,
+      "step": 14079
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.984718174146111e-06,
+      "loss": 1.1126,
+      "step": 14080
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.98337052757358e-06,
+      "loss": 1.0407,
+      "step": 14081
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.982023002739899e-06,
+      "loss": 1.1555,
+      "step": 14082
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9806755996777565e-06,
+      "loss": 1.0161,
+      "step": 14083
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.979328318419862e-06,
+      "loss": 0.9518,
+      "step": 14084
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.977981158998901e-06,
+      "loss": 1.2273,
+      "step": 14085
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.976634121447567e-06,
+      "loss": 0.9706,
+      "step": 14086
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.975287205798551e-06,
+      "loss": 1.1149,
+      "step": 14087
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.973940412084539e-06,
+      "loss": 1.1098,
+      "step": 14088
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.972593740338212e-06,
+      "loss": 0.8797,
+      "step": 14089
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9712471905922465e-06,
+      "loss": 1.1926,
+      "step": 14090
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.969900762879325e-06,
+      "loss": 1.1032,
+      "step": 14091
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.96855445723212e-06,
+      "loss": 1.2817,
+      "step": 14092
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.967208273683302e-06,
+      "loss": 1.1923,
+      "step": 14093
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.965862212265533e-06,
+      "loss": 0.993,
+      "step": 14094
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.964516273011488e-06,
+      "loss": 1.3002,
+      "step": 14095
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.963170455953824e-06,
+      "loss": 1.1773,
+      "step": 14096
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9618247611252e-06,
+      "loss": 1.0518,
+      "step": 14097
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.960479188558268e-06,
+      "loss": 1.1124,
+      "step": 14098
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9591337382856905e-06,
+      "loss": 0.9202,
+      "step": 14099
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.957788410340112e-06,
+      "loss": 1.1645,
+      "step": 14100
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.95644320475418e-06,
+      "loss": 0.9979,
+      "step": 14101
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9550981215605356e-06,
+      "loss": 1.2364,
+      "step": 14102
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.953753160791827e-06,
+      "loss": 1.0347,
+      "step": 14103
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9524083224806875e-06,
+      "loss": 0.9643,
+      "step": 14104
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.951063606659754e-06,
+      "loss": 1.0708,
+      "step": 14105
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.949719013361655e-06,
+      "loss": 1.0782,
+      "step": 14106
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.948374542619026e-06,
+      "loss": 1.0435,
+      "step": 14107
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.947030194464493e-06,
+      "loss": 0.9758,
+      "step": 14108
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.945685968930675e-06,
+      "loss": 1.196,
+      "step": 14109
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.94434186605019e-06,
+      "loss": 1.0097,
+      "step": 14110
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.942997885855666e-06,
+      "loss": 1.0842,
+      "step": 14111
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.94165402837971e-06,
+      "loss": 1.1034,
+      "step": 14112
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.94031029365493e-06,
+      "loss": 1.1945,
+      "step": 14113
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.938966681713944e-06,
+      "loss": 1.0451,
+      "step": 14114
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.937623192589351e-06,
+      "loss": 1.0065,
+      "step": 14115
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.936279826313757e-06,
+      "loss": 1.2243,
+      "step": 14116
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.934936582919756e-06,
+      "loss": 1.2656,
+      "step": 14117
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9335934624399505e-06,
+      "loss": 1.03,
+      "step": 14118
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9322504649069335e-06,
+      "loss": 1.1726,
+      "step": 14119
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9309075903532935e-06,
+      "loss": 1.159,
+      "step": 14120
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.929564838811615e-06,
+      "loss": 1.1095,
+      "step": 14121
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.928222210314489e-06,
+      "loss": 1.1185,
+      "step": 14122
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.926879704894495e-06,
+      "loss": 1.1896,
+      "step": 14123
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9255373225842115e-06,
+      "loss": 1.0622,
+      "step": 14124
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.92419506341621e-06,
+      "loss": 1.1852,
+      "step": 14125
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.92285292742307e-06,
+      "loss": 1.2124,
+      "step": 14126
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.921510914637358e-06,
+      "loss": 1.02,
+      "step": 14127
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.920169025091642e-06,
+      "loss": 1.1041,
+      "step": 14128
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.918827258818479e-06,
+      "loss": 0.8631,
+      "step": 14129
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9174856158504405e-06,
+      "loss": 1.1351,
+      "step": 14130
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.916144096220079e-06,
+      "loss": 1.1664,
+      "step": 14131
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.914802699959949e-06,
+      "loss": 1.0003,
+      "step": 14132
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.913461427102598e-06,
+      "loss": 1.167,
+      "step": 14133
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.912120277680583e-06,
+      "loss": 1.3163,
+      "step": 14134
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.910779251726447e-06,
+      "loss": 1.1336,
+      "step": 14135
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.90943834927273e-06,
+      "loss": 1.22,
+      "step": 14136
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.908097570351971e-06,
+      "loss": 1.1508,
+      "step": 14137
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.906756914996712e-06,
+      "loss": 1.0796,
+      "step": 14138
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.905416383239485e-06,
+      "loss": 1.1264,
+      "step": 14139
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.90407597511282e-06,
+      "loss": 1.1698,
+      "step": 14140
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.90273569064924e-06,
+      "loss": 0.9874,
+      "step": 14141
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.901395529881278e-06,
+      "loss": 1.1549,
+      "step": 14142
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9000554928414525e-06,
+      "loss": 1.142,
+      "step": 14143
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.898715579562281e-06,
+      "loss": 1.1317,
+      "step": 14144
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8973757900762765e-06,
+      "loss": 1.1937,
+      "step": 14145
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.89603612441596e-06,
+      "loss": 1.0326,
+      "step": 14146
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.894696582613835e-06,
+      "loss": 1.2046,
+      "step": 14147
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.89335716470241e-06,
+      "loss": 1.1687,
+      "step": 14148
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8920178707141866e-06,
+      "loss": 1.0569,
+      "step": 14149
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.890678700681669e-06,
+      "loss": 1.13,
+      "step": 14150
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.889339654637355e-06,
+      "loss": 1.0574,
+      "step": 14151
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8880007326137365e-06,
+      "loss": 1.0498,
+      "step": 14152
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.886661934643303e-06,
+      "loss": 1.2722,
+      "step": 14153
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.885323260758551e-06,
+      "loss": 1.2011,
+      "step": 14154
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.883984710991961e-06,
+      "loss": 1.0023,
+      "step": 14155
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.882646285376017e-06,
+      "loss": 1.0631,
+      "step": 14156
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.881307983943193e-06,
+      "loss": 0.9706,
+      "step": 14157
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.879969806725976e-06,
+      "loss": 1.1062,
+      "step": 14158
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.878631753756834e-06,
+      "loss": 1.1015,
+      "step": 14159
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.877293825068233e-06,
+      "loss": 1.2259,
+      "step": 14160
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.875956020692651e-06,
+      "loss": 1.0965,
+      "step": 14161
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.874618340662546e-06,
+      "loss": 1.2226,
+      "step": 14162
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.87328078501038e-06,
+      "loss": 0.9556,
+      "step": 14163
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.87194335376861e-06,
+      "loss": 1.1763,
+      "step": 14164
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8706060469696966e-06,
+      "loss": 1.1102,
+      "step": 14165
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.869268864646086e-06,
+      "loss": 1.2214,
+      "step": 14166
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.867931806830235e-06,
+      "loss": 1.0468,
+      "step": 14167
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.866594873554585e-06,
+      "loss": 1.035,
+      "step": 14168
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.865258064851579e-06,
+      "loss": 1.2377,
+      "step": 14169
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.863921380753661e-06,
+      "loss": 1.1978,
+      "step": 14170
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.862584821293267e-06,
+      "loss": 1.1684,
+      "step": 14171
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.861248386502831e-06,
+      "loss": 1.2008,
+      "step": 14172
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.85991207641478e-06,
+      "loss": 1.0674,
+      "step": 14173
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.858575891061551e-06,
+      "loss": 1.2044,
+      "step": 14174
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.857239830475564e-06,
+      "loss": 1.1141,
+      "step": 14175
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.855903894689242e-06,
+      "loss": 1.0963,
+      "step": 14176
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.854568083735001e-06,
+      "loss": 1.1265,
+      "step": 14177
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.853232397645263e-06,
+      "loss": 1.0801,
+      "step": 14178
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.85189683645244e-06,
+      "loss": 1.1387,
+      "step": 14179
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.85056140018894e-06,
+      "loss": 1.1389,
+      "step": 14180
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.849226088887168e-06,
+      "loss": 1.1124,
+      "step": 14181
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.847890902579533e-06,
+      "loss": 1.1018,
+      "step": 14182
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.846555841298435e-06,
+      "loss": 1.2135,
+      "step": 14183
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.845220905076271e-06,
+      "loss": 1.0029,
+      "step": 14184
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8438860939454325e-06,
+      "loss": 1.1499,
+      "step": 14185
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.842551407938318e-06,
+      "loss": 1.1213,
+      "step": 14186
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.841216847087315e-06,
+      "loss": 1.1297,
+      "step": 14187
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8398824114248055e-06,
+      "loss": 0.9004,
+      "step": 14188
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.838548100983171e-06,
+      "loss": 1.2093,
+      "step": 14189
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8372139157948e-06,
+      "loss": 1.2276,
+      "step": 14190
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.835879855892063e-06,
+      "loss": 1.0953,
+      "step": 14191
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.834545921307335e-06,
+      "loss": 1.0249,
+      "step": 14192
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8332121120729825e-06,
+      "loss": 1.0099,
+      "step": 14193
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.831878428221382e-06,
+      "loss": 1.0806,
+      "step": 14194
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.830544869784891e-06,
+      "loss": 1.1474,
+      "step": 14195
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8292114367958745e-06,
+      "loss": 1.1456,
+      "step": 14196
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8278781292866865e-06,
+      "loss": 1.0921,
+      "step": 14197
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.826544947289689e-06,
+      "loss": 1.1181,
+      "step": 14198
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.82521189083723e-06,
+      "loss": 1.1079,
+      "step": 14199
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.823878959961656e-06,
+      "loss": 1.0524,
+      "step": 14200
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.822546154695322e-06,
+      "loss": 0.9473,
+      "step": 14201
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.821213475070567e-06,
+      "loss": 0.9173,
+      "step": 14202
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.81988092111973e-06,
+      "loss": 1.1747,
+      "step": 14203
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.818548492875145e-06,
+      "loss": 1.1594,
+      "step": 14204
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.817216190369154e-06,
+      "loss": 1.1193,
+      "step": 14205
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.815884013634083e-06,
+      "loss": 1.0133,
+      "step": 14206
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.814551962702263e-06,
+      "loss": 1.0865,
+      "step": 14207
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.813220037606012e-06,
+      "loss": 1.2098,
+      "step": 14208
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.811888238377661e-06,
+      "loss": 1.1436,
+      "step": 14209
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.810556565049524e-06,
+      "loss": 1.1705,
+      "step": 14210
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.809225017653919e-06,
+      "loss": 1.1665,
+      "step": 14211
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.807893596223152e-06,
+      "loss": 1.239,
+      "step": 14212
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.806562300789542e-06,
+      "loss": 1.0728,
+      "step": 14213
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.805231131385393e-06,
+      "loss": 1.1614,
+      "step": 14214
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.803900088043007e-06,
+      "loss": 1.1128,
+      "step": 14215
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.80256917079468e-06,
+      "loss": 1.08,
+      "step": 14216
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.801238379672717e-06,
+      "loss": 1.2248,
+      "step": 14217
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.799907714709411e-06,
+      "loss": 1.1883,
+      "step": 14218
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.798577175937051e-06,
+      "loss": 1.1044,
+      "step": 14219
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.797246763387922e-06,
+      "loss": 0.9603,
+      "step": 14220
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.795916477094319e-06,
+      "loss": 1.2453,
+      "step": 14221
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7945863170885175e-06,
+      "loss": 1.166,
+      "step": 14222
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7932562834027975e-06,
+      "loss": 1.1007,
+      "step": 14223
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.791926376069431e-06,
+      "loss": 0.8608,
+      "step": 14224
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.790596595120699e-06,
+      "loss": 1.2798,
+      "step": 14225
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.789266940588867e-06,
+      "loss": 1.2462,
+      "step": 14226
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.787937412506201e-06,
+      "loss": 0.9294,
+      "step": 14227
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.786608010904964e-06,
+      "loss": 1.1572,
+      "step": 14228
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.785278735817423e-06,
+      "loss": 1.0889,
+      "step": 14229
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.783949587275829e-06,
+      "loss": 1.099,
+      "step": 14230
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.782620565312439e-06,
+      "loss": 1.0549,
+      "step": 14231
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.781291669959502e-06,
+      "loss": 1.1966,
+      "step": 14232
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.77996290124927e-06,
+      "loss": 1.1701,
+      "step": 14233
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.778634259213988e-06,
+      "loss": 1.0366,
+      "step": 14234
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7773057438858975e-06,
+      "loss": 1.1934,
+      "step": 14235
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.775977355297232e-06,
+      "loss": 1.1711,
+      "step": 14236
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.774649093480238e-06,
+      "loss": 0.9776,
+      "step": 14237
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.773320958467143e-06,
+      "loss": 0.9767,
+      "step": 14238
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.771992950290176e-06,
+      "loss": 0.9447,
+      "step": 14239
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.770665068981563e-06,
+      "loss": 1.2139,
+      "step": 14240
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.769337314573533e-06,
+      "loss": 1.2892,
+      "step": 14241
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.768009687098302e-06,
+      "loss": 0.9408,
+      "step": 14242
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.766682186588087e-06,
+      "loss": 1.2045,
+      "step": 14243
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.765354813075109e-06,
+      "loss": 1.1781,
+      "step": 14244
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7640275665915745e-06,
+      "loss": 1.0517,
+      "step": 14245
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.762700447169693e-06,
+      "loss": 1.1741,
+      "step": 14246
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.761373454841665e-06,
+      "loss": 1.2591,
+      "step": 14247
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.760046589639701e-06,
+      "loss": 1.1995,
+      "step": 14248
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.758719851595995e-06,
+      "loss": 0.9225,
+      "step": 14249
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.757393240742742e-06,
+      "loss": 1.3031,
+      "step": 14250
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.75606675711214e-06,
+      "loss": 1.0939,
+      "step": 14251
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.754740400736372e-06,
+      "loss": 1.024,
+      "step": 14252
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.753414171647634e-06,
+      "loss": 1.0609,
+      "step": 14253
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7520880698781026e-06,
+      "loss": 1.1247,
+      "step": 14254
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.75076209545996e-06,
+      "loss": 0.9763,
+      "step": 14255
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.749436248425381e-06,
+      "loss": 1.1935,
+      "step": 14256
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.748110528806547e-06,
+      "loss": 0.908,
+      "step": 14257
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.746784936635624e-06,
+      "loss": 1.2525,
+      "step": 14258
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.745459471944784e-06,
+      "loss": 1.0246,
+      "step": 14259
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.744134134766183e-06,
+      "loss": 1.1305,
+      "step": 14260
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.742808925131994e-06,
+      "loss": 1.1197,
+      "step": 14261
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.741483843074372e-06,
+      "loss": 1.0499,
+      "step": 14262
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.740158888625473e-06,
+      "loss": 1.1546,
+      "step": 14263
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7388340618174444e-06,
+      "loss": 1.1987,
+      "step": 14264
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.737509362682444e-06,
+      "loss": 1.019,
+      "step": 14265
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.736184791252616e-06,
+      "loss": 0.902,
+      "step": 14266
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.734860347560103e-06,
+      "loss": 1.1896,
+      "step": 14267
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.73353603163704e-06,
+      "loss": 1.042,
+      "step": 14268
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.732211843515574e-06,
+      "loss": 1.1008,
+      "step": 14269
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.730887783227834e-06,
+      "loss": 1.131,
+      "step": 14270
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.729563850805953e-06,
+      "loss": 1.0787,
+      "step": 14271
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.728240046282053e-06,
+      "loss": 1.0875,
+      "step": 14272
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.726916369688267e-06,
+      "loss": 1.1296,
+      "step": 14273
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7255928210567125e-06,
+      "loss": 1.2174,
+      "step": 14274
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.72426940041951e-06,
+      "loss": 1.0978,
+      "step": 14275
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.722946107808769e-06,
+      "loss": 1.0832,
+      "step": 14276
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.721622943256612e-06,
+      "loss": 1.0793,
+      "step": 14277
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.720299906795142e-06,
+      "loss": 1.2129,
+      "step": 14278
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.718976998456466e-06,
+      "loss": 1.0258,
+      "step": 14279
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.717654218272684e-06,
+      "loss": 1.1807,
+      "step": 14280
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.716331566275903e-06,
+      "loss": 1.0088,
+      "step": 14281
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.715009042498216e-06,
+      "loss": 1.0826,
+      "step": 14282
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.713686646971712e-06,
+      "loss": 1.1003,
+      "step": 14283
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.712364379728491e-06,
+      "loss": 1.2156,
+      "step": 14284
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.711042240800637e-06,
+      "loss": 1.1197,
+      "step": 14285
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.709720230220233e-06,
+      "loss": 1.1143,
+      "step": 14286
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.708398348019357e-06,
+      "loss": 1.0685,
+      "step": 14287
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7070765942300955e-06,
+      "loss": 1.0922,
+      "step": 14288
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.705754968884518e-06,
+      "loss": 1.2573,
+      "step": 14289
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.704433472014699e-06,
+      "loss": 1.2347,
+      "step": 14290
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.703112103652702e-06,
+      "loss": 1.0429,
+      "step": 14291
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7017908638306e-06,
+      "loss": 1.0949,
+      "step": 14292
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.700469752580451e-06,
+      "loss": 1.0924,
+      "step": 14293
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.699148769934318e-06,
+      "loss": 1.1471,
+      "step": 14294
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.697827915924249e-06,
+      "loss": 1.1421,
+      "step": 14295
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.696507190582309e-06,
+      "loss": 1.1924,
+      "step": 14296
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6951865939405415e-06,
+      "loss": 1.2224,
+      "step": 14297
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.693866126030995e-06,
+      "loss": 1.1459,
+      "step": 14298
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6925457868857096e-06,
+      "loss": 0.9682,
+      "step": 14299
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.691225576536732e-06,
+      "loss": 1.0312,
+      "step": 14300
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.689905495016097e-06,
+      "loss": 1.0262,
+      "step": 14301
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.68858554235584e-06,
+      "loss": 1.0642,
+      "step": 14302
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.687265718587987e-06,
+      "loss": 1.0988,
+      "step": 14303
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.685946023744575e-06,
+      "loss": 1.1046,
+      "step": 14304
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.684626457857625e-06,
+      "loss": 1.0399,
+      "step": 14305
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.683307020959159e-06,
+      "loss": 1.2282,
+      "step": 14306
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6819877130811915e-06,
+      "loss": 1.1856,
+      "step": 14307
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.680668534255747e-06,
+      "loss": 0.9367,
+      "step": 14308
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.679349484514833e-06,
+      "loss": 1.0693,
+      "step": 14309
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.67803056389046e-06,
+      "loss": 1.1222,
+      "step": 14310
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.676711772414631e-06,
+      "loss": 1.1654,
+      "step": 14311
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.675393110119354e-06,
+      "loss": 0.9473,
+      "step": 14312
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.674074577036627e-06,
+      "loss": 1.1494,
+      "step": 14313
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6727561731984485e-06,
+      "loss": 1.2016,
+      "step": 14314
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.671437898636806e-06,
+      "loss": 1.1365,
+      "step": 14315
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6701197533836995e-06,
+      "loss": 1.2041,
+      "step": 14316
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.668801737471111e-06,
+      "loss": 1.0911,
+      "step": 14317
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6674838509310276e-06,
+      "loss": 1.1389,
+      "step": 14318
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.666166093795423e-06,
+      "loss": 1.0319,
+      "step": 14319
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.664848466096286e-06,
+      "loss": 1.0014,
+      "step": 14320
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.663530967865586e-06,
+      "loss": 1.0271,
+      "step": 14321
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.662213599135296e-06,
+      "loss": 1.1448,
+      "step": 14322
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.660896359937381e-06,
+      "loss": 1.0633,
+      "step": 14323
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6595792503038105e-06,
+      "loss": 1.1972,
+      "step": 14324
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.658262270266548e-06,
+      "loss": 1.0298,
+      "step": 14325
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.656945419857551e-06,
+      "loss": 1.1283,
+      "step": 14326
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.655628699108771e-06,
+      "loss": 1.0499,
+      "step": 14327
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6543121080521695e-06,
+      "loss": 0.967,
+      "step": 14328
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6529956467196915e-06,
+      "loss": 1.1369,
+      "step": 14329
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.65167931514328e-06,
+      "loss": 1.0408,
+      "step": 14330
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.650363113354888e-06,
+      "loss": 1.1032,
+      "step": 14331
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.649047041386448e-06,
+      "loss": 1.1548,
+      "step": 14332
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.647731099269898e-06,
+      "loss": 1.0283,
+      "step": 14333
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.646415287037177e-06,
+      "loss": 1.1697,
+      "step": 14334
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.645099604720209e-06,
+      "loss": 1.1455,
+      "step": 14335
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.64378405235093e-06,
+      "loss": 1.0123,
+      "step": 14336
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.642468629961258e-06,
+      "loss": 1.0475,
+      "step": 14337
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.641153337583119e-06,
+      "loss": 1.1642,
+      "step": 14338
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.639838175248422e-06,
+      "loss": 1.205,
+      "step": 14339
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.638523142989094e-06,
+      "loss": 1.0375,
+      "step": 14340
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.637208240837042e-06,
+      "loss": 1.0745,
+      "step": 14341
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6358934688241745e-06,
+      "loss": 1.087,
+      "step": 14342
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.634578826982394e-06,
+      "loss": 1.1585,
+      "step": 14343
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.633264315343609e-06,
+      "loss": 1.0954,
+      "step": 14344
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6319499339397155e-06,
+      "loss": 1.0842,
+      "step": 14345
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.630635682802612e-06,
+      "loss": 1.1489,
+      "step": 14346
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.629321561964185e-06,
+      "loss": 1.1241,
+      "step": 14347
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.628007571456333e-06,
+      "loss": 1.0985,
+      "step": 14348
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.626693711310939e-06,
+      "loss": 1.2476,
+      "step": 14349
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.625379981559885e-06,
+      "loss": 1.1955,
+      "step": 14350
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6240663822350505e-06,
+      "loss": 0.9674,
+      "step": 14351
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.622752913368318e-06,
+      "loss": 1.1467,
+      "step": 14352
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.621439574991559e-06,
+      "loss": 1.1406,
+      "step": 14353
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.620126367136642e-06,
+      "loss": 1.1577,
+      "step": 14354
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6188132898354345e-06,
+      "loss": 0.9831,
+      "step": 14355
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.617500343119806e-06,
+      "loss": 1.0785,
+      "step": 14356
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.616187527021615e-06,
+      "loss": 1.0434,
+      "step": 14357
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.61487484157272e-06,
+      "loss": 1.1325,
+      "step": 14358
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.61356228680497e-06,
+      "loss": 1.2535,
+      "step": 14359
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6122498627502276e-06,
+      "loss": 0.9626,
+      "step": 14360
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6109375694403356e-06,
+      "loss": 1.2726,
+      "step": 14361
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.609625406907141e-06,
+      "loss": 1.0566,
+      "step": 14362
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.60831337518248e-06,
+      "loss": 1.0439,
+      "step": 14363
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.607001474298203e-06,
+      "loss": 1.0156,
+      "step": 14364
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6056897042861385e-06,
+      "loss": 1.1936,
+      "step": 14365
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.60437806517812e-06,
+      "loss": 1.185,
+      "step": 14366
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6030665570059755e-06,
+      "loss": 1.2189,
+      "step": 14367
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.601755179801537e-06,
+      "loss": 1.1679,
+      "step": 14368
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.600443933596625e-06,
+      "loss": 1.1121,
+      "step": 14369
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.599132818423056e-06,
+      "loss": 1.1893,
+      "step": 14370
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.597821834312652e-06,
+      "loss": 1.0919,
+      "step": 14371
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.596510981297226e-06,
+      "loss": 0.9567,
+      "step": 14372
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.595200259408587e-06,
+      "loss": 1.069,
+      "step": 14373
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.59388966867854e-06,
+      "loss": 1.1238,
+      "step": 14374
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.592579209138896e-06,
+      "loss": 1.0052,
+      "step": 14375
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.59126888082145e-06,
+      "loss": 1.0839,
+      "step": 14376
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.589958683758002e-06,
+      "loss": 0.9933,
+      "step": 14377
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.588648617980343e-06,
+      "loss": 1.2486,
+      "step": 14378
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.587338683520273e-06,
+      "loss": 1.2637,
+      "step": 14379
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.586028880409575e-06,
+      "loss": 0.9928,
+      "step": 14380
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.584719208680032e-06,
+      "loss": 1.0025,
+      "step": 14381
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.583409668363425e-06,
+      "loss": 1.0168,
+      "step": 14382
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.582100259491541e-06,
+      "loss": 1.1368,
+      "step": 14383
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.580790982096149e-06,
+      "loss": 1.2006,
+      "step": 14384
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.579481836209023e-06,
+      "loss": 1.1903,
+      "step": 14385
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.578172821861927e-06,
+      "loss": 1.0918,
+      "step": 14386
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.576863939086636e-06,
+      "loss": 1.0609,
+      "step": 14387
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.575555187914908e-06,
+      "loss": 1.1377,
+      "step": 14388
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.574246568378504e-06,
+      "loss": 1.1873,
+      "step": 14389
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.572938080509175e-06,
+      "loss": 1.0729,
+      "step": 14390
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.571629724338681e-06,
+      "loss": 1.0768,
+      "step": 14391
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.57032149989877e-06,
+      "loss": 0.9266,
+      "step": 14392
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.569013407221189e-06,
+      "loss": 1.1453,
+      "step": 14393
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.567705446337676e-06,
+      "loss": 1.0069,
+      "step": 14394
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.566397617279979e-06,
+      "loss": 1.1582,
+      "step": 14395
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.565089920079834e-06,
+      "loss": 1.0013,
+      "step": 14396
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.563782354768972e-06,
+      "loss": 1.1298,
+      "step": 14397
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.562474921379121e-06,
+      "loss": 1.1705,
+      "step": 14398
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.561167619942017e-06,
+      "loss": 1.2473,
+      "step": 14399
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.55986045048938e-06,
+      "loss": 0.9552,
+      "step": 14400
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5585534130529296e-06,
+      "loss": 1.0575,
+      "step": 14401
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.557246507664383e-06,
+      "loss": 1.0166,
+      "step": 14402
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.555939734355459e-06,
+      "loss": 1.0238,
+      "step": 14403
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.554633093157868e-06,
+      "loss": 1.0636,
+      "step": 14404
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.553326584103318e-06,
+      "loss": 1.0565,
+      "step": 14405
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.55202020722351e-06,
+      "loss": 1.0192,
+      "step": 14406
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.550713962550153e-06,
+      "loss": 1.1845,
+      "step": 14407
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.549407850114943e-06,
+      "loss": 1.1619,
+      "step": 14408
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5481018699495735e-06,
+      "loss": 1.0949,
+      "step": 14409
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.546796022085736e-06,
+      "loss": 1.1539,
+      "step": 14410
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.545490306555125e-06,
+      "loss": 1.1206,
+      "step": 14411
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.544184723389423e-06,
+      "loss": 1.0763,
+      "step": 14412
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.542879272620313e-06,
+      "loss": 1.1498,
+      "step": 14413
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.54157395427947e-06,
+      "loss": 1.2064,
+      "step": 14414
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.540268768398578e-06,
+      "loss": 1.0299,
+      "step": 14415
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.538963715009304e-06,
+      "loss": 1.0941,
+      "step": 14416
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.537658794143322e-06,
+      "loss": 1.0694,
+      "step": 14417
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.536354005832296e-06,
+      "loss": 0.9658,
+      "step": 14418
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.535049350107892e-06,
+      "loss": 1.1338,
+      "step": 14419
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.533744827001768e-06,
+      "loss": 1.1477,
+      "step": 14420
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.532440436545583e-06,
+      "loss": 1.2061,
+      "step": 14421
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.531136178770984e-06,
+      "loss": 1.1158,
+      "step": 14422
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.529832053709632e-06,
+      "loss": 1.1727,
+      "step": 14423
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.528528061393167e-06,
+      "loss": 1.1884,
+      "step": 14424
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.527224201853235e-06,
+      "loss": 1.0569,
+      "step": 14425
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.525920475121475e-06,
+      "loss": 0.9236,
+      "step": 14426
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5246168812295286e-06,
+      "loss": 1.1532,
+      "step": 14427
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.523313420209028e-06,
+      "loss": 0.9798,
+      "step": 14428
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.522010092091605e-06,
+      "loss": 1.1279,
+      "step": 14429
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.520706896908882e-06,
+      "loss": 1.0613,
+      "step": 14430
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.519403834692493e-06,
+      "loss": 1.1392,
+      "step": 14431
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.518100905474056e-06,
+      "loss": 1.1296,
+      "step": 14432
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.516798109285186e-06,
+      "loss": 1.1771,
+      "step": 14433
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.515495446157498e-06,
+      "loss": 1.2282,
+      "step": 14434
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.514192916122611e-06,
+      "loss": 1.07,
+      "step": 14435
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.512890519212127e-06,
+      "loss": 1.1093,
+      "step": 14436
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.511588255457655e-06,
+      "loss": 1.1517,
+      "step": 14437
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5102861248907906e-06,
+      "loss": 1.2268,
+      "step": 14438
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5089841275431405e-06,
+      "loss": 0.9039,
+      "step": 14439
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.507682263446297e-06,
+      "loss": 1.0828,
+      "step": 14440
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.506380532631853e-06,
+      "loss": 1.1807,
+      "step": 14441
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.505078935131393e-06,
+      "loss": 1.0659,
+      "step": 14442
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.503777470976511e-06,
+      "loss": 1.1708,
+      "step": 14443
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.502476140198787e-06,
+      "loss": 1.0913,
+      "step": 14444
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.501174942829799e-06,
+      "loss": 1.1394,
+      "step": 14445
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.49987387890112e-06,
+      "loss": 1.2235,
+      "step": 14446
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.498572948444331e-06,
+      "loss": 1.0307,
+      "step": 14447
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.497272151490997e-06,
+      "loss": 1.0919,
+      "step": 14448
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4959714880726845e-06,
+      "loss": 1.1868,
+      "step": 14449
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.494670958220954e-06,
+      "loss": 1.1643,
+      "step": 14450
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.493370561967372e-06,
+      "loss": 1.1059,
+      "step": 14451
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.492070299343492e-06,
+      "loss": 1.3156,
+      "step": 14452
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.490770170380867e-06,
+      "loss": 1.0409,
+      "step": 14453
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.489470175111046e-06,
+      "loss": 0.9449,
+      "step": 14454
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.48817031356558e-06,
+      "loss": 0.9612,
+      "step": 14455
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.486870585776012e-06,
+      "loss": 1.1502,
+      "step": 14456
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.485570991773877e-06,
+      "loss": 1.1168,
+      "step": 14457
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4842715315907205e-06,
+      "loss": 0.9702,
+      "step": 14458
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.482972205258072e-06,
+      "loss": 1.2508,
+      "step": 14459
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4816730128074635e-06,
+      "loss": 1.1954,
+      "step": 14460
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.480373954270419e-06,
+      "loss": 1.1473,
+      "step": 14461
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.47907502967847e-06,
+      "loss": 0.9564,
+      "step": 14462
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.477776239063133e-06,
+      "loss": 1.2531,
+      "step": 14463
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.476477582455926e-06,
+      "loss": 1.1574,
+      "step": 14464
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.475179059888361e-06,
+      "loss": 1.227,
+      "step": 14465
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4738806713919555e-06,
+      "loss": 1.1362,
+      "step": 14466
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.472582416998215e-06,
+      "loss": 1.2052,
+      "step": 14467
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.471284296738643e-06,
+      "loss": 1.0478,
+      "step": 14468
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.469986310644739e-06,
+      "loss": 0.9427,
+      "step": 14469
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.468688458748006e-06,
+      "loss": 1.1315,
+      "step": 14470
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.467390741079939e-06,
+      "loss": 1.158,
+      "step": 14471
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.466093157672027e-06,
+      "loss": 1.0407,
+      "step": 14472
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.464795708555756e-06,
+      "loss": 1.2151,
+      "step": 14473
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.463498393762618e-06,
+      "loss": 1.0063,
+      "step": 14474
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.462201213324092e-06,
+      "loss": 1.2936,
+      "step": 14475
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.460904167271656e-06,
+      "loss": 1.1593,
+      "step": 14476
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4596072556367835e-06,
+      "loss": 1.2096,
+      "step": 14477
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.458310478450951e-06,
+      "loss": 1.1058,
+      "step": 14478
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.457013835745627e-06,
+      "loss": 1.248,
+      "step": 14479
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.455717327552276e-06,
+      "loss": 1.0442,
+      "step": 14480
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.454420953902356e-06,
+      "loss": 1.1296,
+      "step": 14481
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.453124714827334e-06,
+      "loss": 1.2602,
+      "step": 14482
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.451828610358663e-06,
+      "loss": 1.1017,
+      "step": 14483
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4505326405277945e-06,
+      "loss": 1.2393,
+      "step": 14484
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.449236805366175e-06,
+      "loss": 1.109,
+      "step": 14485
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.447941104905258e-06,
+      "loss": 1.0278,
+      "step": 14486
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.446645539176483e-06,
+      "loss": 1.0215,
+      "step": 14487
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.445350108211287e-06,
+      "loss": 1.1089,
+      "step": 14488
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.444054812041107e-06,
+      "loss": 1.2046,
+      "step": 14489
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.442759650697379e-06,
+      "loss": 0.9899,
+      "step": 14490
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.441464624211531e-06,
+      "loss": 1.0422,
+      "step": 14491
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.440169732614991e-06,
+      "loss": 1.0392,
+      "step": 14492
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.438874975939176e-06,
+      "loss": 1.0574,
+      "step": 14493
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.437580354215514e-06,
+      "loss": 1.111,
+      "step": 14494
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.436285867475418e-06,
+      "loss": 1.1029,
+      "step": 14495
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.434991515750303e-06,
+      "loss": 1.0592,
+      "step": 14496
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.433697299071574e-06,
+      "loss": 1.1285,
+      "step": 14497
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.432403217470645e-06,
+      "loss": 1.1558,
+      "step": 14498
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.431109270978916e-06,
+      "loss": 1.1856,
+      "step": 14499
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.429815459627784e-06,
+      "loss": 0.9969,
+      "step": 14500
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.42852178344865e-06,
+      "loss": 1.2034,
+      "step": 14501
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.42722824247291e-06,
+      "loss": 1.0288,
+      "step": 14502
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.425934836731951e-06,
+      "loss": 0.9845,
+      "step": 14503
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.424641566257162e-06,
+      "loss": 1.0613,
+      "step": 14504
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4233484310799215e-06,
+      "loss": 1.1036,
+      "step": 14505
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.422055431231619e-06,
+      "loss": 1.207,
+      "step": 14506
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.420762566743625e-06,
+      "loss": 1.1393,
+      "step": 14507
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4194698376473165e-06,
+      "loss": 1.1465,
+      "step": 14508
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.41817724397406e-06,
+      "loss": 1.195,
+      "step": 14509
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.41688478575523e-06,
+      "loss": 1.2115,
+      "step": 14510
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.415592463022187e-06,
+      "loss": 1.1068,
+      "step": 14511
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.414300275806291e-06,
+      "loss": 1.0481,
+      "step": 14512
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4130082241388974e-06,
+      "loss": 1.0451,
+      "step": 14513
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.411716308051366e-06,
+      "loss": 1.3638,
+      "step": 14514
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.410424527575047e-06,
+      "loss": 1.1681,
+      "step": 14515
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.409132882741285e-06,
+      "loss": 1.1884,
+      "step": 14516
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.407841373581422e-06,
+      "loss": 1.0392,
+      "step": 14517
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.406550000126807e-06,
+      "loss": 1.1109,
+      "step": 14518
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.405258762408774e-06,
+      "loss": 1.2222,
+      "step": 14519
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.403967660458657e-06,
+      "loss": 1.2482,
+      "step": 14520
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.402676694307784e-06,
+      "loss": 1.1872,
+      "step": 14521
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4013858639874905e-06,
+      "loss": 1.1351,
+      "step": 14522
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.400095169529098e-06,
+      "loss": 1.161,
+      "step": 14523
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.398804610963925e-06,
+      "loss": 1.1265,
+      "step": 14524
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.397514188323289e-06,
+      "loss": 1.1453,
+      "step": 14525
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.39622390163851e-06,
+      "loss": 1.0156,
+      "step": 14526
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.394933750940898e-06,
+      "loss": 1.0313,
+      "step": 14527
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.393643736261759e-06,
+      "loss": 0.9036,
+      "step": 14528
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.392353857632395e-06,
+      "loss": 1.1384,
+      "step": 14529
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.391064115084116e-06,
+      "loss": 1.128,
+      "step": 14530
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.389774508648216e-06,
+      "loss": 1.1696,
+      "step": 14531
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.388485038355988e-06,
+      "loss": 1.066,
+      "step": 14532
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3871957042387225e-06,
+      "loss": 1.0926,
+      "step": 14533
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.385906506327714e-06,
+      "loss": 1.1827,
+      "step": 14534
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3846174446542445e-06,
+      "loss": 1.0987,
+      "step": 14535
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.383328519249596e-06,
+      "loss": 0.9017,
+      "step": 14536
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.382039730145041e-06,
+      "loss": 0.9073,
+      "step": 14537
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.380751077371865e-06,
+      "loss": 1.0991,
+      "step": 14538
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.379462560961334e-06,
+      "loss": 1.0813,
+      "step": 14539
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.378174180944714e-06,
+      "loss": 1.1443,
+      "step": 14540
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.376885937353277e-06,
+      "loss": 1.1963,
+      "step": 14541
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.375597830218281e-06,
+      "loss": 1.0677,
+      "step": 14542
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.374309859570986e-06,
+      "loss": 1.2074,
+      "step": 14543
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.37302202544264e-06,
+      "loss": 0.9908,
+      "step": 14544
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.371734327864508e-06,
+      "loss": 0.9969,
+      "step": 14545
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.370446766867829e-06,
+      "loss": 1.0725,
+      "step": 14546
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.369159342483852e-06,
+      "loss": 1.1487,
+      "step": 14547
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.367872054743815e-06,
+      "loss": 1.0269,
+      "step": 14548
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.366584903678962e-06,
+      "loss": 1.1227,
+      "step": 14549
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.365297889320528e-06,
+      "loss": 0.9567,
+      "step": 14550
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.364011011699741e-06,
+      "loss": 1.0872,
+      "step": 14551
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.36272427084783e-06,
+      "loss": 0.9881,
+      "step": 14552
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.361437666796026e-06,
+      "loss": 1.0827,
+      "step": 14553
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.360151199575546e-06,
+      "loss": 1.1707,
+      "step": 14554
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3588648692176115e-06,
+      "loss": 1.0671,
+      "step": 14555
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.357578675753432e-06,
+      "loss": 1.188,
+      "step": 14556
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.356292619214229e-06,
+      "loss": 1.1639,
+      "step": 14557
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.355006699631206e-06,
+      "loss": 1.0292,
+      "step": 14558
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3537209170355685e-06,
+      "loss": 1.0758,
+      "step": 14559
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3524352714585165e-06,
+      "loss": 1.1694,
+      "step": 14560
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.351149762931255e-06,
+      "loss": 1.2306,
+      "step": 14561
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.349864391484976e-06,
+      "loss": 1.0162,
+      "step": 14562
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.348579157150871e-06,
+      "loss": 1.2538,
+      "step": 14563
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.347294059960127e-06,
+      "loss": 1.2517,
+      "step": 14564
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.346009099943936e-06,
+      "loss": 1.1834,
+      "step": 14565
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.344724277133475e-06,
+      "loss": 1.143,
+      "step": 14566
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.343439591559925e-06,
+      "loss": 1.0945,
+      "step": 14567
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3421550432544565e-06,
+      "loss": 1.0291,
+      "step": 14568
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3408706322482505e-06,
+      "loss": 1.0779,
+      "step": 14569
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.339586358572472e-06,
+      "loss": 1.1707,
+      "step": 14570
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.338302222258285e-06,
+      "loss": 1.2224,
+      "step": 14571
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.337018223336849e-06,
+      "loss": 1.1403,
+      "step": 14572
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.33573436183933e-06,
+      "loss": 1.1894,
+      "step": 14573
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3344506377968794e-06,
+      "loss": 1.0931,
+      "step": 14574
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.333167051240652e-06,
+      "loss": 1.0209,
+      "step": 14575
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.33188360220179e-06,
+      "loss": 1.0788,
+      "step": 14576
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.330600290711446e-06,
+      "loss": 1.1149,
+      "step": 14577
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.329317116800762e-06,
+      "loss": 1.1444,
+      "step": 14578
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.328034080500873e-06,
+      "loss": 1.1112,
+      "step": 14579
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3267511818429135e-06,
+      "loss": 1.13,
+      "step": 14580
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.325468420858022e-06,
+      "loss": 1.1106,
+      "step": 14581
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.324185797577324e-06,
+      "loss": 1.0174,
+      "step": 14582
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.32290331203194e-06,
+      "loss": 1.0328,
+      "step": 14583
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.321620964252997e-06,
+      "loss": 1.0592,
+      "step": 14584
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.320338754271618e-06,
+      "loss": 0.9653,
+      "step": 14585
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.319056682118914e-06,
+      "loss": 1.1047,
+      "step": 14586
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.317774747825998e-06,
+      "loss": 1.1289,
+      "step": 14587
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.316492951423973e-06,
+      "loss": 1.1073,
+      "step": 14588
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3152112929439535e-06,
+      "loss": 1.0806,
+      "step": 14589
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.313929772417038e-06,
+      "loss": 1.0728,
+      "step": 14590
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.312648389874325e-06,
+      "loss": 1.0694,
+      "step": 14591
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.311367145346905e-06,
+      "loss": 1.0648,
+      "step": 14592
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3100860388658794e-06,
+      "loss": 1.0933,
+      "step": 14593
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.308805070462332e-06,
+      "loss": 1.0694,
+      "step": 14594
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.307524240167348e-06,
+      "loss": 1.1942,
+      "step": 14595
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.306243548012007e-06,
+      "loss": 1.0204,
+      "step": 14596
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.304962994027393e-06,
+      "loss": 1.0129,
+      "step": 14597
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.303682578244579e-06,
+      "loss": 1.1269,
+      "step": 14598
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.302402300694636e-06,
+      "loss": 1.1605,
+      "step": 14599
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3011221614086295e-06,
+      "loss": 1.079,
+      "step": 14600
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.299842160417633e-06,
+      "loss": 1.1987,
+      "step": 14601
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.298562297752703e-06,
+      "loss": 1.0474,
+      "step": 14602
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.297282573444899e-06,
+      "loss": 1.0935,
+      "step": 14603
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.296002987525272e-06,
+      "loss": 1.1755,
+      "step": 14604
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.294723540024881e-06,
+      "loss": 1.1132,
+      "step": 14605
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2934442309747714e-06,
+      "loss": 1.0237,
+      "step": 14606
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.292165060405988e-06,
+      "loss": 1.1504,
+      "step": 14607
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.290886028349569e-06,
+      "loss": 1.1339,
+      "step": 14608
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.28960713483656e-06,
+      "loss": 0.9578,
+      "step": 14609
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.288328379897991e-06,
+      "loss": 0.8637,
+      "step": 14610
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2870497635648956e-06,
+      "loss": 1.1142,
+      "step": 14611
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.285771285868297e-06,
+      "loss": 1.1149,
+      "step": 14612
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.284492946839229e-06,
+      "loss": 1.241,
+      "step": 14613
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.283214746508708e-06,
+      "loss": 1.1804,
+      "step": 14614
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.281936684907753e-06,
+      "loss": 1.1545,
+      "step": 14615
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.280658762067374e-06,
+      "loss": 1.1206,
+      "step": 14616
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.279380978018591e-06,
+      "loss": 0.9624,
+      "step": 14617
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.278103332792408e-06,
+      "loss": 1.122,
+      "step": 14618
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.276825826419829e-06,
+      "loss": 1.063,
+      "step": 14619
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.275548458931853e-06,
+      "loss": 1.1534,
+      "step": 14620
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.274271230359485e-06,
+      "loss": 1.1791,
+      "step": 14621
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.272994140733714e-06,
+      "loss": 1.0305,
+      "step": 14622
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.271717190085533e-06,
+      "loss": 1.1036,
+      "step": 14623
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.270440378445927e-06,
+      "loss": 1.1894,
+      "step": 14624
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.269163705845886e-06,
+      "loss": 1.1966,
+      "step": 14625
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.267887172316389e-06,
+      "loss": 1.0282,
+      "step": 14626
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.266610777888407e-06,
+      "loss": 1.2258,
+      "step": 14627
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.265334522592926e-06,
+      "loss": 1.0458,
+      "step": 14628
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.264058406460909e-06,
+      "loss": 1.1239,
+      "step": 14629
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.262782429523328e-06,
+      "loss": 1.0831,
+      "step": 14630
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2615065918111396e-06,
+      "loss": 1.1172,
+      "step": 14631
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.260230893355314e-06,
+      "loss": 1.203,
+      "step": 14632
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.258955334186804e-06,
+      "loss": 0.985,
+      "step": 14633
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.257679914336566e-06,
+      "loss": 1.1219,
+      "step": 14634
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2564046338355436e-06,
+      "loss": 1.0648,
+      "step": 14635
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2551294927146925e-06,
+      "loss": 1.2034,
+      "step": 14636
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.253854491004955e-06,
+      "loss": 1.0787,
+      "step": 14637
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.25257962873727e-06,
+      "loss": 0.9351,
+      "step": 14638
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.25130490594257e-06,
+      "loss": 1.0387,
+      "step": 14639
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.250030322651799e-06,
+      "loss": 1.1504,
+      "step": 14640
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2487558788958815e-06,
+      "loss": 1.0493,
+      "step": 14641
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.247481574705744e-06,
+      "loss": 1.143,
+      "step": 14642
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.246207410112309e-06,
+      "loss": 1.0472,
+      "step": 14643
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.244933385146501e-06,
+      "loss": 0.8866,
+      "step": 14644
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.243659499839236e-06,
+      "loss": 1.0369,
+      "step": 14645
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2423857542214254e-06,
+      "loss": 1.0616,
+      "step": 14646
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.241112148323977e-06,
+      "loss": 1.2399,
+      "step": 14647
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.239838682177804e-06,
+      "loss": 1.0839,
+      "step": 14648
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.238565355813805e-06,
+      "loss": 1.052,
+      "step": 14649
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2372921692628825e-06,
+      "loss": 1.0299,
+      "step": 14650
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2360191225559275e-06,
+      "loss": 1.067,
+      "step": 14651
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.23474621572384e-06,
+      "loss": 1.0711,
+      "step": 14652
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.233473448797508e-06,
+      "loss": 1.113,
+      "step": 14653
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2322008218078156e-06,
+      "loss": 1.1738,
+      "step": 14654
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.230928334785644e-06,
+      "loss": 1.1635,
+      "step": 14655
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.229655987761879e-06,
+      "loss": 1.1853,
+      "step": 14656
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2283837807673924e-06,
+      "loss": 1.0664,
+      "step": 14657
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.227111713833059e-06,
+      "loss": 1.0434,
+      "step": 14658
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2258397869897414e-06,
+      "loss": 1.0895,
+      "step": 14659
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2245680002683165e-06,
+      "loss": 1.1631,
+      "step": 14660
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.22329635369964e-06,
+      "loss": 1.1598,
+      "step": 14661
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.222024847314573e-06,
+      "loss": 1.1487,
+      "step": 14662
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.220753481143966e-06,
+      "loss": 1.145,
+      "step": 14663
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.219482255218681e-06,
+      "loss": 1.0002,
+      "step": 14664
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.218211169569562e-06,
+      "loss": 1.1941,
+      "step": 14665
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.216940224227449e-06,
+      "loss": 1.0943,
+      "step": 14666
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.215669419223189e-06,
+      "loss": 0.997,
+      "step": 14667
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.214398754587627e-06,
+      "loss": 1.1068,
+      "step": 14668
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.213128230351591e-06,
+      "loss": 1.3042,
+      "step": 14669
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.211857846545914e-06,
+      "loss": 1.1126,
+      "step": 14670
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.210587603201421e-06,
+      "loss": 1.0832,
+      "step": 14671
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.209317500348946e-06,
+      "loss": 1.1167,
+      "step": 14672
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.208047538019304e-06,
+      "loss": 1.1266,
+      "step": 14673
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2067777162433155e-06,
+      "loss": 0.9715,
+      "step": 14674
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.20550803505179e-06,
+      "loss": 1.1315,
+      "step": 14675
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2042384944755484e-06,
+      "loss": 1.0639,
+      "step": 14676
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.202969094545393e-06,
+      "loss": 1.1516,
+      "step": 14677
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.20169983529213e-06,
+      "loss": 1.0458,
+      "step": 14678
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.200430716746555e-06,
+      "loss": 1.0289,
+      "step": 14679
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.199161738939474e-06,
+      "loss": 1.0495,
+      "step": 14680
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.19789290190168e-06,
+      "loss": 1.0418,
+      "step": 14681
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1966242056639605e-06,
+      "loss": 0.9615,
+      "step": 14682
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.195355650257101e-06,
+      "loss": 1.114,
+      "step": 14683
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.194087235711892e-06,
+      "loss": 1.0772,
+      "step": 14684
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.192818962059112e-06,
+      "loss": 0.9614,
+      "step": 14685
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.191550829329537e-06,
+      "loss": 1.1935,
+      "step": 14686
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.190282837553938e-06,
+      "loss": 1.0597,
+      "step": 14687
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.189014986763093e-06,
+      "loss": 1.0365,
+      "step": 14688
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.187747276987764e-06,
+      "loss": 1.1301,
+      "step": 14689
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.186479708258716e-06,
+      "loss": 0.9435,
+      "step": 14690
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.185212280606704e-06,
+      "loss": 1.0218,
+      "step": 14691
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1839449940624935e-06,
+      "loss": 1.0865,
+      "step": 14692
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.182677848656833e-06,
+      "loss": 0.9684,
+      "step": 14693
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.181410844420473e-06,
+      "loss": 0.919,
+      "step": 14694
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.180143981384157e-06,
+      "loss": 1.0348,
+      "step": 14695
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.178877259578634e-06,
+      "loss": 1.2135,
+      "step": 14696
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.177610679034642e-06,
+      "loss": 1.0517,
+      "step": 14697
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1763442397829145e-06,
+      "loss": 0.936,
+      "step": 14698
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.175077941854182e-06,
+      "loss": 1.1919,
+      "step": 14699
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.173811785279182e-06,
+      "loss": 1.0303,
+      "step": 14700
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1725457700886345e-06,
+      "loss": 1.2168,
+      "step": 14701
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.171279896313264e-06,
+      "loss": 1.0459,
+      "step": 14702
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.170014163983784e-06,
+      "loss": 1.1311,
+      "step": 14703
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1687485731309194e-06,
+      "loss": 1.1598,
+      "step": 14704
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.167483123785378e-06,
+      "loss": 1.1494,
+      "step": 14705
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.166217815977867e-06,
+      "loss": 1.1402,
+      "step": 14706
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.164952649739091e-06,
+      "loss": 1.0678,
+      "step": 14707
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.163687625099757e-06,
+      "loss": 1.1024,
+      "step": 14708
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.16242274209056e-06,
+      "loss": 0.9918,
+      "step": 14709
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.161158000742197e-06,
+      "loss": 1.0972,
+      "step": 14710
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.159893401085353e-06,
+      "loss": 1.1523,
+      "step": 14711
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.158628943150724e-06,
+      "loss": 1.109,
+      "step": 14712
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1573646269689926e-06,
+      "loss": 1.1321,
+      "step": 14713
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.156100452570836e-06,
+      "loss": 1.0423,
+      "step": 14714
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.154836419986939e-06,
+      "loss": 1.0702,
+      "step": 14715
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.15357252924797e-06,
+      "loss": 1.1235,
+      "step": 14716
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.152308780384603e-06,
+      "loss": 1.1288,
+      "step": 14717
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.151045173427502e-06,
+      "loss": 1.1898,
+      "step": 14718
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.149781708407335e-06,
+      "loss": 1.0912,
+      "step": 14719
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.148518385354763e-06,
+      "loss": 0.9819,
+      "step": 14720
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.14725520430044e-06,
+      "loss": 1.2726,
+      "step": 14721
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.145992165275018e-06,
+      "loss": 1.1052,
+      "step": 14722
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.144729268309154e-06,
+      "loss": 1.0448,
+      "step": 14723
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.14346651343349e-06,
+      "loss": 1.1626,
+      "step": 14724
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.14220390067867e-06,
+      "loss": 1.2298,
+      "step": 14725
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.140941430075331e-06,
+      "loss": 1.2225,
+      "step": 14726
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.139679101654116e-06,
+      "loss": 1.1943,
+      "step": 14727
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.138416915445656e-06,
+      "loss": 1.0441,
+      "step": 14728
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.137154871480578e-06,
+      "loss": 1.1539,
+      "step": 14729
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.135892969789506e-06,
+      "loss": 1.1179,
+      "step": 14730
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.134631210403068e-06,
+      "loss": 1.1788,
+      "step": 14731
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.133369593351884e-06,
+      "loss": 1.0674,
+      "step": 14732
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.132108118666565e-06,
+      "loss": 1.1561,
+      "step": 14733
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.130846786377721e-06,
+      "loss": 1.1125,
+      "step": 14734
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.129585596515969e-06,
+      "loss": 1.1865,
+      "step": 14735
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.128324549111911e-06,
+      "loss": 1.0838,
+      "step": 14736
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.127063644196148e-06,
+      "loss": 1.2366,
+      "step": 14737
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.125802881799275e-06,
+      "loss": 1.2466,
+      "step": 14738
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.124542261951894e-06,
+      "loss": 1.0989,
+      "step": 14739
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.123281784684593e-06,
+      "loss": 1.0674,
+      "step": 14740
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.122021450027961e-06,
+      "loss": 1.1091,
+      "step": 14741
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.120761258012578e-06,
+      "loss": 1.154,
+      "step": 14742
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.119501208669033e-06,
+      "loss": 0.9421,
+      "step": 14743
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.118241302027899e-06,
+      "loss": 1.0959,
+      "step": 14744
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.116981538119751e-06,
+      "loss": 1.0547,
+      "step": 14745
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.115721916975156e-06,
+      "loss": 1.0152,
+      "step": 14746
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.114462438624689e-06,
+      "loss": 1.078,
+      "step": 14747
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.11320310309891e-06,
+      "loss": 1.1101,
+      "step": 14748
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.11194391042838e-06,
+      "loss": 1.0181,
+      "step": 14749
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.11068486064365e-06,
+      "loss": 1.0954,
+      "step": 14750
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.109425953775279e-06,
+      "loss": 0.9421,
+      "step": 14751
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.108167189853821e-06,
+      "loss": 1.0001,
+      "step": 14752
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.106908568909818e-06,
+      "loss": 0.9871,
+      "step": 14753
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.105650090973809e-06,
+      "loss": 1.069,
+      "step": 14754
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.10439175607634e-06,
+      "loss": 1.142,
+      "step": 14755
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.103133564247947e-06,
+      "loss": 1.1067,
+      "step": 14756
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.101875515519158e-06,
+      "loss": 1.2092,
+      "step": 14757
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.100617609920503e-06,
+      "loss": 1.0971,
+      "step": 14758
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.099359847482511e-06,
+      "loss": 1.0928,
+      "step": 14759
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0981022282357045e-06,
+      "loss": 0.947,
+      "step": 14760
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.096844752210598e-06,
+      "loss": 1.0577,
+      "step": 14761
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.095587419437706e-06,
+      "loss": 1.1038,
+      "step": 14762
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.094330229947546e-06,
+      "loss": 1.002,
+      "step": 14763
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.093073183770625e-06,
+      "loss": 1.0192,
+      "step": 14764
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.091816280937445e-06,
+      "loss": 1.1786,
+      "step": 14765
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.090559521478505e-06,
+      "loss": 1.0137,
+      "step": 14766
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.089302905424311e-06,
+      "loss": 1.1412,
+      "step": 14767
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.088046432805352e-06,
+      "loss": 1.1313,
+      "step": 14768
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0867901036521186e-06,
+      "loss": 1.1103,
+      "step": 14769
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.085533917995097e-06,
+      "loss": 1.0528,
+      "step": 14770
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.084277875864776e-06,
+      "loss": 1.1022,
+      "step": 14771
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0830219772916336e-06,
+      "loss": 0.969,
+      "step": 14772
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0817662223061475e-06,
+      "loss": 0.9939,
+      "step": 14773
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0805106109387856e-06,
+      "loss": 1.0813,
+      "step": 14774
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.079255143220027e-06,
+      "loss": 1.1629,
+      "step": 14775
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.077999819180334e-06,
+      "loss": 1.1086,
+      "step": 14776
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0767446388501685e-06,
+      "loss": 1.0242,
+      "step": 14777
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.075489602259988e-06,
+      "loss": 1.1228,
+      "step": 14778
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.074234709440254e-06,
+      "loss": 1.1991,
+      "step": 14779
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.072979960421418e-06,
+      "loss": 1.0813,
+      "step": 14780
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.071725355233928e-06,
+      "loss": 1.0205,
+      "step": 14781
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.070470893908224e-06,
+      "loss": 1.0803,
+      "step": 14782
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0692165764747585e-06,
+      "loss": 1.2553,
+      "step": 14783
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.067962402963963e-06,
+      "loss": 1.1277,
+      "step": 14784
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.066708373406275e-06,
+      "loss": 1.1679,
+      "step": 14785
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0654544878321235e-06,
+      "loss": 1.1882,
+      "step": 14786
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.06420074627194e-06,
+      "loss": 1.0379,
+      "step": 14787
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.06294714875615e-06,
+      "loss": 1.13,
+      "step": 14788
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.061693695315171e-06,
+      "loss": 1.0691,
+      "step": 14789
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0604403859794195e-06,
+      "loss": 1.0822,
+      "step": 14790
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.059187220779315e-06,
+      "loss": 1.0777,
+      "step": 14791
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.057934199745266e-06,
+      "loss": 1.1253,
+      "step": 14792
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.056681322907678e-06,
+      "loss": 1.1736,
+      "step": 14793
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.055428590296953e-06,
+      "loss": 1.1925,
+      "step": 14794
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.054176001943497e-06,
+      "loss": 1.1565,
+      "step": 14795
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.052923557877704e-06,
+      "loss": 0.9755,
+      "step": 14796
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.051671258129966e-06,
+      "loss": 1.2191,
+      "step": 14797
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0504191027306695e-06,
+      "loss": 1.0212,
+      "step": 14798
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.049167091710208e-06,
+      "loss": 1.1137,
+      "step": 14799
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.047915225098959e-06,
+      "loss": 1.128,
+      "step": 14800
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.046663502927302e-06,
+      "loss": 1.0384,
+      "step": 14801
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.045411925225615e-06,
+      "loss": 1.1867,
+      "step": 14802
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0441604920242705e-06,
+      "loss": 1.1784,
+      "step": 14803
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.042909203353634e-06,
+      "loss": 1.0668,
+      "step": 14804
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.041658059244069e-06,
+      "loss": 1.1395,
+      "step": 14805
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0404070597259434e-06,
+      "loss": 1.0699,
+      "step": 14806
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.039156204829613e-06,
+      "loss": 1.216,
+      "step": 14807
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.03790549458543e-06,
+      "loss": 1.0482,
+      "step": 14808
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.036654929023743e-06,
+      "loss": 0.9225,
+      "step": 14809
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.035404508174908e-06,
+      "loss": 1.2243,
+      "step": 14810
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.034154232069264e-06,
+      "loss": 1.1048,
+      "step": 14811
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.032904100737152e-06,
+      "loss": 1.1834,
+      "step": 14812
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.031654114208905e-06,
+      "loss": 1.2324,
+      "step": 14813
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.030404272514864e-06,
+      "loss": 0.9688,
+      "step": 14814
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.029154575685354e-06,
+      "loss": 1.0993,
+      "step": 14815
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.027905023750706e-06,
+      "loss": 1.0753,
+      "step": 14816
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.026655616741233e-06,
+      "loss": 1.2349,
+      "step": 14817
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.025406354687267e-06,
+      "loss": 1.102,
+      "step": 14818
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.024157237619118e-06,
+      "loss": 1.2502,
+      "step": 14819
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.022908265567097e-06,
+      "loss": 1.1097,
+      "step": 14820
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.021659438561511e-06,
+      "loss": 1.2968,
+      "step": 14821
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.020410756632674e-06,
+      "loss": 0.7953,
+      "step": 14822
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0191622198108806e-06,
+      "loss": 1.1583,
+      "step": 14823
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.017913828126432e-06,
+      "loss": 1.1471,
+      "step": 14824
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.016665581609619e-06,
+      "loss": 1.1384,
+      "step": 14825
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0154174802907386e-06,
+      "loss": 1.0566,
+      "step": 14826
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.014169524200076e-06,
+      "loss": 1.1983,
+      "step": 14827
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.012921713367917e-06,
+      "loss": 1.0947,
+      "step": 14828
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.011674047824535e-06,
+      "loss": 1.1987,
+      "step": 14829
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.010426527600218e-06,
+      "loss": 1.1285,
+      "step": 14830
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.009179152725233e-06,
+      "loss": 1.2388,
+      "step": 14831
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.007931923229853e-06,
+      "loss": 0.9859,
+      "step": 14832
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.006684839144339e-06,
+      "loss": 1.1975,
+      "step": 14833
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.005437900498959e-06,
+      "loss": 1.2277,
+      "step": 14834
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.004191107323974e-06,
+      "loss": 1.0978,
+      "step": 14835
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.002944459649638e-06,
+      "loss": 1.1059,
+      "step": 14836
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0016979575062e-06,
+      "loss": 1.1138,
+      "step": 14837
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.000451600923915e-06,
+      "loss": 1.1248,
+      "step": 14838
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.999205389933028e-06,
+      "loss": 1.1797,
+      "step": 14839
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.997959324563776e-06,
+      "loss": 0.9794,
+      "step": 14840
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.996713404846397e-06,
+      "loss": 0.906,
+      "step": 14841
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.995467630811133e-06,
+      "loss": 1.041,
+      "step": 14842
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.994222002488209e-06,
+      "loss": 0.9763,
+      "step": 14843
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9929765199078565e-06,
+      "loss": 1.0373,
+      "step": 14844
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.991731183100293e-06,
+      "loss": 0.9845,
+      "step": 14845
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.99048599209575e-06,
+      "loss": 0.9402,
+      "step": 14846
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9892409469244365e-06,
+      "loss": 1.043,
+      "step": 14847
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.987996047616568e-06,
+      "loss": 1.0438,
+      "step": 14848
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.986751294202352e-06,
+      "loss": 1.1475,
+      "step": 14849
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.985506686712001e-06,
+      "loss": 1.1241,
+      "step": 14850
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.984262225175714e-06,
+      "loss": 1.0851,
+      "step": 14851
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9830179096236935e-06,
+      "loss": 1.1384,
+      "step": 14852
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.981773740086128e-06,
+      "loss": 1.1912,
+      "step": 14853
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.980529716593218e-06,
+      "loss": 1.2077,
+      "step": 14854
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.979285839175149e-06,
+      "loss": 1.1267,
+      "step": 14855
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.978042107862107e-06,
+      "loss": 0.9853,
+      "step": 14856
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.97679852268427e-06,
+      "loss": 1.1712,
+      "step": 14857
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.975555083671821e-06,
+      "loss": 1.0429,
+      "step": 14858
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.974311790854934e-06,
+      "loss": 1.1531,
+      "step": 14859
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.97306864426378e-06,
+      "loss": 1.2086,
+      "step": 14860
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.971825643928521e-06,
+      "loss": 1.0582,
+      "step": 14861
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.970582789879329e-06,
+      "loss": 1.2464,
+      "step": 14862
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.96934008214636e-06,
+      "loss": 1.1988,
+      "step": 14863
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.968097520759773e-06,
+      "loss": 1.1051,
+      "step": 14864
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.966855105749716e-06,
+      "loss": 1.0989,
+      "step": 14865
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.965612837146346e-06,
+      "loss": 1.0866,
+      "step": 14866
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.964370714979806e-06,
+      "loss": 1.1234,
+      "step": 14867
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.963128739280239e-06,
+      "loss": 1.0778,
+      "step": 14868
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.961886910077779e-06,
+      "loss": 1.1538,
+      "step": 14869
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.960645227402571e-06,
+      "loss": 1.2043,
+      "step": 14870
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.959403691284741e-06,
+      "loss": 1.1603,
+      "step": 14871
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.95816230175442e-06,
+      "loss": 0.9918,
+      "step": 14872
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.956921058841727e-06,
+      "loss": 1.1524,
+      "step": 14873
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.955679962576793e-06,
+      "loss": 1.1112,
+      "step": 14874
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9544390129897295e-06,
+      "loss": 1.1715,
+      "step": 14875
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.953198210110651e-06,
+      "loss": 1.0404,
+      "step": 14876
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9519575539696665e-06,
+      "loss": 1.1712,
+      "step": 14877
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.950717044596889e-06,
+      "loss": 1.1586,
+      "step": 14878
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9494766820224175e-06,
+      "loss": 0.9691,
+      "step": 14879
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.948236466276354e-06,
+      "loss": 1.236,
+      "step": 14880
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.946996397388791e-06,
+      "loss": 1.0659,
+      "step": 14881
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.945756475389826e-06,
+      "loss": 1.1396,
+      "step": 14882
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.944516700309547e-06,
+      "loss": 0.9654,
+      "step": 14883
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.943277072178035e-06,
+      "loss": 1.1695,
+      "step": 14884
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9420375910253816e-06,
+      "loss": 1.1048,
+      "step": 14885
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.940798256881658e-06,
+      "loss": 1.311,
+      "step": 14886
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.939559069776943e-06,
+      "loss": 1.1917,
+      "step": 14887
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.938320029741302e-06,
+      "loss": 1.0348,
+      "step": 14888
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.937081136804811e-06,
+      "loss": 1.0979,
+      "step": 14889
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.93584239099753e-06,
+      "loss": 1.1766,
+      "step": 14890
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.93460379234952e-06,
+      "loss": 1.1553,
+      "step": 14891
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.933365340890834e-06,
+      "loss": 1.0032,
+      "step": 14892
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9321270366515355e-06,
+      "loss": 1.1442,
+      "step": 14893
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.930888879661667e-06,
+      "loss": 1.1854,
+      "step": 14894
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.929650869951278e-06,
+      "loss": 1.0235,
+      "step": 14895
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.928413007550406e-06,
+      "loss": 1.1289,
+      "step": 14896
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.927175292489098e-06,
+      "loss": 1.0975,
+      "step": 14897
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.925937724797386e-06,
+      "loss": 1.1216,
+      "step": 14898
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.924700304505302e-06,
+      "loss": 1.098,
+      "step": 14899
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.923463031642873e-06,
+      "loss": 1.1152,
+      "step": 14900
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.922225906240127e-06,
+      "loss": 1.1195,
+      "step": 14901
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.920988928327085e-06,
+      "loss": 1.0586,
+      "step": 14902
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.919752097933764e-06,
+      "loss": 0.9783,
+      "step": 14903
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.918515415090175e-06,
+      "loss": 1.2533,
+      "step": 14904
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.917278879826335e-06,
+      "loss": 1.1884,
+      "step": 14905
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.916042492172248e-06,
+      "loss": 1.1078,
+      "step": 14906
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.914806252157918e-06,
+      "loss": 1.1434,
+      "step": 14907
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9135701598133405e-06,
+      "loss": 1.1091,
+      "step": 14908
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.912334215168519e-06,
+      "loss": 1.1411,
+      "step": 14909
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.911098418253443e-06,
+      "loss": 1.2389,
+      "step": 14910
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.909862769098102e-06,
+      "loss": 1.0751,
+      "step": 14911
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.908627267732477e-06,
+      "loss": 1.2564,
+      "step": 14912
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.907391914186557e-06,
+      "loss": 1.1464,
+      "step": 14913
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.906156708490319e-06,
+      "loss": 1.0829,
+      "step": 14914
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.904921650673735e-06,
+      "loss": 1.1192,
+      "step": 14915
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.903686740766774e-06,
+      "loss": 1.2201,
+      "step": 14916
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.902451978799409e-06,
+      "loss": 1.0516,
+      "step": 14917
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.901217364801605e-06,
+      "loss": 1.2366,
+      "step": 14918
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.899982898803319e-06,
+      "loss": 1.1446,
+      "step": 14919
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.898748580834508e-06,
+      "loss": 1.1399,
+      "step": 14920
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.897514410925124e-06,
+      "loss": 1.2271,
+      "step": 14921
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.896280389105121e-06,
+      "loss": 1.1338,
+      "step": 14922
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.895046515404442e-06,
+      "loss": 1.1489,
+      "step": 14923
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.893812789853028e-06,
+      "loss": 1.1577,
+      "step": 14924
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.892579212480823e-06,
+      "loss": 0.8839,
+      "step": 14925
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.891345783317759e-06,
+      "loss": 1.2461,
+      "step": 14926
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.890112502393768e-06,
+      "loss": 1.1174,
+      "step": 14927
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.888879369738774e-06,
+      "loss": 1.1186,
+      "step": 14928
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.88764638538271e-06,
+      "loss": 1.1386,
+      "step": 14929
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.886413549355491e-06,
+      "loss": 1.2336,
+      "step": 14930
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8851808616870365e-06,
+      "loss": 1.0948,
+      "step": 14931
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.883948322407256e-06,
+      "loss": 1.0479,
+      "step": 14932
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.882715931546067e-06,
+      "loss": 1.0384,
+      "step": 14933
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.881483689133372e-06,
+      "loss": 1.0895,
+      "step": 14934
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.880251595199074e-06,
+      "loss": 1.1802,
+      "step": 14935
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.879019649773068e-06,
+      "loss": 1.2267,
+      "step": 14936
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.877787852885258e-06,
+      "loss": 1.0571,
+      "step": 14937
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.876556204565531e-06,
+      "loss": 1.0291,
+      "step": 14938
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8753247048437785e-06,
+      "loss": 1.3024,
+      "step": 14939
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.874093353749878e-06,
+      "loss": 1.0614,
+      "step": 14940
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.872862151313721e-06,
+      "loss": 1.0927,
+      "step": 14941
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.871631097565179e-06,
+      "loss": 1.2031,
+      "step": 14942
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.870400192534128e-06,
+      "loss": 1.0254,
+      "step": 14943
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.869169436250435e-06,
+      "loss": 0.9758,
+      "step": 14944
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.867938828743973e-06,
+      "loss": 1.2743,
+      "step": 14945
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.866708370044602e-06,
+      "loss": 1.0708,
+      "step": 14946
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.865478060182181e-06,
+      "loss": 1.1913,
+      "step": 14947
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.864247899186563e-06,
+      "loss": 1.0865,
+      "step": 14948
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.863017887087606e-06,
+      "loss": 1.099,
+      "step": 14949
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.861788023915159e-06,
+      "loss": 1.143,
+      "step": 14950
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.860558309699064e-06,
+      "loss": 1.0744,
+      "step": 14951
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.859328744469159e-06,
+      "loss": 1.1262,
+      "step": 14952
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.858099328255291e-06,
+      "loss": 1.1671,
+      "step": 14953
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.856870061087289e-06,
+      "loss": 1.0155,
+      "step": 14954
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.855640942994985e-06,
+      "loss": 1.0952,
+      "step": 14955
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8544119740082005e-06,
+      "loss": 1.1212,
+      "step": 14956
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.853183154156769e-06,
+      "loss": 1.1325,
+      "step": 14957
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.851954483470505e-06,
+      "loss": 1.1321,
+      "step": 14958
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.850725961979223e-06,
+      "loss": 0.9343,
+      "step": 14959
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8494975897127365e-06,
+      "loss": 1.1302,
+      "step": 14960
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.848269366700857e-06,
+      "loss": 1.0413,
+      "step": 14961
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.84704129297339e-06,
+      "loss": 1.151,
+      "step": 14962
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.845813368560135e-06,
+      "loss": 1.1591,
+      "step": 14963
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.844585593490887e-06,
+      "loss": 0.9635,
+      "step": 14964
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.843357967795448e-06,
+      "loss": 1.0613,
+      "step": 14965
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.842130491503605e-06,
+      "loss": 1.0542,
+      "step": 14966
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.840903164645145e-06,
+      "loss": 1.1175,
+      "step": 14967
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.83967598724985e-06,
+      "loss": 1.1072,
+      "step": 14968
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8384489593475035e-06,
+      "loss": 1.0917,
+      "step": 14969
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.837222080967882e-06,
+      "loss": 1.1468,
+      "step": 14970
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.835995352140751e-06,
+      "loss": 1.0438,
+      "step": 14971
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8347687728958906e-06,
+      "loss": 1.1699,
+      "step": 14972
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.833542343263059e-06,
+      "loss": 1.1611,
+      "step": 14973
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.832316063272021e-06,
+      "loss": 1.0364,
+      "step": 14974
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.83108993295253e-06,
+      "loss": 1.1094,
+      "step": 14975
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.829863952334348e-06,
+      "loss": 1.1244,
+      "step": 14976
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.828638121447221e-06,
+      "loss": 1.092,
+      "step": 14977
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.827412440320898e-06,
+      "loss": 1.0741,
+      "step": 14978
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.826186908985119e-06,
+      "loss": 1.1057,
+      "step": 14979
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.82496152746963e-06,
+      "loss": 1.0558,
+      "step": 14980
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.823736295804166e-06,
+      "loss": 1.0614,
+      "step": 14981
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.822511214018457e-06,
+      "loss": 1.1054,
+      "step": 14982
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8212862821422305e-06,
+      "loss": 1.1267,
+      "step": 14983
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.820061500205219e-06,
+      "loss": 1.1185,
+      "step": 14984
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.818836868237139e-06,
+      "loss": 1.1923,
+      "step": 14985
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.81761238626771e-06,
+      "loss": 1.2296,
+      "step": 14986
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.816388054326645e-06,
+      "loss": 1.0222,
+      "step": 14987
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.815163872443659e-06,
+      "loss": 1.0916,
+      "step": 14988
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.813939840648457e-06,
+      "loss": 1.0965,
+      "step": 14989
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8127159589707417e-06,
+      "loss": 1.0657,
+      "step": 14990
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.811492227440212e-06,
+      "loss": 1.116,
+      "step": 14991
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.810268646086569e-06,
+      "loss": 1.1966,
+      "step": 14992
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.809045214939502e-06,
+      "loss": 1.1821,
+      "step": 14993
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8078219340287016e-06,
+      "loss": 1.1332,
+      "step": 14994
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8065988033838475e-06,
+      "loss": 1.136,
+      "step": 14995
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8053758230346315e-06,
+      "loss": 1.1219,
+      "step": 14996
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8041529930107255e-06,
+      "loss": 1.2042,
+      "step": 14997
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.802930313341805e-06,
+      "loss": 1.1183,
+      "step": 14998
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.801707784057538e-06,
+      "loss": 1.0644,
+      "step": 14999
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.800485405187594e-06,
+      "loss": 1.1712,
+      "step": 15000
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7992631767616406e-06,
+      "loss": 1.0128,
+      "step": 15001
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.798041098809334e-06,
+      "loss": 1.1828,
+      "step": 15002
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7968191713603308e-06,
+      "loss": 1.1782,
+      "step": 15003
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7955973944442804e-06,
+      "loss": 1.1711,
+      "step": 15004
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7943757680908377e-06,
+      "loss": 1.1129,
+      "step": 15005
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.793154292329646e-06,
+      "loss": 1.1859,
+      "step": 15006
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7919329671903447e-06,
+      "loss": 1.0393,
+      "step": 15007
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7907117927025706e-06,
+      "loss": 1.2516,
+      "step": 15008
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7894907688959625e-06,
+      "loss": 0.9261,
+      "step": 15009
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.788269895800151e-06,
+      "loss": 1.0378,
+      "step": 15010
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7870491734447557e-06,
+      "loss": 0.9554,
+      "step": 15011
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.785828601859409e-06,
+      "loss": 1.0966,
+      "step": 15012
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.784608181073728e-06,
+      "loss": 1.0924,
+      "step": 15013
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7833879111173276e-06,
+      "loss": 1.1461,
+      "step": 15014
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7821677920198164e-06,
+      "loss": 1.13,
+      "step": 15015
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.78094782381081e-06,
+      "loss": 1.1478,
+      "step": 15016
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7797280065199102e-06,
+      "loss": 1.3082,
+      "step": 15017
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7785083401767187e-06,
+      "loss": 1.2054,
+      "step": 15018
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.77728882481083e-06,
+      "loss": 1.0688,
+      "step": 15019
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7760694604518445e-06,
+      "loss": 1.2391,
+      "step": 15020
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7748502471293492e-06,
+      "loss": 1.171,
+      "step": 15021
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.77363118487293e-06,
+      "loss": 0.9953,
+      "step": 15022
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7724122737121692e-06,
+      "loss": 1.0346,
+      "step": 15023
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.77119351367665e-06,
+      "loss": 1.053,
+      "step": 15024
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7699749047959466e-06,
+      "loss": 1.1582,
+      "step": 15025
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7687564470996296e-06,
+      "loss": 1.1095,
+      "step": 15026
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.767538140617265e-06,
+      "loss": 1.141,
+      "step": 15027
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.766319985378425e-06,
+      "loss": 1.224,
+      "step": 15028
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7651019814126656e-06,
+      "loss": 1.123,
+      "step": 15029
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7638841287495455e-06,
+      "loss": 1.1335,
+      "step": 15030
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.762666427418614e-06,
+      "loss": 1.055,
+      "step": 15031
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7614488774494283e-06,
+      "loss": 1.0404,
+      "step": 15032
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.760231478871531e-06,
+      "loss": 1.1772,
+      "step": 15033
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7590142317144663e-06,
+      "loss": 1.1307,
+      "step": 15034
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.757797136007768e-06,
+      "loss": 1.0903,
+      "step": 15035
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7565801917809796e-06,
+      "loss": 1.1319,
+      "step": 15036
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7553633990636274e-06,
+      "loss": 1.2609,
+      "step": 15037
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.754146757885242e-06,
+      "loss": 1.1314,
+      "step": 15038
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7529302682753422e-06,
+      "loss": 1.2025,
+      "step": 15039
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7517139302634564e-06,
+      "loss": 1.084,
+      "step": 15040
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.750497743879098e-06,
+      "loss": 1.0536,
+      "step": 15041
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.749281709151781e-06,
+      "loss": 1.2484,
+      "step": 15042
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7480658261110105e-06,
+      "loss": 1.0849,
+      "step": 15043
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7468500947862987e-06,
+      "loss": 1.1014,
+      "step": 15044
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.745634515207146e-06,
+      "loss": 1.0369,
+      "step": 15045
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7444190874030496e-06,
+      "loss": 1.0924,
+      "step": 15046
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.743203811403502e-06,
+      "loss": 0.9934,
+      "step": 15047
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7419886872380007e-06,
+      "loss": 1.0043,
+      "step": 15048
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.740773714936029e-06,
+      "loss": 1.1777,
+      "step": 15049
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.739558894527072e-06,
+      "loss": 1.1144,
+      "step": 15050
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7383442260406067e-06,
+      "loss": 1.1023,
+      "step": 15051
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7371297095061143e-06,
+      "loss": 1.0714,
+      "step": 15052
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7359153449530673e-06,
+      "loss": 0.8436,
+      "step": 15053
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7347011324109307e-06,
+      "loss": 1.0336,
+      "step": 15054
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7334870719091697e-06,
+      "loss": 0.945,
+      "step": 15055
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7322731634772514e-06,
+      "loss": 1.0556,
+      "step": 15056
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7310594071446317e-06,
+      "loss": 1.1443,
+      "step": 15057
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.729845802940759e-06,
+      "loss": 0.9818,
+      "step": 15058
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7286323508950928e-06,
+      "loss": 1.1569,
+      "step": 15059
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7274190510370766e-06,
+      "loss": 1.0064,
+      "step": 15060
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.726205903396153e-06,
+      "loss": 0.9504,
+      "step": 15061
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7249929080017576e-06,
+      "loss": 1.1521,
+      "step": 15062
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7237800648833333e-06,
+      "loss": 1.1175,
+      "step": 15063
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7225673740703094e-06,
+      "loss": 1.0842,
+      "step": 15064
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.721354835592115e-06,
+      "loss": 1.0743,
+      "step": 15065
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.720142449478169e-06,
+      "loss": 1.1333,
+      "step": 15066
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7189302157579023e-06,
+      "loss": 1.1813,
+      "step": 15067
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.717718134460726e-06,
+      "loss": 1.1241,
+      "step": 15068
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.716506205616055e-06,
+      "loss": 1.1135,
+      "step": 15069
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7152944292532966e-06,
+      "loss": 1.0601,
+      "step": 15070
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.714082805401863e-06,
+      "loss": 1.155,
+      "step": 15071
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.712871334091154e-06,
+      "loss": 0.9131,
+      "step": 15072
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7116600153505665e-06,
+      "loss": 1.2117,
+      "step": 15073
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7104488492094947e-06,
+      "loss": 1.0398,
+      "step": 15074
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7092378356973357e-06,
+      "loss": 0.981,
+      "step": 15075
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.708026974843474e-06,
+      "loss": 0.9642,
+      "step": 15076
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7068162666772945e-06,
+      "loss": 1.0925,
+      "step": 15077
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.705605711228172e-06,
+      "loss": 1.0886,
+      "step": 15078
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.704395308525491e-06,
+      "loss": 1.0743,
+      "step": 15079
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7031850585986217e-06,
+      "loss": 0.9949,
+      "step": 15080
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.701974961476933e-06,
+      "loss": 1.2378,
+      "step": 15081
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7007650171897856e-06,
+      "loss": 1.1254,
+      "step": 15082
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.699555225766547e-06,
+      "loss": 1.039,
+      "step": 15083
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.698345587236577e-06,
+      "loss": 1.2827,
+      "step": 15084
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6971361016292274e-06,
+      "loss": 1.0473,
+      "step": 15085
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6959267689738488e-06,
+      "loss": 1.1756,
+      "step": 15086
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6947175892997846e-06,
+      "loss": 1.132,
+      "step": 15087
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6935085626363833e-06,
+      "loss": 0.9364,
+      "step": 15088
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6922996890129836e-06,
+      "loss": 1.1056,
+      "step": 15089
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6910909684589203e-06,
+      "loss": 1.0086,
+      "step": 15090
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6898824010035216e-06,
+      "loss": 1.0658,
+      "step": 15091
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6886739866761235e-06,
+      "loss": 1.0918,
+      "step": 15092
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6874657255060466e-06,
+      "loss": 1.1291,
+      "step": 15093
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6862576175226118e-06,
+      "loss": 1.0728,
+      "step": 15094
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6850496627551322e-06,
+      "loss": 1.0277,
+      "step": 15095
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6838418612329306e-06,
+      "loss": 1.3211,
+      "step": 15096
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6826342129853108e-06,
+      "loss": 1.0482,
+      "step": 15097
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6814267180415776e-06,
+      "loss": 1.0065,
+      "step": 15098
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.680219376431039e-06,
+      "loss": 1.2294,
+      "step": 15099
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.67901218818299e-06,
+      "loss": 1.0332,
+      "step": 15100
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6778051533267257e-06,
+      "loss": 1.1646,
+      "step": 15101
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.676598271891534e-06,
+      "loss": 1.0786,
+      "step": 15102
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.675391543906709e-06,
+      "loss": 1.0305,
+      "step": 15103
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.674184969401532e-06,
+      "loss": 1.0176,
+      "step": 15104
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6729785484052815e-06,
+      "loss": 1.0334,
+      "step": 15105
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.671772280947231e-06,
+      "loss": 1.204,
+      "step": 15106
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6705661670566604e-06,
+      "loss": 0.916,
+      "step": 15107
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.669360206762835e-06,
+      "loss": 1.2284,
+      "step": 15108
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.66815440009502e-06,
+      "loss": 1.0737,
+      "step": 15109
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6669487470824717e-06,
+      "loss": 1.1171,
+      "step": 15110
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6657432477544562e-06,
+      "loss": 1.132,
+      "step": 15111
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.664537902140225e-06,
+      "loss": 1.0224,
+      "step": 15112
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6633327102690253e-06,
+      "loss": 1.1212,
+      "step": 15113
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6621276721701025e-06,
+      "loss": 0.9256,
+      "step": 15114
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6609227878727062e-06,
+      "loss": 1.2644,
+      "step": 15115
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6597180574060698e-06,
+      "loss": 0.9459,
+      "step": 15116
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.658513480799432e-06,
+      "loss": 0.9913,
+      "step": 15117
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.657309058082017e-06,
+      "loss": 1.012,
+      "step": 15118
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6561047892830627e-06,
+      "loss": 1.2506,
+      "step": 15119
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6549006744317882e-06,
+      "loss": 1.049,
+      "step": 15120
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.653696713557413e-06,
+      "loss": 1.1583,
+      "step": 15121
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6524929066891524e-06,
+      "loss": 1.1687,
+      "step": 15122
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.651289253856225e-06,
+      "loss": 1.0691,
+      "step": 15123
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6500857550878366e-06,
+      "loss": 1.036,
+      "step": 15124
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6488824104131927e-06,
+      "loss": 1.0648,
+      "step": 15125
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.647679219861491e-06,
+      "loss": 0.9845,
+      "step": 15126
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6464761834619356e-06,
+      "loss": 1.0845,
+      "step": 15127
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.645273301243719e-06,
+      "loss": 1.1864,
+      "step": 15128
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6440705732360315e-06,
+      "loss": 1.0847,
+      "step": 15129
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.642867999468055e-06,
+      "loss": 1.0012,
+      "step": 15130
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6416655799689794e-06,
+      "loss": 1.2042,
+      "step": 15131
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6404633147679813e-06,
+      "loss": 1.0811,
+      "step": 15132
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6392612038942364e-06,
+      "loss": 1.1963,
+      "step": 15133
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6380592473769115e-06,
+      "loss": 1.0121,
+      "step": 15134
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6368574452451834e-06,
+      "loss": 1.1518,
+      "step": 15135
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6356557975282104e-06,
+      "loss": 1.1524,
+      "step": 15136
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.634454304255155e-06,
+      "loss": 1.1549,
+      "step": 15137
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.633252965455171e-06,
+      "loss": 1.2033,
+      "step": 15138
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6320517811574154e-06,
+      "loss": 1.0551,
+      "step": 15139
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.630850751391036e-06,
+      "loss": 1.1676,
+      "step": 15140
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.629649876185175e-06,
+      "loss": 1.2348,
+      "step": 15141
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6284491555689795e-06,
+      "loss": 1.0605,
+      "step": 15142
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.627248589571586e-06,
+      "loss": 1.1187,
+      "step": 15143
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.626048178222127e-06,
+      "loss": 1.1239,
+      "step": 15144
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.62484792154973e-06,
+      "loss": 1.1198,
+      "step": 15145
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.623647819583529e-06,
+      "loss": 1.0272,
+      "step": 15146
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6224478723526425e-06,
+      "loss": 1.335,
+      "step": 15147
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6212480798861915e-06,
+      "loss": 1.2257,
+      "step": 15148
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6200484422132865e-06,
+      "loss": 1.1786,
+      "step": 15149
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.618848959363046e-06,
+      "loss": 1.0162,
+      "step": 15150
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6176496313645757e-06,
+      "loss": 1.1808,
+      "step": 15151
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6164504582469796e-06,
+      "loss": 0.9801,
+      "step": 15152
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6152514400393535e-06,
+      "loss": 1.1729,
+      "step": 15153
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6140525767708014e-06,
+      "loss": 0.9467,
+      "step": 15154
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6128538684704117e-06,
+      "loss": 0.889,
+      "step": 15155
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.611655315167276e-06,
+      "loss": 1.0662,
+      "step": 15156
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6104569168904747e-06,
+      "loss": 1.0664,
+      "step": 15157
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.609258673669097e-06,
+      "loss": 1.198,
+      "step": 15158
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.608060585532216e-06,
+      "loss": 1.1363,
+      "step": 15159
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.606862652508907e-06,
+      "loss": 1.096,
+      "step": 15160
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.605664874628236e-06,
+      "loss": 1.1086,
+      "step": 15161
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.604467251919278e-06,
+      "loss": 1.1448,
+      "step": 15162
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6032697844110896e-06,
+      "loss": 1.017,
+      "step": 15163
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.602072472132733e-06,
+      "loss": 0.9793,
+      "step": 15164
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.600875315113257e-06,
+      "loss": 1.0925,
+      "step": 15165
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5996783133817192e-06,
+      "loss": 1.0689,
+      "step": 15166
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5984814669671687e-06,
+      "loss": 0.9611,
+      "step": 15167
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5972847758986475e-06,
+      "loss": 1.2285,
+      "step": 15168
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5960882402051945e-06,
+      "loss": 1.0402,
+      "step": 15169
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5948918599158432e-06,
+      "loss": 1.0213,
+      "step": 15170
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.593695635059633e-06,
+      "loss": 1.1975,
+      "step": 15171
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.592499565665589e-06,
+      "loss": 1.0009,
+      "step": 15172
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.591303651762739e-06,
+      "loss": 1.1839,
+      "step": 15173
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5901078933800957e-06,
+      "loss": 1.0609,
+      "step": 15174
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.588912290546689e-06,
+      "loss": 1.0397,
+      "step": 15175
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5877168432915256e-06,
+      "loss": 1.1202,
+      "step": 15176
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5865215516436157e-06,
+      "loss": 1.02,
+      "step": 15177
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5853264156319645e-06,
+      "loss": 1.1881,
+      "step": 15178
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.584131435285579e-06,
+      "loss": 1.1213,
+      "step": 15179
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.582936610633455e-06,
+      "loss": 1.0858,
+      "step": 15180
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.581741941704584e-06,
+      "loss": 1.1329,
+      "step": 15181
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.580547428527965e-06,
+      "loss": 1.2248,
+      "step": 15182
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5793530711325807e-06,
+      "loss": 1.1267,
+      "step": 15183
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5781588695474146e-06,
+      "loss": 1.0615,
+      "step": 15184
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5769648238014433e-06,
+      "loss": 1.0099,
+      "step": 15185
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5757709339236492e-06,
+      "loss": 1.0188,
+      "step": 15186
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5745771999430022e-06,
+      "loss": 1.093,
+      "step": 15187
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.57338362188847e-06,
+      "loss": 1.0925,
+      "step": 15188
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.572190199789014e-06,
+      "loss": 0.9102,
+      "step": 15189
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5709969336736017e-06,
+      "loss": 1.0735,
+      "step": 15190
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5698038235711875e-06,
+      "loss": 1.1038,
+      "step": 15191
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5686108695107223e-06,
+      "loss": 1.0425,
+      "step": 15192
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5674180715211548e-06,
+      "loss": 0.9813,
+      "step": 15193
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.566225429631437e-06,
+      "loss": 0.9195,
+      "step": 15194
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5650329438705067e-06,
+      "loss": 1.2372,
+      "step": 15195
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5638406142673022e-06,
+      "loss": 1.1186,
+      "step": 15196
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5626484408507543e-06,
+      "loss": 1.2096,
+      "step": 15197
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5614564236498006e-06,
+      "loss": 1.2163,
+      "step": 15198
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5602645626933642e-06,
+      "loss": 1.2442,
+      "step": 15199
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5590728580103685e-06,
+      "loss": 1.0738,
+      "step": 15200
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5578813096297293e-06,
+      "loss": 1.1443,
+      "step": 15201
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5566899175803683e-06,
+      "loss": 1.1874,
+      "step": 15202
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.555498681891194e-06,
+      "loss": 0.9408,
+      "step": 15203
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5543076025911136e-06,
+      "loss": 1.1659,
+      "step": 15204
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.553116679709029e-06,
+      "loss": 0.9583,
+      "step": 15205
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.551925913273846e-06,
+      "loss": 1.1,
+      "step": 15206
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.550735303314459e-06,
+      "loss": 1.1777,
+      "step": 15207
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5495448498597584e-06,
+      "loss": 1.1059,
+      "step": 15208
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5483545529386307e-06,
+      "loss": 1.1872,
+      "step": 15209
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5471644125799675e-06,
+      "loss": 1.1302,
+      "step": 15210
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5459744288126472e-06,
+      "loss": 0.9924,
+      "step": 15211
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5447846016655475e-06,
+      "loss": 1.0824,
+      "step": 15212
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.543594931167538e-06,
+      "loss": 1.0573,
+      "step": 15213
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5424054173474955e-06,
+      "loss": 0.9539,
+      "step": 15214
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5412160602342816e-06,
+      "loss": 1.1055,
+      "step": 15215
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.54002685985676e-06,
+      "loss": 1.0069,
+      "step": 15216
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.538837816243784e-06,
+      "loss": 1.1514,
+      "step": 15217
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5376489294242167e-06,
+      "loss": 1.243,
+      "step": 15218
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5364601994269055e-06,
+      "loss": 0.9184,
+      "step": 15219
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5352716262806953e-06,
+      "loss": 1.1425,
+      "step": 15220
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5340832100144284e-06,
+      "loss": 1.2017,
+      "step": 15221
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5328949506569488e-06,
+      "loss": 1.1268,
+      "step": 15222
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5317068482370907e-06,
+      "loss": 1.0816,
+      "step": 15223
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.530518902783684e-06,
+      "loss": 1.0984,
+      "step": 15224
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.529331114325555e-06,
+      "loss": 1.0771,
+      "step": 15225
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.528143482891533e-06,
+      "loss": 1.0785,
+      "step": 15226
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5269560085104347e-06,
+      "loss": 1.078,
+      "step": 15227
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5257686912110758e-06,
+      "loss": 1.0767,
+      "step": 15228
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5245815310222733e-06,
+      "loss": 1.1321,
+      "step": 15229
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.523394527972833e-06,
+      "loss": 0.9849,
+      "step": 15230
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5222076820915618e-06,
+      "loss": 1.0541,
+      "step": 15231
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.521020993407256e-06,
+      "loss": 1.04,
+      "step": 15232
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.519834461948719e-06,
+      "loss": 1.1747,
+      "step": 15233
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5186480877447437e-06,
+      "loss": 1.2703,
+      "step": 15234
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.517461870824117e-06,
+      "loss": 1.119,
+      "step": 15235
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.516275811215624e-06,
+      "loss": 1.1172,
+      "step": 15236
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5150899089480516e-06,
+      "loss": 1.2005,
+      "step": 15237
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.513904164050176e-06,
+      "loss": 1.1489,
+      "step": 15238
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5127185765507708e-06,
+      "loss": 1.0381,
+      "step": 15239
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5115331464786053e-06,
+      "loss": 1.1937,
+      "step": 15240
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5103478738624506e-06,
+      "loss": 1.1,
+      "step": 15241
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5091627587310682e-06,
+      "loss": 1.0504,
+      "step": 15242
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.507977801113216e-06,
+      "loss": 1.0771,
+      "step": 15243
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5067930010376484e-06,
+      "loss": 1.2411,
+      "step": 15244
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5056083585331204e-06,
+      "loss": 1.1495,
+      "step": 15245
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.50442387362838e-06,
+      "loss": 0.8853,
+      "step": 15246
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5032395463521674e-06,
+      "loss": 1.0084,
+      "step": 15247
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5020553767332233e-06,
+      "loss": 0.9612,
+      "step": 15248
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.500871364800288e-06,
+      "loss": 1.0877,
+      "step": 15249
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.499687510582087e-06,
+      "loss": 1.1533,
+      "step": 15250
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.498503814107358e-06,
+      "loss": 1.1598,
+      "step": 15251
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4973202754048207e-06,
+      "loss": 1.1052,
+      "step": 15252
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4961368945031936e-06,
+      "loss": 1.0912,
+      "step": 15253
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4949536714312005e-06,
+      "loss": 1.2209,
+      "step": 15254
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4937706062175513e-06,
+      "loss": 1.1144,
+      "step": 15255
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4925876988909544e-06,
+      "loss": 1.2186,
+      "step": 15256
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4914049494801137e-06,
+      "loss": 1.0741,
+      "step": 15257
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4902223580137372e-06,
+      "loss": 0.9981,
+      "step": 15258
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4890399245205186e-06,
+      "loss": 1.0544,
+      "step": 15259
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4878576490291537e-06,
+      "loss": 1.2063,
+      "step": 15260
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.486675531568329e-06,
+      "loss": 1.3029,
+      "step": 15261
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.485493572166737e-06,
+      "loss": 1.0375,
+      "step": 15262
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4843117708530583e-06,
+      "loss": 1.0917,
+      "step": 15263
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4831301276559706e-06,
+      "loss": 1.057,
+      "step": 15264
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.481948642604146e-06,
+      "loss": 0.9069,
+      "step": 15265
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.480767315726261e-06,
+      "loss": 1.0416,
+      "step": 15266
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.479586147050983e-06,
+      "loss": 1.197,
+      "step": 15267
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.478405136606968e-06,
+      "loss": 1.1999,
+      "step": 15268
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4772242844228866e-06,
+      "loss": 1.0587,
+      "step": 15269
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.476043590527387e-06,
+      "loss": 1.0656,
+      "step": 15270
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4748630549491237e-06,
+      "loss": 1.0064,
+      "step": 15271
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.473682677716742e-06,
+      "loss": 1.0987,
+      "step": 15272
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4725024588588906e-06,
+      "loss": 1.1649,
+      "step": 15273
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4713223984042086e-06,
+      "loss": 1.1971,
+      "step": 15274
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4701424963813313e-06,
+      "loss": 1.1147,
+      "step": 15275
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4689627528188883e-06,
+      "loss": 1.1702,
+      "step": 15276
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4677831677455166e-06,
+      "loss": 1.0772,
+      "step": 15277
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.466603741189837e-06,
+      "loss": 1.0711,
+      "step": 15278
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4654244731804697e-06,
+      "loss": 1.1466,
+      "step": 15279
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4642453637460295e-06,
+      "loss": 1.1105,
+      "step": 15280
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.463066412915137e-06,
+      "loss": 1.0856,
+      "step": 15281
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4618876207163987e-06,
+      "loss": 0.9523,
+      "step": 15282
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4607089871784193e-06,
+      "loss": 1.0937,
+      "step": 15283
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4595305123297984e-06,
+      "loss": 1.2354,
+      "step": 15284
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.458352196199141e-06,
+      "loss": 1.1051,
+      "step": 15285
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4571740388150365e-06,
+      "loss": 0.9515,
+      "step": 15286
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4559960402060764e-06,
+      "loss": 1.0966,
+      "step": 15287
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4548182004008447e-06,
+      "loss": 1.1275,
+      "step": 15288
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4536405194279286e-06,
+      "loss": 1.1668,
+      "step": 15289
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4524629973159063e-06,
+      "loss": 1.0399,
+      "step": 15290
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4512856340933507e-06,
+      "loss": 1.1659,
+      "step": 15291
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.450108429788831e-06,
+      "loss": 0.8573,
+      "step": 15292
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4489313844309204e-06,
+      "loss": 1.1983,
+      "step": 15293
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.447754498048178e-06,
+      "loss": 1.0657,
+      "step": 15294
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.446577770669165e-06,
+      "loss": 1.1511,
+      "step": 15295
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4454012023224337e-06,
+      "loss": 1.0478,
+      "step": 15296
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4442247930365423e-06,
+      "loss": 1.1345,
+      "step": 15297
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.443048542840035e-06,
+      "loss": 1.2661,
+      "step": 15298
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4418724517614565e-06,
+      "loss": 1.0477,
+      "step": 15299
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4406965198293428e-06,
+      "loss": 1.2149,
+      "step": 15300
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.439520747072238e-06,
+      "loss": 1.1982,
+      "step": 15301
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.438345133518671e-06,
+      "loss": 1.1041,
+      "step": 15302
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4371696791971708e-06,
+      "loss": 1.0548,
+      "step": 15303
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4359943841362586e-06,
+      "loss": 1.1816,
+      "step": 15304
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4348192483644614e-06,
+      "loss": 1.0832,
+      "step": 15305
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4336442719102934e-06,
+      "loss": 1.0297,
+      "step": 15306
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4324694548022686e-06,
+      "loss": 1.0413,
+      "step": 15307
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4312947970688927e-06,
+      "loss": 1.0169,
+      "step": 15308
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4301202987386772e-06,
+      "loss": 1.1504,
+      "step": 15309
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4289459598401197e-06,
+      "loss": 1.1903,
+      "step": 15310
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4277717804017208e-06,
+      "loss": 1.0299,
+      "step": 15311
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4265977604519686e-06,
+      "loss": 1.1539,
+      "step": 15312
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.425423900019359e-06,
+      "loss": 1.0568,
+      "step": 15313
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.424250199132377e-06,
+      "loss": 1.0626,
+      "step": 15314
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4230766578195016e-06,
+      "loss": 1.121,
+      "step": 15315
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4219032761092163e-06,
+      "loss": 1.1384,
+      "step": 15316
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4207300540299935e-06,
+      "loss": 1.2451,
+      "step": 15317
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.419556991610303e-06,
+      "loss": 1.1039,
+      "step": 15318
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.418384088878609e-06,
+      "loss": 1.0534,
+      "step": 15319
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4172113458633803e-06,
+      "loss": 1.1591,
+      "step": 15320
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4160387625930723e-06,
+      "loss": 1.0828,
+      "step": 15321
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.414866339096142e-06,
+      "loss": 1.1224,
+      "step": 15322
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.413694075401035e-06,
+      "loss": 1.0412,
+      "step": 15323
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.412521971536208e-06,
+      "loss": 0.9814,
+      "step": 15324
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.411350027530098e-06,
+      "loss": 1.2074,
+      "step": 15325
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4101782434111476e-06,
+      "loss": 1.1807,
+      "step": 15326
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4090066192077864e-06,
+      "loss": 1.1062,
+      "step": 15327
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4078351549484555e-06,
+      "loss": 1.1128,
+      "step": 15328
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.406663850661579e-06,
+      "loss": 1.2424,
+      "step": 15329
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4054927063755793e-06,
+      "loss": 0.9431,
+      "step": 15330
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4043217221188764e-06,
+      "loss": 1.097,
+      "step": 15331
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.40315089791989e-06,
+      "loss": 1.1116,
+      "step": 15332
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4019802338070276e-06,
+      "loss": 1.0993,
+      "step": 15333
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.400809729808704e-06,
+      "loss": 1.1046,
+      "step": 15334
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3996393859533204e-06,
+      "loss": 1.1553,
+      "step": 15335
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3984692022692756e-06,
+      "loss": 1.1178,
+      "step": 15336
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.397299178784972e-06,
+      "loss": 1.1184,
+      "step": 15337
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.396129315528799e-06,
+      "loss": 0.9853,
+      "step": 15338
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3949596125291463e-06,
+      "loss": 1.1138,
+      "step": 15339
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3937900698143967e-06,
+      "loss": 1.0499,
+      "step": 15340
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3926206874129363e-06,
+      "loss": 1.0591,
+      "step": 15341
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3914514653531404e-06,
+      "loss": 1.053,
+      "step": 15342
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.390282403663382e-06,
+      "loss": 1.1394,
+      "step": 15343
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.389113502372028e-06,
+      "loss": 1.0721,
+      "step": 15344
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3879447615074514e-06,
+      "loss": 1.0532,
+      "step": 15345
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.386776181098009e-06,
+      "loss": 1.047,
+      "step": 15346
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3856077611720607e-06,
+      "loss": 1.1679,
+      "step": 15347
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3844395017579566e-06,
+      "loss": 1.1025,
+      "step": 15348
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.383271402884053e-06,
+      "loss": 1.1784,
+      "step": 15349
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3821034645786934e-06,
+      "loss": 0.8376,
+      "step": 15350
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3809356868702202e-06,
+      "loss": 0.989,
+      "step": 15351
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3797680697869685e-06,
+      "loss": 1.0365,
+      "step": 15352
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3786006133572803e-06,
+      "loss": 1.0232,
+      "step": 15353
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.377433317609482e-06,
+      "loss": 1.0639,
+      "step": 15354
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.376266182571897e-06,
+      "loss": 1.0891,
+      "step": 15355
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.375099208272856e-06,
+      "loss": 1.0288,
+      "step": 15356
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3739323947406744e-06,
+      "loss": 1.1119,
+      "step": 15357
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3727657420036665e-06,
+      "loss": 1.1096,
+      "step": 15358
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3715992500901405e-06,
+      "loss": 1.173,
+      "step": 15359
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3704329190284104e-06,
+      "loss": 1.0411,
+      "step": 15360
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.369266748846777e-06,
+      "loss": 1.0315,
+      "step": 15361
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.368100739573539e-06,
+      "loss": 1.1801,
+      "step": 15362
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.366934891236989e-06,
+      "loss": 1.0297,
+      "step": 15363
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3657692038654256e-06,
+      "loss": 1.0965,
+      "step": 15364
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3646036774871338e-06,
+      "loss": 1.0216,
+      "step": 15365
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.363438312130396e-06,
+      "loss": 1.0301,
+      "step": 15366
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.362273107823489e-06,
+      "loss": 1.0173,
+      "step": 15367
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.361108064594697e-06,
+      "loss": 1.0953,
+      "step": 15368
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3599431824722873e-06,
+      "loss": 1.1092,
+      "step": 15369
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.35877846148453e-06,
+      "loss": 1.0793,
+      "step": 15370
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3576139016596843e-06,
+      "loss": 0.9283,
+      "step": 15371
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.356449503026018e-06,
+      "loss": 1.0788,
+      "step": 15372
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.355285265611784e-06,
+      "loss": 1.0801,
+      "step": 15373
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.354121189445235e-06,
+      "loss": 1.0848,
+      "step": 15374
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.352957274554618e-06,
+      "loss": 1.1436,
+      "step": 15375
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.351793520968182e-06,
+      "loss": 1.22,
+      "step": 15376
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.350629928714165e-06,
+      "loss": 1.0794,
+      "step": 15377
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3494664978208056e-06,
+      "loss": 1.0346,
+      "step": 15378
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.348303228316332e-06,
+      "loss": 1.0914,
+      "step": 15379
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.34714012022898e-06,
+      "loss": 1.0474,
+      "step": 15380
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.345977173586973e-06,
+      "loss": 1.2185,
+      "step": 15381
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3448143884185302e-06,
+      "loss": 0.8458,
+      "step": 15382
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3436517647518673e-06,
+      "loss": 1.0444,
+      "step": 15383
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3424893026152027e-06,
+      "loss": 1.1479,
+      "step": 15384
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.341327002036745e-06,
+      "loss": 0.9791,
+      "step": 15385
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3401648630446983e-06,
+      "loss": 1.0149,
+      "step": 15386
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3390028856672607e-06,
+      "loss": 1.0039,
+      "step": 15387
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.337841069932637e-06,
+      "loss": 1.0763,
+      "step": 15388
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3366794158690187e-06,
+      "loss": 1.0433,
+      "step": 15389
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3355179235045944e-06,
+      "loss": 1.2009,
+      "step": 15390
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.334356592867547e-06,
+      "loss": 1.0102,
+      "step": 15391
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3331954239860655e-06,
+      "loss": 1.0726,
+      "step": 15392
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3320344168883245e-06,
+      "loss": 1.1763,
+      "step": 15393
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.330873571602499e-06,
+      "loss": 1.1811,
+      "step": 15394
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.329712888156755e-06,
+      "loss": 1.1527,
+      "step": 15395
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3285523665792663e-06,
+      "loss": 1.1095,
+      "step": 15396
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3273920068981913e-06,
+      "loss": 1.1053,
+      "step": 15397
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.326231809141686e-06,
+      "loss": 1.1864,
+      "step": 15398
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.32507177333791e-06,
+      "loss": 1.145,
+      "step": 15399
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.323911899515013e-06,
+      "loss": 1.0452,
+      "step": 15400
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3227521877011405e-06,
+      "loss": 1.1968,
+      "step": 15401
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3215926379244314e-06,
+      "loss": 0.9596,
+      "step": 15402
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.320433250213032e-06,
+      "loss": 1.0395,
+      "step": 15403
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.319274024595074e-06,
+      "loss": 1.049,
+      "step": 15404
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3181149610986874e-06,
+      "loss": 1.1497,
+      "step": 15405
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.316956059751997e-06,
+      "loss": 1.0996,
+      "step": 15406
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3157973205831327e-06,
+      "loss": 1.0886,
+      "step": 15407
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3146387436202086e-06,
+      "loss": 1.0303,
+      "step": 15408
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3134803288913433e-06,
+      "loss": 1.1849,
+      "step": 15409
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.312322076424641e-06,
+      "loss": 1.0209,
+      "step": 15410
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.311163986248218e-06,
+      "loss": 1.0858,
+      "step": 15411
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3100060583901737e-06,
+      "loss": 1.1612,
+      "step": 15412
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3088482928786093e-06,
+      "loss": 1.147,
+      "step": 15413
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.307690689741614e-06,
+      "loss": 1.1347,
+      "step": 15414
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3065332490072877e-06,
+      "loss": 1.0434,
+      "step": 15415
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3053759707037116e-06,
+      "loss": 1.1505,
+      "step": 15416
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.304218854858976e-06,
+      "loss": 1.1991,
+      "step": 15417
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3030619015011567e-06,
+      "loss": 1.0366,
+      "step": 15418
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3019051106583278e-06,
+      "loss": 0.8852,
+      "step": 15419
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3007484823585655e-06,
+      "loss": 1.0899,
+      "step": 15420
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2995920166299367e-06,
+      "loss": 0.9723,
+      "step": 15421
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.298435713500504e-06,
+      "loss": 1.1814,
+      "step": 15422
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2972795729983242e-06,
+      "loss": 1.2077,
+      "step": 15423
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2961235951514616e-06,
+      "loss": 1.1643,
+      "step": 15424
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.294967779987963e-06,
+      "loss": 1.0918,
+      "step": 15425
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.293812127535878e-06,
+      "loss": 1.1821,
+      "step": 15426
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2926566378232474e-06,
+      "loss": 1.0934,
+      "step": 15427
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.291501310878117e-06,
+      "loss": 1.1675,
+      "step": 15428
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2903461467285216e-06,
+      "loss": 1.2191,
+      "step": 15429
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.289191145402493e-06,
+      "loss": 1.1303,
+      "step": 15430
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2880363069280553e-06,
+      "loss": 1.0399,
+      "step": 15431
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.286881631333242e-06,
+      "loss": 1.0199,
+      "step": 15432
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2857271186460672e-06,
+      "loss": 1.1701,
+      "step": 15433
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2845727688945504e-06,
+      "loss": 1.1291,
+      "step": 15434
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2834185821067e-06,
+      "loss": 1.1365,
+      "step": 15435
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2822645583105306e-06,
+      "loss": 1.1039,
+      "step": 15436
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2811106975340445e-06,
+      "loss": 1.0726,
+      "step": 15437
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.279956999805238e-06,
+      "loss": 1.1156,
+      "step": 15438
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.278803465152116e-06,
+      "loss": 1.1138,
+      "step": 15439
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.277650093602668e-06,
+      "loss": 1.1109,
+      "step": 15440
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2764968851848822e-06,
+      "loss": 1.1461,
+      "step": 15441
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2753438399267414e-06,
+      "loss": 1.0476,
+      "step": 15442
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2741909578562314e-06,
+      "loss": 1.0261,
+      "step": 15443
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.273038239001327e-06,
+      "loss": 1.0786,
+      "step": 15444
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2718856833900013e-06,
+      "loss": 1.117,
+      "step": 15445
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.270733291050221e-06,
+      "loss": 1.1164,
+      "step": 15446
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2695810620099566e-06,
+      "loss": 1.1406,
+      "step": 15447
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.268428996297166e-06,
+      "loss": 1.114,
+      "step": 15448
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.267277093939807e-06,
+      "loss": 1.0928,
+      "step": 15449
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.266125354965829e-06,
+      "loss": 1.0873,
+      "step": 15450
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.264973779403189e-06,
+      "loss": 0.9513,
+      "step": 15451
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2638223672798274e-06,
+      "loss": 1.0727,
+      "step": 15452
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.262671118623687e-06,
+      "loss": 1.1601,
+      "step": 15453
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2615200334627017e-06,
+      "loss": 1.0002,
+      "step": 15454
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2603691118248115e-06,
+      "loss": 1.165,
+      "step": 15455
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.259218353737942e-06,
+      "loss": 1.1534,
+      "step": 15456
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2580677592300192e-06,
+      "loss": 1.0546,
+      "step": 15457
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.256917328328961e-06,
+      "loss": 1.0217,
+      "step": 15458
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2557670610626924e-06,
+      "loss": 1.0876,
+      "step": 15459
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2546169574591233e-06,
+      "loss": 0.949,
+      "step": 15460
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2534670175461626e-06,
+      "loss": 1.0971,
+      "step": 15461
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2523172413517134e-06,
+      "loss": 1.077,
+      "step": 15462
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2511676289036844e-06,
+      "loss": 1.1739,
+      "step": 15463
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.250018180229969e-06,
+      "loss": 1.0013,
+      "step": 15464
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2488688953584612e-06,
+      "loss": 1.0575,
+      "step": 15465
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2477197743170476e-06,
+      "loss": 1.1763,
+      "step": 15466
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2465708171336206e-06,
+      "loss": 1.0508,
+      "step": 15467
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.24542202383606e-06,
+      "loss": 1.2224,
+      "step": 15468
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.244273394452241e-06,
+      "loss": 1.1679,
+      "step": 15469
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2431249290100365e-06,
+      "loss": 1.1104,
+      "step": 15470
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.241976627537321e-06,
+      "loss": 1.1773,
+      "step": 15471
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.240828490061958e-06,
+      "loss": 1.0523,
+      "step": 15472
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2396805166118108e-06,
+      "loss": 1.0746,
+      "step": 15473
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2385327072147323e-06,
+      "loss": 1.0636,
+      "step": 15474
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2373850618985826e-06,
+      "loss": 1.2359,
+      "step": 15475
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2362375806912106e-06,
+      "loss": 1.0817,
+      "step": 15476
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2350902636204594e-06,
+      "loss": 1.1342,
+      "step": 15477
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2339431107141707e-06,
+      "loss": 1.1973,
+      "step": 15478
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.232796122000188e-06,
+      "loss": 1.0526,
+      "step": 15479
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2316492975063407e-06,
+      "loss": 1.1865,
+      "step": 15480
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2305026372604596e-06,
+      "loss": 0.9408,
+      "step": 15481
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2293561412903697e-06,
+      "loss": 0.9071,
+      "step": 15482
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2282098096238967e-06,
+      "loss": 1.0201,
+      "step": 15483
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2270636422888556e-06,
+      "loss": 1.0355,
+      "step": 15484
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2259176393130596e-06,
+      "loss": 1.0764,
+      "step": 15485
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2247718007243235e-06,
+      "loss": 1.042,
+      "step": 15486
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2236261265504508e-06,
+      "loss": 1.1099,
+      "step": 15487
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2224806168192423e-06,
+      "loss": 1.073,
+      "step": 15488
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2213352715584955e-06,
+      "loss": 1.0875,
+      "step": 15489
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.220190090796008e-06,
+      "loss": 0.9242,
+      "step": 15490
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2190450745595693e-06,
+      "loss": 1.0137,
+      "step": 15491
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2179002228769653e-06,
+      "loss": 1.1663,
+      "step": 15492
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.216755535775973e-06,
+      "loss": 1.142,
+      "step": 15493
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2156110132843775e-06,
+      "loss": 1.1104,
+      "step": 15494
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.214466655429952e-06,
+      "loss": 1.0235,
+      "step": 15495
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2133224622404648e-06,
+      "loss": 0.973,
+      "step": 15496
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.21217843374368e-06,
+      "loss": 1.018,
+      "step": 15497
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.211034569967365e-06,
+      "loss": 1.0173,
+      "step": 15498
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.209890870939275e-06,
+      "loss": 1.1391,
+      "step": 15499
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2087473366871615e-06,
+      "loss": 1.0591,
+      "step": 15500
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2076039672387817e-06,
+      "loss": 1.0455,
+      "step": 15501
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2064607626218737e-06,
+      "loss": 1.0796,
+      "step": 15502
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2053177228641872e-06,
+      "loss": 1.1378,
+      "step": 15503
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2041748479934575e-06,
+      "loss": 1.1939,
+      "step": 15504
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2030321380374185e-06,
+      "loss": 1.181,
+      "step": 15505
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.201889593023797e-06,
+      "loss": 0.9916,
+      "step": 15506
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.200747212980326e-06,
+      "loss": 1.1603,
+      "step": 15507
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.199604997934724e-06,
+      "loss": 1.1229,
+      "step": 15508
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.19846294791471e-06,
+      "loss": 1.0807,
+      "step": 15509
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1973210629479936e-06,
+      "loss": 1.086,
+      "step": 15510
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.196179343062292e-06,
+      "loss": 1.187,
+      "step": 15511
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.19503778828531e-06,
+      "loss": 1.1008,
+      "step": 15512
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.193896398644747e-06,
+      "loss": 1.1386,
+      "step": 15513
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1927551741682995e-06,
+      "loss": 1.1531,
+      "step": 15514
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1916141148836678e-06,
+      "loss": 0.8919,
+      "step": 15515
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1904732208185384e-06,
+      "loss": 1.0251,
+      "step": 15516
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1893324920005974e-06,
+      "loss": 1.1343,
+      "step": 15517
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.188191928457526e-06,
+      "loss": 1.2503,
+      "step": 15518
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.187051530217005e-06,
+      "loss": 1.0679,
+      "step": 15519
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1859112973067086e-06,
+      "loss": 1.037,
+      "step": 15520
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.184771229754304e-06,
+      "loss": 1.1101,
+      "step": 15521
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1836313275874564e-06,
+      "loss": 1.1679,
+      "step": 15522
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.182491590833834e-06,
+      "loss": 1.0271,
+      "step": 15523
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1813520195210903e-06,
+      "loss": 1.0416,
+      "step": 15524
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1802126136768776e-06,
+      "loss": 0.9847,
+      "step": 15525
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1790733733288513e-06,
+      "loss": 1.0341,
+      "step": 15526
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.177934298504655e-06,
+      "loss": 0.9805,
+      "step": 15527
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1767953892319294e-06,
+      "loss": 1.0192,
+      "step": 15528
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1756566455383107e-06,
+      "loss": 1.1673,
+      "step": 15529
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1745180674514388e-06,
+      "loss": 0.9865,
+      "step": 15530
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1733796549989403e-06,
+      "loss": 1.1246,
+      "step": 15531
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.172241408208442e-06,
+      "loss": 1.0758,
+      "step": 15532
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.171103327107561e-06,
+      "loss": 1.0291,
+      "step": 15533
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.169965411723922e-06,
+      "loss": 0.9996,
+      "step": 15534
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1688276620851368e-06,
+      "loss": 1.2398,
+      "step": 15535
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1676900782188136e-06,
+      "loss": 1.0996,
+      "step": 15536
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1665526601525567e-06,
+      "loss": 1.1503,
+      "step": 15537
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.165415407913973e-06,
+      "loss": 1.0988,
+      "step": 15538
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1642783215306585e-06,
+      "loss": 1.2175,
+      "step": 15539
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.163141401030205e-06,
+      "loss": 1.0264,
+      "step": 15540
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1620046464401997e-06,
+      "loss": 0.9498,
+      "step": 15541
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1608680577882367e-06,
+      "loss": 1.1044,
+      "step": 15542
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1597316351018914e-06,
+      "loss": 1.0975,
+      "step": 15543
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.158595378408743e-06,
+      "loss": 1.1261,
+      "step": 15544
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.157459287736362e-06,
+      "loss": 1.1074,
+      "step": 15545
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1563233631123237e-06,
+      "loss": 1.093,
+      "step": 15546
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1551876045641915e-06,
+      "loss": 1.0733,
+      "step": 15547
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.154052012119526e-06,
+      "loss": 0.8367,
+      "step": 15548
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.152916585805882e-06,
+      "loss": 0.9794,
+      "step": 15549
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1517813256508177e-06,
+      "loss": 1.0542,
+      "step": 15550
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.150646231681882e-06,
+      "loss": 1.1907,
+      "step": 15551
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.149511303926619e-06,
+      "loss": 1.2659,
+      "step": 15552
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.148376542412567e-06,
+      "loss": 1.1427,
+      "step": 15553
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.147241947167269e-06,
+      "loss": 1.1706,
+      "step": 15554
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1461075182182565e-06,
+      "loss": 1.0272,
+      "step": 15555
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1449732555930578e-06,
+      "loss": 1.1794,
+      "step": 15556
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1438391593191943e-06,
+      "loss": 1.1575,
+      "step": 15557
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1427052294241943e-06,
+      "loss": 1.0927,
+      "step": 15558
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1415714659355736e-06,
+      "loss": 1.2162,
+      "step": 15559
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.140437868880841e-06,
+      "loss": 0.9943,
+      "step": 15560
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.139304438287507e-06,
+      "loss": 0.9982,
+      "step": 15561
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.13817117418308e-06,
+      "loss": 0.9706,
+      "step": 15562
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1370380765950583e-06,
+      "loss": 1.1891,
+      "step": 15563
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1359051455509393e-06,
+      "loss": 1.1047,
+      "step": 15564
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.134772381078214e-06,
+      "loss": 1.198,
+      "step": 15565
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.133639783204375e-06,
+      "loss": 1.0073,
+      "step": 15566
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1325073519569047e-06,
+      "loss": 1.0858,
+      "step": 15567
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1313750873632854e-06,
+      "loss": 1.2118,
+      "step": 15568
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.13024298945099e-06,
+      "loss": 1.1152,
+      "step": 15569
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.129111058247497e-06,
+      "loss": 0.9963,
+      "step": 15570
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.127979293780271e-06,
+      "loss": 1.0491,
+      "step": 15571
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1268476960767756e-06,
+      "loss": 1.1851,
+      "step": 15572
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1257162651644756e-06,
+      "loss": 1.1119,
+      "step": 15573
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1245850010708266e-06,
+      "loss": 1.2117,
+      "step": 15574
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.123453903823279e-06,
+      "loss": 1.1431,
+      "step": 15575
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.122322973449279e-06,
+      "loss": 0.9976,
+      "step": 15576
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.121192209976277e-06,
+      "loss": 1.0464,
+      "step": 15577
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.12006161343171e-06,
+      "loss": 1.1599,
+      "step": 15578
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1189311838430147e-06,
+      "loss": 1.1768,
+      "step": 15579
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1178009212376193e-06,
+      "loss": 1.0436,
+      "step": 15580
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.116670825642959e-06,
+      "loss": 1.0993,
+      "step": 15581
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1155408970864555e-06,
+      "loss": 1.0427,
+      "step": 15582
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1144111355955243e-06,
+      "loss": 1.1193,
+      "step": 15583
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.113281541197588e-06,
+      "loss": 1.1094,
+      "step": 15584
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.112152113920053e-06,
+      "loss": 0.9941,
+      "step": 15585
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.111022853790332e-06,
+      "loss": 1.0735,
+      "step": 15586
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.109893760835827e-06,
+      "loss": 1.0884,
+      "step": 15587
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1087648350839382e-06,
+      "loss": 1.2567,
+      "step": 15588
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.107636076562056e-06,
+      "loss": 1.112,
+      "step": 15589
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1065074852975806e-06,
+      "loss": 0.9368,
+      "step": 15590
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1053790613178945e-06,
+      "loss": 1.1438,
+      "step": 15591
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.104250804650384e-06,
+      "loss": 1.036,
+      "step": 15592
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1031227153224217e-06,
+      "loss": 1.138,
+      "step": 15593
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1019947933613927e-06,
+      "loss": 1.2349,
+      "step": 15594
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1008670387946625e-06,
+      "loss": 1.1464,
+      "step": 15595
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0997394516496004e-06,
+      "loss": 1.1782,
+      "step": 15596
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0986120319535664e-06,
+      "loss": 1.0808,
+      "step": 15597
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0974847797339235e-06,
+      "loss": 1.1565,
+      "step": 15598
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0963576950180275e-06,
+      "loss": 1.1411,
+      "step": 15599
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0952307778332256e-06,
+      "loss": 1.2266,
+      "step": 15600
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0941040282068645e-06,
+      "loss": 0.8995,
+      "step": 15601
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.092977446166292e-06,
+      "loss": 1.1061,
+      "step": 15602
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.091851031738844e-06,
+      "loss": 0.8312,
+      "step": 15603
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.090724784951855e-06,
+      "loss": 1.0662,
+      "step": 15604
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0895987058326528e-06,
+      "loss": 1.0135,
+      "step": 15605
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0884727944085714e-06,
+      "loss": 1.2255,
+      "step": 15606
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.087347050706928e-06,
+      "loss": 1.1183,
+      "step": 15607
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0862214747550434e-06,
+      "loss": 1.1743,
+      "step": 15608
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0850960665802267e-06,
+      "loss": 1.1564,
+      "step": 15609
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0839708262097955e-06,
+      "loss": 1.0933,
+      "step": 15610
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0828457536710544e-06,
+      "loss": 1.0582,
+      "step": 15611
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0817208489913e-06,
+      "loss": 1.0586,
+      "step": 15612
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0805961121978377e-06,
+      "loss": 1.0268,
+      "step": 15613
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.07947154331796e-06,
+      "loss": 1.1499,
+      "step": 15614
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0783471423789536e-06,
+      "loss": 1.0303,
+      "step": 15615
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0772229094081042e-06,
+      "loss": 1.219,
+      "step": 15616
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.076098844432698e-06,
+      "loss": 1.0042,
+      "step": 15617
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0749749474800107e-06,
+      "loss": 1.0187,
+      "step": 15618
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0738512185773163e-06,
+      "loss": 1.0702,
+      "step": 15619
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0727276577518795e-06,
+      "loss": 1.1291,
+      "step": 15620
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.071604265030974e-06,
+      "loss": 1.1311,
+      "step": 15621
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.070481040441857e-06,
+      "loss": 1.0495,
+      "step": 15622
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.069357984011786e-06,
+      "loss": 1.1567,
+      "step": 15623
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.068235095768012e-06,
+      "loss": 1.0346,
+      "step": 15624
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0671123757377898e-06,
+      "loss": 1.0435,
+      "step": 15625
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.065989823948361e-06,
+      "loss": 1.0138,
+      "step": 15626
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.064867440426966e-06,
+      "loss": 1.2193,
+      "step": 15627
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0637452252008404e-06,
+      "loss": 1.1422,
+      "step": 15628
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0626231782972227e-06,
+      "loss": 1.0944,
+      "step": 15629
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.061501299743338e-06,
+      "loss": 1.1276,
+      "step": 15630
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0603795895664125e-06,
+      "loss": 1.0499,
+      "step": 15631
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0592580477936606e-06,
+      "loss": 1.0909,
+      "step": 15632
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.058136674452309e-06,
+      "loss": 1.1864,
+      "step": 15633
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0570154695695642e-06,
+      "loss": 1.0481,
+      "step": 15634
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.055894433172635e-06,
+      "loss": 1.1711,
+      "step": 15635
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0547735652887236e-06,
+      "loss": 1.0085,
+      "step": 15636
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0536528659450348e-06,
+      "loss": 1.0715,
+      "step": 15637
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.052532335168763e-06,
+      "loss": 1.0772,
+      "step": 15638
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0514119729870995e-06,
+      "loss": 0.9948,
+      "step": 15639
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.050291779427229e-06,
+      "loss": 1.1029,
+      "step": 15640
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0491717545163426e-06,
+      "loss": 1.1177,
+      "step": 15641
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0480518982816155e-06,
+      "loss": 1.0696,
+      "step": 15642
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.046932210750224e-06,
+      "loss": 1.1812,
+      "step": 15643
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.045812691949336e-06,
+      "loss": 1.0164,
+      "step": 15644
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0446933419061263e-06,
+      "loss": 1.1506,
+      "step": 15645
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0435741606477543e-06,
+      "loss": 1.2299,
+      "step": 15646
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0424551482013787e-06,
+      "loss": 1.2413,
+      "step": 15647
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0413363045941525e-06,
+      "loss": 1.1279,
+      "step": 15648
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.040217629853233e-06,
+      "loss": 1.0565,
+      "step": 15649
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.039099124005762e-06,
+      "loss": 1.1132,
+      "step": 15650
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0379807870788857e-06,
+      "loss": 0.9906,
+      "step": 15651
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0368626190997374e-06,
+      "loss": 1.13,
+      "step": 15652
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0357446200954586e-06,
+      "loss": 1.0614,
+      "step": 15653
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.034626790093177e-06,
+      "loss": 1.1479,
+      "step": 15654
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0335091291200193e-06,
+      "loss": 1.1175,
+      "step": 15655
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0323916372031024e-06,
+      "loss": 0.8603,
+      "step": 15656
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0312743143695544e-06,
+      "loss": 1.0845,
+      "step": 15657
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.030157160646483e-06,
+      "loss": 1.2837,
+      "step": 15658
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0290401760609976e-06,
+      "loss": 1.0642,
+      "step": 15659
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0279233606402094e-06,
+      "loss": 0.9817,
+      "step": 15660
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0268067144112167e-06,
+      "loss": 0.9703,
+      "step": 15661
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0256902374011177e-06,
+      "loss": 1.0369,
+      "step": 15662
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0245739296370035e-06,
+      "loss": 1.2147,
+      "step": 15663
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0234577911459686e-06,
+      "loss": 1.1587,
+      "step": 15664
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.022341821955096e-06,
+      "loss": 1.196,
+      "step": 15665
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0212260220914636e-06,
+      "loss": 0.9941,
+      "step": 15666
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.020110391582154e-06,
+      "loss": 1.1008,
+      "step": 15667
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0189949304542365e-06,
+      "loss": 1.1422,
+      "step": 15668
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.017879638734784e-06,
+      "loss": 1.0581,
+      "step": 15669
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0167645164508584e-06,
+      "loss": 1.1645,
+      "step": 15670
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0156495636295224e-06,
+      "loss": 1.1904,
+      "step": 15671
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0145347802978274e-06,
+      "loss": 1.101,
+      "step": 15672
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0134201664828324e-06,
+      "loss": 1.242,
+      "step": 15673
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0123057222115835e-06,
+      "loss": 1.0822,
+      "step": 15674
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0111914475111247e-06,
+      "loss": 1.1434,
+      "step": 15675
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.010077342408494e-06,
+      "loss": 0.9934,
+      "step": 15676
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0089634069307316e-06,
+      "loss": 1.076,
+      "step": 15677
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0078496411048675e-06,
+      "loss": 1.0592,
+      "step": 15678
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0067360449579297e-06,
+      "loss": 0.9917,
+      "step": 15679
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.005622618516938e-06,
+      "loss": 1.1071,
+      "step": 15680
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.004509361808919e-06,
+      "loss": 1.0466,
+      "step": 15681
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.003396274860886e-06,
+      "loss": 1.018,
+      "step": 15682
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.002283357699848e-06,
+      "loss": 1.1041,
+      "step": 15683
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.001170610352809e-06,
+      "loss": 1.1194,
+      "step": 15684
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.000058032846782e-06,
+      "loss": 1.1528,
+      "step": 15685
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.998945625208759e-06,
+      "loss": 1.1724,
+      "step": 15686
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9978333874657373e-06,
+      "loss": 1.058,
+      "step": 15687
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.996721319644702e-06,
+      "loss": 1.0686,
+      "step": 15688
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.995609421772648e-06,
+      "loss": 0.9634,
+      "step": 15689
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9944976938765557e-06,
+      "loss": 1.1334,
+      "step": 15690
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9933861359834004e-06,
+      "loss": 1.1254,
+      "step": 15691
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9922747481201554e-06,
+      "loss": 1.1629,
+      "step": 15692
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9911635303137964e-06,
+      "loss": 1.14,
+      "step": 15693
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.990052482591286e-06,
+      "loss": 1.0361,
+      "step": 15694
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9889416049795826e-06,
+      "loss": 1.1065,
+      "step": 15695
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9878308975056513e-06,
+      "loss": 1.1363,
+      "step": 15696
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9867203601964425e-06,
+      "loss": 1.1069,
+      "step": 15697
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.985609993078905e-06,
+      "loss": 1.1998,
+      "step": 15698
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9844997961799814e-06,
+      "loss": 1.1072,
+      "step": 15699
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.983389769526619e-06,
+      "loss": 1.2226,
+      "step": 15700
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.982279913145751e-06,
+      "loss": 1.206,
+      "step": 15701
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.981170227064312e-06,
+      "loss": 1.2386,
+      "step": 15702
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9800607113092263e-06,
+      "loss": 1.0645,
+      "step": 15703
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9789513659074254e-06,
+      "loss": 1.0966,
+      "step": 15704
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9778421908858267e-06,
+      "loss": 1.0326,
+      "step": 15705
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.976733186271348e-06,
+      "loss": 1.081,
+      "step": 15706
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9756243520908957e-06,
+      "loss": 1.1822,
+      "step": 15707
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9745156883713856e-06,
+      "loss": 0.949,
+      "step": 15708
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9734071951397182e-06,
+      "loss": 0.9894,
+      "step": 15709
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.972298872422794e-06,
+      "loss": 1.0591,
+      "step": 15710
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9711907202475054e-06,
+      "loss": 1.1424,
+      "step": 15711
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9700827386407504e-06,
+      "loss": 1.1062,
+      "step": 15712
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9689749276294133e-06,
+      "loss": 1.1271,
+      "step": 15713
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.967867287240377e-06,
+      "loss": 1.1734,
+      "step": 15714
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9667598175005176e-06,
+      "loss": 1.199,
+      "step": 15715
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.965652518436716e-06,
+      "loss": 1.0622,
+      "step": 15716
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9645453900758415e-06,
+      "loss": 1.1201,
+      "step": 15717
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9634384324447596e-06,
+      "loss": 1.0402,
+      "step": 15718
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.96233164557033e-06,
+      "loss": 1.1203,
+      "step": 15719
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9612250294794166e-06,
+      "loss": 0.8911,
+      "step": 15720
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9601185841988722e-06,
+      "loss": 1.169,
+      "step": 15721
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9590123097555466e-06,
+      "loss": 1.156,
+      "step": 15722
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.957906206176282e-06,
+      "loss": 1.221,
+      "step": 15723
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9568002734879264e-06,
+      "loss": 1.1693,
+      "step": 15724
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9556945117173154e-06,
+      "loss": 1.0708,
+      "step": 15725
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9545889208912816e-06,
+      "loss": 1.1539,
+      "step": 15726
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9534835010366515e-06,
+      "loss": 1.115,
+      "step": 15727
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.952378252180258e-06,
+      "loss": 0.9884,
+      "step": 15728
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9512731743489165e-06,
+      "loss": 1.0894,
+      "step": 15729
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9501682675694466e-06,
+      "loss": 1.1217,
+      "step": 15730
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9490635318686567e-06,
+      "loss": 1.0816,
+      "step": 15731
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.947958967273362e-06,
+      "loss": 1.0729,
+      "step": 15732
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9468545738103636e-06,
+      "loss": 1.1178,
+      "step": 15733
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9457503515064624e-06,
+      "loss": 1.1049,
+      "step": 15734
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.944646300388452e-06,
+      "loss": 1.2704,
+      "step": 15735
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9435424204831285e-06,
+      "loss": 1.1319,
+      "step": 15736
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.942438711817279e-06,
+      "loss": 1.0284,
+      "step": 15737
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9413351744176855e-06,
+      "loss": 1.1376,
+      "step": 15738
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.940231808311127e-06,
+      "loss": 1.1976,
+      "step": 15739
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9391286135243823e-06,
+      "loss": 1.1459,
+      "step": 15740
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9380255900842213e-06,
+      "loss": 1.1601,
+      "step": 15741
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.936922738017407e-06,
+      "loss": 1.0609,
+      "step": 15742
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.93582005735071e-06,
+      "loss": 1.0769,
+      "step": 15743
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9347175481108857e-06,
+      "loss": 1.0893,
+      "step": 15744
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9336152103246873e-06,
+      "loss": 1.0701,
+      "step": 15745
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.932513044018863e-06,
+      "loss": 1.0842,
+      "step": 15746
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9314110492201665e-06,
+      "loss": 1.0347,
+      "step": 15747
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9303092259553346e-06,
+      "loss": 1.0655,
+      "step": 15748
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.929207574251107e-06,
+      "loss": 1.0952,
+      "step": 15749
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9281060941342143e-06,
+      "loss": 1.1623,
+      "step": 15750
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.927004785631389e-06,
+      "loss": 1.0991,
+      "step": 15751
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9259036487693604e-06,
+      "loss": 1.1049,
+      "step": 15752
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9248026835748444e-06,
+      "loss": 0.987,
+      "step": 15753
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9237018900745617e-06,
+      "loss": 1.1884,
+      "step": 15754
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9226012682952186e-06,
+      "loss": 1.0023,
+      "step": 15755
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.921500818263533e-06,
+      "loss": 1.224,
+      "step": 15756
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9204005400062052e-06,
+      "loss": 1.1132,
+      "step": 15757
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9193004335499365e-06,
+      "loss": 0.9176,
+      "step": 15758
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9182004989214187e-06,
+      "loss": 1.127,
+      "step": 15759
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9171007361473512e-06,
+      "loss": 1.2811,
+      "step": 15760
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.916001145254419e-06,
+      "loss": 1.1616,
+      "step": 15761
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9149017262693056e-06,
+      "loss": 1.136,
+      "step": 15762
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.913802479218688e-06,
+      "loss": 1.0952,
+      "step": 15763
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.912703404129248e-06,
+      "loss": 1.0722,
+      "step": 15764
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.911604501027653e-06,
+      "loss": 0.9938,
+      "step": 15765
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9105057699405704e-06,
+      "loss": 1.0637,
+      "step": 15766
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9094072108946592e-06,
+      "loss": 1.048,
+      "step": 15767
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9083088239165867e-06,
+      "loss": 1.029,
+      "step": 15768
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.907210609033003e-06,
+      "loss": 1.1887,
+      "step": 15769
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.906112566270558e-06,
+      "loss": 1.078,
+      "step": 15770
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9050146956558968e-06,
+      "loss": 1.0866,
+      "step": 15771
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9039169972156656e-06,
+      "loss": 1.0986,
+      "step": 15772
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9028194709765e-06,
+      "loss": 1.0968,
+      "step": 15773
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9017221169650334e-06,
+      "loss": 0.9898,
+      "step": 15774
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.900624935207893e-06,
+      "loss": 1.0143,
+      "step": 15775
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8995279257317097e-06,
+      "loss": 0.896,
+      "step": 15776
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.898431088563103e-06,
+      "loss": 1.1077,
+      "step": 15777
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8973344237286883e-06,
+      "loss": 1.309,
+      "step": 15778
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8962379312550746e-06,
+      "loss": 1.1118,
+      "step": 15779
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.89514161116888e-06,
+      "loss": 0.9828,
+      "step": 15780
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.894045463496703e-06,
+      "loss": 1.0369,
+      "step": 15781
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.892949488265141e-06,
+      "loss": 1.1165,
+      "step": 15782
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8918536855007975e-06,
+      "loss": 0.974,
+      "step": 15783
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8907580552302607e-06,
+      "loss": 1.1653,
+      "step": 15784
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8896625974801184e-06,
+      "loss": 1.2038,
+      "step": 15785
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8885673122769507e-06,
+      "loss": 1.1099,
+      "step": 15786
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8874721996473444e-06,
+      "loss": 1.0747,
+      "step": 15787
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8863772596178697e-06,
+      "loss": 1.1375,
+      "step": 15788
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.885282492215099e-06,
+      "loss": 1.2012,
+      "step": 15789
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8841878974655957e-06,
+      "loss": 1.1504,
+      "step": 15790
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.883093475395927e-06,
+      "loss": 0.9894,
+      "step": 15791
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.881999226032651e-06,
+      "loss": 1.1867,
+      "step": 15792
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.88090514940232e-06,
+      "loss": 1.0614,
+      "step": 15793
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.879811245531482e-06,
+      "loss": 1.0079,
+      "step": 15794
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8787175144466873e-06,
+      "loss": 1.1275,
+      "step": 15795
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8776239561744768e-06,
+      "loss": 1.1349,
+      "step": 15796
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.876530570741386e-06,
+      "loss": 1.0756,
+      "step": 15797
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.875437358173946e-06,
+      "loss": 1.1631,
+      "step": 15798
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.874344318498691e-06,
+      "loss": 1.0884,
+      "step": 15799
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.873251451742145e-06,
+      "loss": 1.1426,
+      "step": 15800
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8721587579308265e-06,
+      "loss": 1.1333,
+      "step": 15801
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.871066237091249e-06,
+      "loss": 1.2289,
+      "step": 15802
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.869973889249933e-06,
+      "loss": 1.0011,
+      "step": 15803
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8688817144333816e-06,
+      "loss": 1.2281,
+      "step": 15804
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8677897126680988e-06,
+      "loss": 1.1157,
+      "step": 15805
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8666978839805805e-06,
+      "loss": 1.1194,
+      "step": 15806
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8656062283973307e-06,
+      "loss": 1.1532,
+      "step": 15807
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8645147459448363e-06,
+      "loss": 1.0268,
+      "step": 15808
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.863423436649584e-06,
+      "loss": 1.1537,
+      "step": 15809
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.862332300538053e-06,
+      "loss": 1.1547,
+      "step": 15810
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.861241337636731e-06,
+      "loss": 0.9526,
+      "step": 15811
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.860150547972086e-06,
+      "loss": 1.1032,
+      "step": 15812
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8590599315705892e-06,
+      "loss": 1.1431,
+      "step": 15813
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.857969488458705e-06,
+      "loss": 1.2218,
+      "step": 15814
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8568792186629e-06,
+      "loss": 1.1329,
+      "step": 15815
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.85578912220963e-06,
+      "loss": 1.0165,
+      "step": 15816
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.854699199125347e-06,
+      "loss": 1.1201,
+      "step": 15817
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8536094494364976e-06,
+      "loss": 1.1728,
+      "step": 15818
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8525198731695334e-06,
+      "loss": 1.1422,
+      "step": 15819
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.851430470350892e-06,
+      "loss": 1.056,
+      "step": 15820
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8503412410070097e-06,
+      "loss": 1.165,
+      "step": 15821
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8492521851643164e-06,
+      "loss": 1.2608,
+      "step": 15822
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.848163302849245e-06,
+      "loss": 1.0964,
+      "step": 15823
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8470745940882182e-06,
+      "loss": 1.0224,
+      "step": 15824
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.845986058907655e-06,
+      "loss": 1.1192,
+      "step": 15825
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8448976973339682e-06,
+      "loss": 1.0526,
+      "step": 15826
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8438095093935746e-06,
+      "loss": 1.1118,
+      "step": 15827
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.842721495112879e-06,
+      "loss": 1.0879,
+      "step": 15828
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.84163365451828e-06,
+      "loss": 1.1533,
+      "step": 15829
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8405459876361842e-06,
+      "loss": 1.1895,
+      "step": 15830
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8394584944929813e-06,
+      "loss": 1.0018,
+      "step": 15831
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.838371175115062e-06,
+      "loss": 1.0141,
+      "step": 15832
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8372840295288106e-06,
+      "loss": 1.0864,
+      "step": 15833
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8361970577606113e-06,
+      "loss": 1.1492,
+      "step": 15834
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8351102598368442e-06,
+      "loss": 1.1712,
+      "step": 15835
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.834023635783879e-06,
+      "loss": 0.9313,
+      "step": 15836
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8329371856280865e-06,
+      "loss": 1.3149,
+      "step": 15837
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.831850909395828e-06,
+      "loss": 1.1106,
+      "step": 15838
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.83076480711347e-06,
+      "loss": 1.2685,
+      "step": 15839
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.829678878807366e-06,
+      "loss": 1.1231,
+      "step": 15840
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.828593124503869e-06,
+      "loss": 1.0709,
+      "step": 15841
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.827507544229322e-06,
+      "loss": 0.9603,
+      "step": 15842
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8264221380100778e-06,
+      "loss": 1.0332,
+      "step": 15843
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8253369058724713e-06,
+      "loss": 1.1627,
+      "step": 15844
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.824251847842838e-06,
+      "loss": 0.9645,
+      "step": 15845
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8231669639475068e-06,
+      "loss": 1.2283,
+      "step": 15846
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.822082254212809e-06,
+      "loss": 1.2103,
+      "step": 15847
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8209977186650663e-06,
+      "loss": 1.118,
+      "step": 15848
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.819913357330596e-06,
+      "loss": 1.1082,
+      "step": 15849
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8188291702357095e-06,
+      "loss": 1.0483,
+      "step": 15850
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.817745157406724e-06,
+      "loss": 1.0694,
+      "step": 15851
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.81666131886994e-06,
+      "loss": 1.261,
+      "step": 15852
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8155776546516612e-06,
+      "loss": 1.2357,
+      "step": 15853
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8144941647781797e-06,
+      "loss": 1.047,
+      "step": 15854
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.813410849275797e-06,
+      "loss": 1.1042,
+      "step": 15855
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8123277081707977e-06,
+      "loss": 1.1329,
+      "step": 15856
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.811244741489467e-06,
+      "loss": 1.0278,
+      "step": 15857
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.810161949258082e-06,
+      "loss": 1.0183,
+      "step": 15858
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8090793315029254e-06,
+      "loss": 1.056,
+      "step": 15859
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.807996888250265e-06,
+      "loss": 1.036,
+      "step": 15860
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8069146195263695e-06,
+      "loss": 1.0523,
+      "step": 15861
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.805832525357499e-06,
+      "loss": 1.0374,
+      "step": 15862
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.804750605769919e-06,
+      "loss": 1.1075,
+      "step": 15863
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.803668860789881e-06,
+      "loss": 0.9855,
+      "step": 15864
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.802587290443637e-06,
+      "loss": 1.2177,
+      "step": 15865
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.801505894757428e-06,
+      "loss": 0.9882,
+      "step": 15866
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8004246737575057e-06,
+      "loss": 1.1284,
+      "step": 15867
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7993436274701024e-06,
+      "loss": 1.111,
+      "step": 15868
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7982627559214504e-06,
+      "loss": 0.8193,
+      "step": 15869
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7971820591377842e-06,
+      "loss": 1.1379,
+      "step": 15870
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.796101537145327e-06,
+      "loss": 1.0247,
+      "step": 15871
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7950211899702996e-06,
+      "loss": 1.2014,
+      "step": 15872
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.793941017638915e-06,
+      "loss": 1.1133,
+      "step": 15873
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7928610201773933e-06,
+      "loss": 1.0444,
+      "step": 15874
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.791781197611939e-06,
+      "loss": 1.1656,
+      "step": 15875
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7907015499687573e-06,
+      "loss": 1.1283,
+      "step": 15876
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7896220772740434e-06,
+      "loss": 1.0552,
+      "step": 15877
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7885427795539997e-06,
+      "loss": 1.0906,
+      "step": 15878
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7874636568348144e-06,
+      "loss": 1.0953,
+      "step": 15879
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7863847091426755e-06,
+      "loss": 0.9717,
+      "step": 15880
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.785305936503763e-06,
+      "loss": 1.1448,
+      "step": 15881
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7842273389442586e-06,
+      "loss": 1.0327,
+      "step": 15882
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.783148916490338e-06,
+      "loss": 1.254,
+      "step": 15883
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7820706691681676e-06,
+      "loss": 1.1254,
+      "step": 15884
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.780992597003913e-06,
+      "loss": 1.1691,
+      "step": 15885
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7799147000237404e-06,
+      "loss": 0.9315,
+      "step": 15886
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7788369782538038e-06,
+      "loss": 1.1172,
+      "step": 15887
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7777594317202573e-06,
+      "loss": 1.1045,
+      "step": 15888
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7766820604492474e-06,
+      "loss": 1.0137,
+      "step": 15889
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.775604864466923e-06,
+      "loss": 1.1538,
+      "step": 15890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7745278437994227e-06,
+      "loss": 1.1377,
+      "step": 15891
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.773450998472883e-06,
+      "loss": 1.2797,
+      "step": 15892
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7723743285134307e-06,
+      "loss": 1.1231,
+      "step": 15893
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.771297833947202e-06,
+      "loss": 0.9489,
+      "step": 15894
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7702215148003153e-06,
+      "loss": 1.0283,
+      "step": 15895
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.76914537109889e-06,
+      "loss": 1.133,
+      "step": 15896
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7680694028690393e-06,
+      "loss": 0.9928,
+      "step": 15897
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.766993610136878e-06,
+      "loss": 1.1939,
+      "step": 15898
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.765917992928511e-06,
+      "loss": 1.0505,
+      "step": 15899
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7648425512700393e-06,
+      "loss": 1.1587,
+      "step": 15900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7637672851875584e-06,
+      "loss": 1.0961,
+      "step": 15901
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.762692194707167e-06,
+      "loss": 1.2048,
+      "step": 15902
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7616172798549523e-06,
+      "loss": 1.187,
+      "step": 15903
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7605425406569984e-06,
+      "loss": 1.2461,
+      "step": 15904
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.759467977139384e-06,
+      "loss": 1.1493,
+      "step": 15905
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.758393589328191e-06,
+      "loss": 1.008,
+      "step": 15906
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.757319377249489e-06,
+      "loss": 1.1066,
+      "step": 15907
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7562453409293456e-06,
+      "loss": 1.0878,
+      "step": 15908
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7551714803938214e-06,
+      "loss": 1.0635,
+      "step": 15909
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7540977956689817e-06,
+      "loss": 1.1683,
+      "step": 15910
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.75302428678088e-06,
+      "loss": 1.1357,
+      "step": 15911
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7519509537555665e-06,
+      "loss": 1.2315,
+      "step": 15912
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.750877796619084e-06,
+      "loss": 1.1014,
+      "step": 15913
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.749804815397481e-06,
+      "loss": 1.1702,
+      "step": 15914
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7487320101167935e-06,
+      "loss": 1.2188,
+      "step": 15915
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.747659380803053e-06,
+      "loss": 1.0993,
+      "step": 15916
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.746586927482289e-06,
+      "loss": 1.1762,
+      "step": 15917
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.745514650180533e-06,
+      "loss": 1.017,
+      "step": 15918
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7444425489238014e-06,
+      "loss": 1.052,
+      "step": 15919
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.743370623738112e-06,
+      "loss": 1.0537,
+      "step": 15920
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7422988746494726e-06,
+      "loss": 1.1381,
+      "step": 15921
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7412273016838986e-06,
+      "loss": 1.1576,
+      "step": 15922
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7401559048673907e-06,
+      "loss": 1.1702,
+      "step": 15923
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7390846842259487e-06,
+      "loss": 1.2044,
+      "step": 15924
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.738013639785565e-06,
+      "loss": 1.1441,
+      "step": 15925
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7369427715722362e-06,
+      "loss": 0.9402,
+      "step": 15926
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7358720796119465e-06,
+      "loss": 1.2887,
+      "step": 15927
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.734801563930679e-06,
+      "loss": 1.1724,
+      "step": 15928
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.733731224554408e-06,
+      "loss": 1.0291,
+      "step": 15929
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.732661061509113e-06,
+      "loss": 1.0097,
+      "step": 15930
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.731591074820763e-06,
+      "loss": 1.1135,
+      "step": 15931
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7305212645153213e-06,
+      "loss": 1.0849,
+      "step": 15932
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7294516306187457e-06,
+      "loss": 1.1505,
+      "step": 15933
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7283821731570014e-06,
+      "loss": 1.1032,
+      "step": 15934
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7273128921560365e-06,
+      "loss": 1.0845,
+      "step": 15935
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7262437876418e-06,
+      "loss": 1.0031,
+      "step": 15936
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7251748596402317e-06,
+      "loss": 1.1455,
+      "step": 15937
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.724106108177278e-06,
+      "loss": 1.1631,
+      "step": 15938
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7230375332788717e-06,
+      "loss": 0.9754,
+      "step": 15939
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7219691349709433e-06,
+      "loss": 1.0103,
+      "step": 15940
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7209009132794174e-06,
+      "loss": 1.0493,
+      "step": 15941
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7198328682302223e-06,
+      "loss": 1.1165,
+      "step": 15942
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7187649998492737e-06,
+      "loss": 0.9831,
+      "step": 15943
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.717697308162485e-06,
+      "loss": 1.1586,
+      "step": 15944
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7166297931957632e-06,
+      "loss": 1.081,
+      "step": 15945
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.715562454975019e-06,
+      "loss": 1.0719,
+      "step": 15946
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7144952935261517e-06,
+      "loss": 1.1259,
+      "step": 15947
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.713428308875058e-06,
+      "loss": 0.9568,
+      "step": 15948
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7123615010476266e-06,
+      "loss": 1.039,
+      "step": 15949
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7112948700697528e-06,
+      "loss": 0.9995,
+      "step": 15950
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7102284159673165e-06,
+      "loss": 1.1937,
+      "step": 15951
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7091621387661992e-06,
+      "loss": 1.2353,
+      "step": 15952
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7080960384922716e-06,
+      "loss": 1.215,
+      "step": 15953
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7070301151714117e-06,
+      "loss": 0.9504,
+      "step": 15954
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7059643688294824e-06,
+      "loss": 1.1369,
+      "step": 15955
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.704898799492345e-06,
+      "loss": 1.1163,
+      "step": 15956
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.703833407185862e-06,
+      "loss": 1.0073,
+      "step": 15957
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.702768191935885e-06,
+      "loss": 1.1494,
+      "step": 15958
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7017031537682638e-06,
+      "loss": 1.1064,
+      "step": 15959
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.700638292708839e-06,
+      "loss": 1.0916,
+      "step": 15960
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6995736087834614e-06,
+      "loss": 1.0078,
+      "step": 15961
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.698509102017961e-06,
+      "loss": 1.2347,
+      "step": 15962
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6974447724381727e-06,
+      "loss": 1.1351,
+      "step": 15963
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6963806200699206e-06,
+      "loss": 1.0779,
+      "step": 15964
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6953166449390344e-06,
+      "loss": 1.2145,
+      "step": 15965
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6942528470713315e-06,
+      "loss": 1.0867,
+      "step": 15966
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.693189226492625e-06,
+      "loss": 1.0564,
+      "step": 15967
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.692125783228726e-06,
+      "loss": 1.1314,
+      "step": 15968
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.691062517305445e-06,
+      "loss": 1.0285,
+      "step": 15969
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.689999428748582e-06,
+      "loss": 1.1259,
+      "step": 15970
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6889365175839344e-06,
+      "loss": 0.959,
+      "step": 15971
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6878737838372946e-06,
+      "loss": 1.0151,
+      "step": 15972
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.686811227534456e-06,
+      "loss": 1.181,
+      "step": 15973
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.685748848701202e-06,
+      "loss": 1.1752,
+      "step": 15974
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6846866473633126e-06,
+      "loss": 1.1012,
+      "step": 15975
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6836246235465625e-06,
+      "loss": 1.2161,
+      "step": 15976
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.682562777276728e-06,
+      "loss": 1.1604,
+      "step": 15977
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.681501108579575e-06,
+      "loss": 1.0896,
+      "step": 15978
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6804396174808676e-06,
+      "loss": 1.0282,
+      "step": 15979
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6793783040063627e-06,
+      "loss": 1.2288,
+      "step": 15980
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.678317168181819e-06,
+      "loss": 1.0119,
+      "step": 15981
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.677256210032986e-06,
+      "loss": 1.1952,
+      "step": 15982
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6761954295856087e-06,
+      "loss": 0.9974,
+      "step": 15983
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6751348268654266e-06,
+      "loss": 1.0831,
+      "step": 15984
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6740744018981846e-06,
+      "loss": 1.133,
+      "step": 15985
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.673014154709611e-06,
+      "loss": 1.1558,
+      "step": 15986
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6719540853254366e-06,
+      "loss": 1.1659,
+      "step": 15987
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.670894193771383e-06,
+      "loss": 1.1714,
+      "step": 15988
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6698344800731756e-06,
+      "loss": 1.1737,
+      "step": 15989
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.668774944256528e-06,
+      "loss": 1.0261,
+      "step": 15990
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6677155863471525e-06,
+      "loss": 1.1375,
+      "step": 15991
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.666656406370752e-06,
+      "loss": 1.0207,
+      "step": 15992
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6655974043530385e-06,
+      "loss": 0.9945,
+      "step": 15993
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6645385803197065e-06,
+      "loss": 1.1584,
+      "step": 15994
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.663479934296449e-06,
+      "loss": 1.3168,
+      "step": 15995
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6624214663089555e-06,
+      "loss": 1.1847,
+      "step": 15996
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6613631763829164e-06,
+      "loss": 0.9523,
+      "step": 15997
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.660305064544011e-06,
+      "loss": 0.9714,
+      "step": 15998
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6592471308179136e-06,
+      "loss": 1.1847,
+      "step": 15999
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6581893752303e-06,
+      "loss": 1.0586,
+      "step": 16000
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6571317978068425e-06,
+      "loss": 1.0937,
+      "step": 16001
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6560743985731994e-06,
+      "loss": 1.0045,
+      "step": 16002
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.655017177555035e-06,
+      "loss": 1.0651,
+      "step": 16003
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6539601347779997e-06,
+      "loss": 1.0375,
+      "step": 16004
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6529032702677503e-06,
+      "loss": 1.2268,
+      "step": 16005
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.651846584049932e-06,
+      "loss": 1.106,
+      "step": 16006
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.650790076150186e-06,
+      "loss": 1.0991,
+      "step": 16007
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6497337465941496e-06,
+      "loss": 1.1678,
+      "step": 16008
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6486775954074617e-06,
+      "loss": 1.0495,
+      "step": 16009
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.647621622615749e-06,
+      "loss": 1.2112,
+      "step": 16010
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.646565828244637e-06,
+      "loss": 1.1134,
+      "step": 16011
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6455102123197428e-06,
+      "loss": 1.0461,
+      "step": 16012
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6444547748666916e-06,
+      "loss": 1.0563,
+      "step": 16013
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6433995159110902e-06,
+      "loss": 1.1958,
+      "step": 16014
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.642344435478548e-06,
+      "loss": 1.0675,
+      "step": 16015
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.641289533594664e-06,
+      "loss": 1.0986,
+      "step": 16016
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.640234810285046e-06,
+      "loss": 1.0873,
+      "step": 16017
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6391802655752853e-06,
+      "loss": 1.0151,
+      "step": 16018
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.638125899490971e-06,
+      "loss": 0.9864,
+      "step": 16019
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6370717120576885e-06,
+      "loss": 1.0194,
+      "step": 16020
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.636017703301024e-06,
+      "loss": 1.1276,
+      "step": 16021
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6349638732465534e-06,
+      "loss": 0.9505,
+      "step": 16022
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.63391022191985e-06,
+      "loss": 1.1356,
+      "step": 16023
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6328567493464796e-06,
+      "loss": 1.0519,
+      "step": 16024
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6318034555520134e-06,
+      "loss": 1.1459,
+      "step": 16025
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.630750340562008e-06,
+      "loss": 1.1684,
+      "step": 16026
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6296974044020195e-06,
+      "loss": 1.1198,
+      "step": 16027
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.628644647097597e-06,
+      "loss": 1.114,
+      "step": 16028
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.627592068674293e-06,
+      "loss": 1.1533,
+      "step": 16029
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6265396691576485e-06,
+      "loss": 1.0767,
+      "step": 16030
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6254874485732006e-06,
+      "loss": 1.0399,
+      "step": 16031
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.624435406946483e-06,
+      "loss": 1.1004,
+      "step": 16032
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6233835443030296e-06,
+      "loss": 1.2755,
+      "step": 16033
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6223318606683647e-06,
+      "loss": 1.1438,
+      "step": 16034
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.621280356068008e-06,
+      "loss": 1.1777,
+      "step": 16035
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6202290305274737e-06,
+      "loss": 1.2225,
+      "step": 16036
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6191778840722804e-06,
+      "loss": 1.0237,
+      "step": 16037
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6181269167279344e-06,
+      "loss": 1.1494,
+      "step": 16038
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.617076128519935e-06,
+      "loss": 1.0091,
+      "step": 16039
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6160255194737895e-06,
+      "loss": 1.1582,
+      "step": 16040
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6149750896149884e-06,
+      "loss": 0.9529,
+      "step": 16041
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.613924838969024e-06,
+      "loss": 1.0519,
+      "step": 16042
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6128747675613788e-06,
+      "loss": 1.0572,
+      "step": 16043
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6118248754175413e-06,
+      "loss": 1.2359,
+      "step": 16044
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6107751625629863e-06,
+      "loss": 1.2262,
+      "step": 16045
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6097256290231866e-06,
+      "loss": 1.064,
+      "step": 16046
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6086762748236093e-06,
+      "loss": 1.0292,
+      "step": 16047
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6076270999897245e-06,
+      "loss": 1.2913,
+      "step": 16048
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6065781045469897e-06,
+      "loss": 0.9864,
+      "step": 16049
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6055292885208615e-06,
+      "loss": 1.0866,
+      "step": 16050
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6044806519367884e-06,
+      "loss": 1.0268,
+      "step": 16051
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6034321948202223e-06,
+      "loss": 0.9406,
+      "step": 16052
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6023839171966047e-06,
+      "loss": 1.1719,
+      "step": 16053
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.601335819091374e-06,
+      "loss": 1.1125,
+      "step": 16054
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.600287900529962e-06,
+      "loss": 1.0852,
+      "step": 16055
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5992401615378027e-06,
+      "loss": 1.1241,
+      "step": 16056
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5981926021403203e-06,
+      "loss": 1.0955,
+      "step": 16057
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.597145222362936e-06,
+      "loss": 0.9171,
+      "step": 16058
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.596098022231063e-06,
+      "loss": 1.0178,
+      "step": 16059
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5950510017701194e-06,
+      "loss": 1.0524,
+      "step": 16060
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.594004161005511e-06,
+      "loss": 1.1584,
+      "step": 16061
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5929574999626416e-06,
+      "loss": 1.0954,
+      "step": 16062
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5919110186669072e-06,
+      "loss": 1.126,
+      "step": 16063
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5908647171437083e-06,
+      "loss": 1.2063,
+      "step": 16064
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5898185954184342e-06,
+      "loss": 1.2561,
+      "step": 16065
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.58877265351647e-06,
+      "loss": 1.2134,
+      "step": 16066
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.587726891463195e-06,
+      "loss": 1.1455,
+      "step": 16067
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5866813092839915e-06,
+      "loss": 1.1292,
+      "step": 16068
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5856359070042324e-06,
+      "loss": 1.1479,
+      "step": 16069
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.584590684649284e-06,
+      "loss": 1.1329,
+      "step": 16070
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5835456422445093e-06,
+      "loss": 1.1287,
+      "step": 16071
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5825007798152737e-06,
+      "loss": 1.0901,
+      "step": 16072
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.58145609738693e-06,
+      "loss": 0.9955,
+      "step": 16073
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5804115949848298e-06,
+      "loss": 1.1118,
+      "step": 16074
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.579367272634317e-06,
+      "loss": 1.2341,
+      "step": 16075
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5783231303607393e-06,
+      "loss": 1.0908,
+      "step": 16076
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.577279168189434e-06,
+      "loss": 1.1869,
+      "step": 16077
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5762353861457335e-06,
+      "loss": 1.1707,
+      "step": 16078
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5751917842549647e-06,
+      "loss": 1.0374,
+      "step": 16079
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.574148362542458e-06,
+      "loss": 1.205,
+      "step": 16080
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.573105121033532e-06,
+      "loss": 1.1576,
+      "step": 16081
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.572062059753504e-06,
+      "loss": 1.1279,
+      "step": 16082
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5710191787276783e-06,
+      "loss": 1.1567,
+      "step": 16083
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5699764779813763e-06,
+      "loss": 1.1209,
+      "step": 16084
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.568933957539893e-06,
+      "loss": 1.1891,
+      "step": 16085
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5678916174285296e-06,
+      "loss": 1.0871,
+      "step": 16086
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5668494576725754e-06,
+      "loss": 1.0752,
+      "step": 16087
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5658074782973288e-06,
+      "loss": 1.2003,
+      "step": 16088
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.564765679328071e-06,
+      "loss": 1.0562,
+      "step": 16089
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5637240607900837e-06,
+      "loss": 0.9025,
+      "step": 16090
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.562682622708642e-06,
+      "loss": 1.2168,
+      "step": 16091
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.561641365109023e-06,
+      "loss": 1.1462,
+      "step": 16092
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5606002880164925e-06,
+      "loss": 1.0913,
+      "step": 16093
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5595593914563155e-06,
+      "loss": 1.1047,
+      "step": 16094
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.558518675453746e-06,
+      "loss": 1.1228,
+      "step": 16095
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5574781400340477e-06,
+      "loss": 1.1525,
+      "step": 16096
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5564377852224676e-06,
+      "loss": 1.0392,
+      "step": 16097
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.555397611044251e-06,
+      "loss": 1.0733,
+      "step": 16098
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5543576175246378e-06,
+      "loss": 1.1846,
+      "step": 16099
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.553317804688872e-06,
+      "loss": 1.0358,
+      "step": 16100
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5522781725621814e-06,
+      "loss": 0.9669,
+      "step": 16101
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5512387211697988e-06,
+      "loss": 0.931,
+      "step": 16102
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.550199450536942e-06,
+      "loss": 1.1062,
+      "step": 16103
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5491603606888384e-06,
+      "loss": 1.2445,
+      "step": 16104
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5481214516507015e-06,
+      "loss": 1.1452,
+      "step": 16105
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5470827234477414e-06,
+      "loss": 1.0157,
+      "step": 16106
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5460441761051625e-06,
+      "loss": 1.2153,
+      "step": 16107
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5450058096481733e-06,
+      "loss": 1.2295,
+      "step": 16108
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5439676241019684e-06,
+      "loss": 1.0656,
+      "step": 16109
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.542929619491742e-06,
+      "loss": 0.9761,
+      "step": 16110
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.54189179584268e-06,
+      "loss": 1.1451,
+      "step": 16111
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5408541531799735e-06,
+      "loss": 0.9924,
+      "step": 16112
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5398166915288004e-06,
+      "loss": 1.0254,
+      "step": 16113
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.538779410914337e-06,
+      "loss": 1.1232,
+      "step": 16114
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.537742311361752e-06,
+      "loss": 1.1422,
+      "step": 16115
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.536705392896217e-06,
+      "loss": 1.1794,
+      "step": 16116
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5356686555428935e-06,
+      "loss": 1.0969,
+      "step": 16117
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5346320993269402e-06,
+      "loss": 0.8766,
+      "step": 16118
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5335957242735086e-06,
+      "loss": 1.1835,
+      "step": 16119
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5325595304077532e-06,
+      "loss": 0.9109,
+      "step": 16120
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.531523517754818e-06,
+      "loss": 0.9001,
+      "step": 16121
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.530487686339842e-06,
+      "loss": 0.9457,
+      "step": 16122
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5294520361879606e-06,
+      "loss": 1.1726,
+      "step": 16123
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.528416567324311e-06,
+      "loss": 1.085,
+      "step": 16124
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.527381279774017e-06,
+      "loss": 0.9671,
+      "step": 16125
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.526346173562202e-06,
+      "loss": 1.0667,
+      "step": 16126
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5253112487139873e-06,
+      "loss": 0.9695,
+      "step": 16127
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.524276505254488e-06,
+      "loss": 1.2548,
+      "step": 16128
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.523241943208812e-06,
+      "loss": 1.0865,
+      "step": 16129
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5222075626020624e-06,
+      "loss": 1.2508,
+      "step": 16130
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.521173363459347e-06,
+      "loss": 1.1815,
+      "step": 16131
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.52013934580576e-06,
+      "loss": 1.1392,
+      "step": 16132
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.519105509666393e-06,
+      "loss": 1.1909,
+      "step": 16133
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5180718550663317e-06,
+      "loss": 1.2036,
+      "step": 16134
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5170383820306667e-06,
+      "loss": 1.2387,
+      "step": 16135
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5160050905844724e-06,
+      "loss": 1.1275,
+      "step": 16136
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.514971980752826e-06,
+      "loss": 0.9781,
+      "step": 16137
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5139390525607934e-06,
+      "loss": 1.0747,
+      "step": 16138
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.512906306033448e-06,
+      "loss": 1.0557,
+      "step": 16139
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.511873741195847e-06,
+      "loss": 1.1733,
+      "step": 16140
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5108413580730486e-06,
+      "loss": 1.1285,
+      "step": 16141
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.509809156690103e-06,
+      "loss": 0.9917,
+      "step": 16142
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.508777137072064e-06,
+      "loss": 1.0258,
+      "step": 16143
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5077452992439733e-06,
+      "loss": 1.0305,
+      "step": 16144
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.506713643230869e-06,
+      "loss": 1.1056,
+      "step": 16145
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.505682169057786e-06,
+      "loss": 0.9803,
+      "step": 16146
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.50465087674976e-06,
+      "loss": 1.0591,
+      "step": 16147
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.503619766331813e-06,
+      "loss": 1.0823,
+      "step": 16148
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5025888378289687e-06,
+      "loss": 1.0749,
+      "step": 16149
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.501558091266242e-06,
+      "loss": 1.1695,
+      "step": 16150
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.50052752666865e-06,
+      "loss": 1.2896,
+      "step": 16151
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4994971440612004e-06,
+      "loss": 1.0084,
+      "step": 16152
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.498466943468897e-06,
+      "loss": 1.0482,
+      "step": 16153
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4974369249167375e-06,
+      "loss": 1.134,
+      "step": 16154
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.496407088429721e-06,
+      "loss": 1.1623,
+      "step": 16155
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4953774340328373e-06,
+      "loss": 1.1269,
+      "step": 16156
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4943479617510745e-06,
+      "loss": 1.1445,
+      "step": 16157
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4933186716094093e-06,
+      "loss": 1.1484,
+      "step": 16158
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4922895636328257e-06,
+      "loss": 1.1477,
+      "step": 16159
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.491260637846297e-06,
+      "loss": 1.0728,
+      "step": 16160
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4902318942747883e-06,
+      "loss": 1.0861,
+      "step": 16161
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.489203332943264e-06,
+      "loss": 1.2271,
+      "step": 16162
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4881749538766895e-06,
+      "loss": 1.1352,
+      "step": 16163
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.487146757100017e-06,
+      "loss": 1.1103,
+      "step": 16164
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4861187426381993e-06,
+      "loss": 1.0729,
+      "step": 16165
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4850909105161747e-06,
+      "loss": 0.8009,
+      "step": 16166
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4840632607589e-06,
+      "loss": 1.168,
+      "step": 16167
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4830357933913063e-06,
+      "loss": 1.0487,
+      "step": 16168
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4820085084383273e-06,
+      "loss": 1.1365,
+      "step": 16169
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4809814059248882e-06,
+      "loss": 1.0373,
+      "step": 16170
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4799544858759218e-06,
+      "loss": 1.1003,
+      "step": 16171
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4789277483163443e-06,
+      "loss": 1.0546,
+      "step": 16172
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4779011932710717e-06,
+      "loss": 1.0456,
+      "step": 16173
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.476874820765012e-06,
+      "loss": 1.1603,
+      "step": 16174
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.475848630823079e-06,
+      "loss": 0.9317,
+      "step": 16175
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4748226234701723e-06,
+      "loss": 1.0635,
+      "step": 16176
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4737967987311893e-06,
+      "loss": 1.0981,
+      "step": 16177
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4727711566310207e-06,
+      "loss": 1.1735,
+      "step": 16178
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.471745697194563e-06,
+      "loss": 1.1328,
+      "step": 16179
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.470720420446697e-06,
+      "loss": 1.1407,
+      "step": 16180
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4696953264123048e-06,
+      "loss": 0.9648,
+      "step": 16181
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4686704151162575e-06,
+      "loss": 1.1089,
+      "step": 16182
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.467645686583433e-06,
+      "loss": 1.1419,
+      "step": 16183
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4666211408386963e-06,
+      "loss": 1.0541,
+      "step": 16184
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.465596777906909e-06,
+      "loss": 1.0006,
+      "step": 16185
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4645725978129277e-06,
+      "loss": 1.114,
+      "step": 16186
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4635486005816113e-06,
+      "loss": 1.1344,
+      "step": 16187
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4625247862378067e-06,
+      "loss": 1.0951,
+      "step": 16188
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4615011548063582e-06,
+      "loss": 1.0058,
+      "step": 16189
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.460477706312103e-06,
+      "loss": 1.2646,
+      "step": 16190
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.459454440779886e-06,
+      "loss": 1.0802,
+      "step": 16191
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.458431358234532e-06,
+      "loss": 1.2123,
+      "step": 16192
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4574084587008706e-06,
+      "loss": 1.1345,
+      "step": 16193
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.456385742203721e-06,
+      "loss": 1.0526,
+      "step": 16194
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.455363208767908e-06,
+      "loss": 1.1742,
+      "step": 16195
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4543408584182404e-06,
+      "loss": 1.0783,
+      "step": 16196
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4533186911795303e-06,
+      "loss": 1.0875,
+      "step": 16197
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4522967070765792e-06,
+      "loss": 0.973,
+      "step": 16198
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.451274906134191e-06,
+      "loss": 1.1938,
+      "step": 16199
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.450253288377162e-06,
+      "loss": 0.9539,
+      "step": 16200
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4492318538302837e-06,
+      "loss": 1.1275,
+      "step": 16201
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4482106025183383e-06,
+      "loss": 1.0138,
+      "step": 16202
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4471895344661154e-06,
+      "loss": 1.0685,
+      "step": 16203
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4461686496983906e-06,
+      "loss": 1.1421,
+      "step": 16204
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4451479482399377e-06,
+      "loss": 1.0226,
+      "step": 16205
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.444127430115523e-06,
+      "loss": 1.087,
+      "step": 16206
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.443107095349918e-06,
+      "loss": 1.1084,
+      "step": 16207
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4420869439678796e-06,
+      "loss": 0.9854,
+      "step": 16208
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.441066975994163e-06,
+      "loss": 1.1429,
+      "step": 16209
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4400471914535193e-06,
+      "loss": 1.2744,
+      "step": 16210
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4390275903706995e-06,
+      "loss": 1.1221,
+      "step": 16211
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.438008172770444e-06,
+      "loss": 1.1469,
+      "step": 16212
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.436988938677489e-06,
+      "loss": 1.1482,
+      "step": 16213
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4359698881165726e-06,
+      "loss": 0.997,
+      "step": 16214
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.434951021112423e-06,
+      "loss": 1.0939,
+      "step": 16215
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4339323376897626e-06,
+      "loss": 1.1241,
+      "step": 16216
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.432913837873311e-06,
+      "loss": 1.2568,
+      "step": 16217
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.431895521687789e-06,
+      "loss": 1.1598,
+      "step": 16218
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.430877389157907e-06,
+      "loss": 1.2325,
+      "step": 16219
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4298594403083696e-06,
+      "loss": 1.085,
+      "step": 16220
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.428841675163879e-06,
+      "loss": 0.9568,
+      "step": 16221
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.427824093749137e-06,
+      "loss": 1.1404,
+      "step": 16222
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4268066960888346e-06,
+      "loss": 1.1969,
+      "step": 16223
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.425789482207663e-06,
+      "loss": 1.1718,
+      "step": 16224
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4247724521303017e-06,
+      "loss": 1.1274,
+      "step": 16225
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.423755605881439e-06,
+      "loss": 0.8886,
+      "step": 16226
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4227389434857463e-06,
+      "loss": 0.9941,
+      "step": 16227
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.421722464967895e-06,
+      "loss": 1.099,
+      "step": 16228
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4207061703525505e-06,
+      "loss": 1.0425,
+      "step": 16229
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4196900596643804e-06,
+      "loss": 1.0266,
+      "step": 16230
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4186741329280394e-06,
+      "loss": 1.1592,
+      "step": 16231
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4176583901681803e-06,
+      "loss": 1.1463,
+      "step": 16232
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4166428314094514e-06,
+      "loss": 0.9867,
+      "step": 16233
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4156274566765014e-06,
+      "loss": 1.0143,
+      "step": 16234
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4146122659939687e-06,
+      "loss": 1.1919,
+      "step": 16235
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.413597259386489e-06,
+      "loss": 1.0202,
+      "step": 16236
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.412582436878689e-06,
+      "loss": 1.0939,
+      "step": 16237
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4115677984952023e-06,
+      "loss": 0.959,
+      "step": 16238
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.41055334426065e-06,
+      "loss": 0.9453,
+      "step": 16239
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.409539074199646e-06,
+      "loss": 1.0772,
+      "step": 16240
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4085249883368043e-06,
+      "loss": 1.0964,
+      "step": 16241
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.407511086696739e-06,
+      "loss": 1.1861,
+      "step": 16242
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4064973693040495e-06,
+      "loss": 0.9974,
+      "step": 16243
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4054838361833376e-06,
+      "loss": 0.9997,
+      "step": 16244
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4044704873591964e-06,
+      "loss": 1.1104,
+      "step": 16245
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4034573228562206e-06,
+      "loss": 1.0314,
+      "step": 16246
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.402444342698995e-06,
+      "loss": 1.2687,
+      "step": 16247
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.401431546912102e-06,
+      "loss": 0.9257,
+      "step": 16248
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4004189355201156e-06,
+      "loss": 1.025,
+      "step": 16249
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3994065085476125e-06,
+      "loss": 1.2764,
+      "step": 16250
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3983942660191628e-06,
+      "loss": 1.2432,
+      "step": 16251
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.39738220795933e-06,
+      "loss": 1.0815,
+      "step": 16252
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.396370334392669e-06,
+      "loss": 1.0578,
+      "step": 16253
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.395358645343742e-06,
+      "loss": 1.1447,
+      "step": 16254
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3943471408370943e-06,
+      "loss": 1.0584,
+      "step": 16255
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3933358208972755e-06,
+      "loss": 1.1289,
+      "step": 16256
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3923246855488236e-06,
+      "loss": 1.0433,
+      "step": 16257
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3913137348162805e-06,
+      "loss": 1.1776,
+      "step": 16258
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3903029687241763e-06,
+      "loss": 1.0961,
+      "step": 16259
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3892923872970406e-06,
+      "loss": 1.1448,
+      "step": 16260
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.388281990559392e-06,
+      "loss": 1.1203,
+      "step": 16261
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3872717785357593e-06,
+      "loss": 1.1331,
+      "step": 16262
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3862617512506504e-06,
+      "loss": 1.1155,
+      "step": 16263
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3852519087285787e-06,
+      "loss": 1.1209,
+      "step": 16264
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3842422509940454e-06,
+      "loss": 1.0615,
+      "step": 16265
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.383232778071559e-06,
+      "loss": 1.137,
+      "step": 16266
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3822234899856133e-06,
+      "loss": 0.9447,
+      "step": 16267
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3812143867606996e-06,
+      "loss": 1.126,
+      "step": 16268
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.380205468421304e-06,
+      "loss": 1.1558,
+      "step": 16269
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3791967349919152e-06,
+      "loss": 1.0387,
+      "step": 16270
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3781881864970102e-06,
+      "loss": 1.1649,
+      "step": 16271
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3771798229610632e-06,
+      "loss": 0.9291,
+      "step": 16272
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3761716444085404e-06,
+      "loss": 1.0994,
+      "step": 16273
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.375163650863914e-06,
+      "loss": 1.239,
+      "step": 16274
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.374155842351643e-06,
+      "loss": 0.9978,
+      "step": 16275
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.373148218896182e-06,
+      "loss": 1.249,
+      "step": 16276
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.372140780521981e-06,
+      "loss": 1.2014,
+      "step": 16277
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.371133527253493e-06,
+      "loss": 1.0225,
+      "step": 16278
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3701264591151596e-06,
+      "loss": 1.0977,
+      "step": 16279
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3691195761314178e-06,
+      "loss": 1.0941,
+      "step": 16280
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3681128783266995e-06,
+      "loss": 1.0562,
+      "step": 16281
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.36710636572544e-06,
+      "loss": 1.0975,
+      "step": 16282
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3661000383520616e-06,
+      "loss": 0.8986,
+      "step": 16283
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3650938962309845e-06,
+      "loss": 1.169,
+      "step": 16284
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.364087939386622e-06,
+      "loss": 0.9982,
+      "step": 16285
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3630821678433925e-06,
+      "loss": 1.0707,
+      "step": 16286
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3620765816256984e-06,
+      "loss": 1.0737,
+      "step": 16287
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3610711807579443e-06,
+      "loss": 1.173,
+      "step": 16288
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3600659652645243e-06,
+      "loss": 1.0294,
+      "step": 16289
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3590609351698367e-06,
+      "loss": 1.1062,
+      "step": 16290
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3580560904982706e-06,
+      "loss": 0.9991,
+      "step": 16291
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3570514312742087e-06,
+      "loss": 1.2083,
+      "step": 16292
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.356046957522028e-06,
+      "loss": 1.1068,
+      "step": 16293
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.355042669266111e-06,
+      "loss": 1.0106,
+      "step": 16294
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.354038566530825e-06,
+      "loss": 1.1098,
+      "step": 16295
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3530346493405343e-06,
+      "loss": 1.0798,
+      "step": 16296
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3520309177196065e-06,
+      "loss": 1.2382,
+      "step": 16297
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3510273716923972e-06,
+      "loss": 1.0373,
+      "step": 16298
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3500240112832575e-06,
+      "loss": 1.0362,
+      "step": 16299
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3490208365165356e-06,
+      "loss": 1.0923,
+      "step": 16300
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3480178474165795e-06,
+      "loss": 1.1007,
+      "step": 16301
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3470150440077267e-06,
+      "loss": 0.9781,
+      "step": 16302
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3460124263143124e-06,
+      "loss": 1.1206,
+      "step": 16303
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.345009994360663e-06,
+      "loss": 1.1683,
+      "step": 16304
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.344007748171113e-06,
+      "loss": 1.0732,
+      "step": 16305
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.343005687769978e-06,
+      "loss": 1.275,
+      "step": 16306
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.342003813181578e-06,
+      "loss": 1.1158,
+      "step": 16307
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3410021244302205e-06,
+      "loss": 1.1799,
+      "step": 16308
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3400006215402195e-06,
+      "loss": 1.1264,
+      "step": 16309
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.338999304535877e-06,
+      "loss": 1.159,
+      "step": 16310
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3379981734414913e-06,
+      "loss": 1.222,
+      "step": 16311
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3369972282813534e-06,
+      "loss": 1.1698,
+      "step": 16312
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3359964690797587e-06,
+      "loss": 1.2367,
+      "step": 16313
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3349958958609918e-06,
+      "loss": 1.243,
+      "step": 16314
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3339955086493327e-06,
+      "loss": 1.1439,
+      "step": 16315
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3329953074690546e-06,
+      "loss": 1.1322,
+      "step": 16316
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3319952923444355e-06,
+      "loss": 1.1024,
+      "step": 16317
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.330995463299739e-06,
+      "loss": 1.1117,
+      "step": 16318
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.32999582035923e-06,
+      "loss": 1.1636,
+      "step": 16319
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3289963635471622e-06,
+      "loss": 1.2317,
+      "step": 16320
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.327997092887796e-06,
+      "loss": 1.119,
+      "step": 16321
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.326998008405378e-06,
+      "loss": 1.1037,
+      "step": 16322
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3259991101241532e-06,
+      "loss": 1.0076,
+      "step": 16323
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.325000398068358e-06,
+      "loss": 1.1467,
+      "step": 16324
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3240018722622347e-06,
+      "loss": 1.2567,
+      "step": 16325
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3230035327300125e-06,
+      "loss": 1.0241,
+      "step": 16326
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3220053794959163e-06,
+      "loss": 1.1975,
+      "step": 16327
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3210074125841674e-06,
+      "loss": 1.0682,
+      "step": 16328
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.320009632018988e-06,
+      "loss": 1.1913,
+      "step": 16329
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3190120378245884e-06,
+      "loss": 1.0948,
+      "step": 16330
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.318014630025178e-06,
+      "loss": 0.9956,
+      "step": 16331
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.317017408644957e-06,
+      "loss": 1.214,
+      "step": 16332
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3160203737081297e-06,
+      "loss": 1.0226,
+      "step": 16333
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.315023525238893e-06,
+      "loss": 1.0521,
+      "step": 16334
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3140268632614337e-06,
+      "loss": 1.1933,
+      "step": 16335
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.313030387799937e-06,
+      "loss": 0.969,
+      "step": 16336
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.312034098878588e-06,
+      "loss": 1.0832,
+      "step": 16337
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3110379965215637e-06,
+      "loss": 1.1847,
+      "step": 16338
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.310042080753033e-06,
+      "loss": 1.15,
+      "step": 16339
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.309046351597164e-06,
+      "loss": 1.106,
+      "step": 16340
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.308050809078124e-06,
+      "loss": 0.7862,
+      "step": 16341
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3070554532200693e-06,
+      "loss": 1.2188,
+      "step": 16342
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3060602840471558e-06,
+      "loss": 1.1235,
+      "step": 16343
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3050653015835278e-06,
+      "loss": 1.1575,
+      "step": 16344
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3040705058533384e-06,
+      "loss": 1.1447,
+      "step": 16345
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.303075896880725e-06,
+      "loss": 1.2352,
+      "step": 16346
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3020814746898248e-06,
+      "loss": 1.2097,
+      "step": 16347
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3010872393047644e-06,
+      "loss": 1.189,
+      "step": 16348
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.300093190749677e-06,
+      "loss": 1.1649,
+      "step": 16349
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.299099329048685e-06,
+      "loss": 1.1316,
+      "step": 16350
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2981056542259037e-06,
+      "loss": 1.1303,
+      "step": 16351
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.297112166305445e-06,
+      "loss": 1.1699,
+      "step": 16352
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.296118865311424e-06,
+      "loss": 1.061,
+      "step": 16353
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2951257512679413e-06,
+      "loss": 1.1058,
+      "step": 16354
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.294132824199098e-06,
+      "loss": 1.0544,
+      "step": 16355
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.293140084128985e-06,
+      "loss": 1.0168,
+      "step": 16356
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2921475310817008e-06,
+      "loss": 0.9954,
+      "step": 16357
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.291155165081328e-06,
+      "loss": 1.0295,
+      "step": 16358
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.290162986151949e-06,
+      "loss": 1.1982,
+      "step": 16359
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2891709943176366e-06,
+      "loss": 1.0015,
+      "step": 16360
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2881791896024706e-06,
+      "loss": 1.0635,
+      "step": 16361
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2871875720305158e-06,
+      "loss": 1.1503,
+      "step": 16362
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.286196141625837e-06,
+      "loss": 1.0919,
+      "step": 16363
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2852048984124885e-06,
+      "loss": 1.202,
+      "step": 16364
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.284213842414532e-06,
+      "loss": 1.1964,
+      "step": 16365
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.283222973656014e-06,
+      "loss": 1.0442,
+      "step": 16366
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.282232292160981e-06,
+      "loss": 1.1409,
+      "step": 16367
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2812417979534696e-06,
+      "loss": 1.1009,
+      "step": 16368
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2802514910575223e-06,
+      "loss": 1.0272,
+      "step": 16369
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.279261371497169e-06,
+      "loss": 1.3455,
+      "step": 16370
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2782714392964354e-06,
+      "loss": 1.1734,
+      "step": 16371
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.277281694479343e-06,
+      "loss": 1.0849,
+      "step": 16372
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2762921370699152e-06,
+      "loss": 1.0312,
+      "step": 16373
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2753027670921623e-06,
+      "loss": 1.0284,
+      "step": 16374
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2743135845700935e-06,
+      "loss": 1.0026,
+      "step": 16375
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2733245895277113e-06,
+      "loss": 1.1775,
+      "step": 16376
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2723357819890214e-06,
+      "loss": 1.0469,
+      "step": 16377
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2713471619780148e-06,
+      "loss": 1.023,
+      "step": 16378
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2703587295186836e-06,
+      "loss": 1.0849,
+      "step": 16379
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2693704846350108e-06,
+      "loss": 1.0607,
+      "step": 16380
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2683824273509847e-06,
+      "loss": 1.1461,
+      "step": 16381
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.267394557690579e-06,
+      "loss": 1.1457,
+      "step": 16382
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2664068756777637e-06,
+      "loss": 1.106,
+      "step": 16383
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2654193813365133e-06,
+      "loss": 1.1793,
+      "step": 16384
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2644320746907867e-06,
+      "loss": 1.0229,
+      "step": 16385
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2634449557645454e-06,
+      "loss": 0.7182,
+      "step": 16386
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2624580245817394e-06,
+      "loss": 1.1031,
+      "step": 16387
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.261471281166324e-06,
+      "loss": 0.8801,
+      "step": 16388
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2604847255422436e-06,
+      "loss": 0.9332,
+      "step": 16389
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2594983577334374e-06,
+      "loss": 0.9476,
+      "step": 16390
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2585121777638387e-06,
+      "loss": 0.9768,
+      "step": 16391
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.257526185657386e-06,
+      "loss": 1.1759,
+      "step": 16392
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2565403814380026e-06,
+      "loss": 1.0737,
+      "step": 16393
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.255554765129612e-06,
+      "loss": 1.2766,
+      "step": 16394
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2545693367561282e-06,
+      "loss": 1.1754,
+      "step": 16395
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2535840963414714e-06,
+      "loss": 1.1547,
+      "step": 16396
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.252599043909547e-06,
+      "loss": 1.0161,
+      "step": 16397
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2516141794842594e-06,
+      "loss": 1.1744,
+      "step": 16398
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.250629503089505e-06,
+      "loss": 0.9428,
+      "step": 16399
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2496450147491867e-06,
+      "loss": 1.1827,
+      "step": 16400
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.24866071448719e-06,
+      "loss": 1.1655,
+      "step": 16401
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.247676602327402e-06,
+      "loss": 0.9777,
+      "step": 16402
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2466926782937014e-06,
+      "loss": 1.0569,
+      "step": 16403
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.245708942409971e-06,
+      "loss": 1.1559,
+      "step": 16404
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.244725394700079e-06,
+      "loss": 1.2183,
+      "step": 16405
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2437420351878947e-06,
+      "loss": 1.2576,
+      "step": 16406
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2427588638972784e-06,
+      "loss": 1.111,
+      "step": 16407
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.241775880852094e-06,
+      "loss": 1.1275,
+      "step": 16408
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2407930860761926e-06,
+      "loss": 0.9994,
+      "step": 16409
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.239810479593424e-06,
+      "loss": 1.1071,
+      "step": 16410
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2388280614276315e-06,
+      "loss": 1.0901,
+      "step": 16411
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2378458316026588e-06,
+      "loss": 1.0491,
+      "step": 16412
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2368637901423395e-06,
+      "loss": 1.1177,
+      "step": 16413
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2358819370705067e-06,
+      "loss": 1.1269,
+      "step": 16414
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2349002724109826e-06,
+      "loss": 1.1322,
+      "step": 16415
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.233918796187592e-06,
+      "loss": 0.985,
+      "step": 16416
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2329375084241568e-06,
+      "loss": 1.1811,
+      "step": 16417
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2319564091444857e-06,
+      "loss": 1.2112,
+      "step": 16418
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2309754983723876e-06,
+      "loss": 1.188,
+      "step": 16419
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.229994776131663e-06,
+      "loss": 1.0947,
+      "step": 16420
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2290142424461182e-06,
+      "loss": 1.1798,
+      "step": 16421
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.228033897339543e-06,
+      "loss": 1.1418,
+      "step": 16422
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.227053740835726e-06,
+      "loss": 1.0722,
+      "step": 16423
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.226073772958459e-06,
+      "loss": 1.1141,
+      "step": 16424
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2250939937315186e-06,
+      "loss": 1.0629,
+      "step": 16425
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2241144031786823e-06,
+      "loss": 1.1267,
+      "step": 16426
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2231350013237175e-06,
+      "loss": 1.2223,
+      "step": 16427
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.222155788190399e-06,
+      "loss": 1.1266,
+      "step": 16428
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.221176763802485e-06,
+      "loss": 1.0825,
+      "step": 16429
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.220197928183735e-06,
+      "loss": 1.2031,
+      "step": 16430
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2192192813578983e-06,
+      "loss": 1.0812,
+      "step": 16431
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2182408233487296e-06,
+      "loss": 1.1811,
+      "step": 16432
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.217262554179972e-06,
+      "loss": 1.1734,
+      "step": 16433
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.216284473875363e-06,
+      "loss": 0.909,
+      "step": 16434
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2153065824586363e-06,
+      "loss": 1.1845,
+      "step": 16435
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.214328879953528e-06,
+      "loss": 1.0004,
+      "step": 16436
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2133513663837604e-06,
+      "loss": 1.1894,
+      "step": 16437
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2123740417730556e-06,
+      "loss": 1.1366,
+      "step": 16438
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2113969061451278e-06,
+      "loss": 1.0903,
+      "step": 16439
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2104199595236943e-06,
+      "loss": 1.0688,
+      "step": 16440
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2094432019324595e-06,
+      "loss": 1.0617,
+      "step": 16441
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.208466633395128e-06,
+      "loss": 1.1351,
+      "step": 16442
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.207490253935394e-06,
+      "loss": 1.0937,
+      "step": 16443
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2065140635769568e-06,
+      "loss": 1.063,
+      "step": 16444
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.205538062343504e-06,
+      "loss": 1.0269,
+      "step": 16445
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2045622502587195e-06,
+      "loss": 1.0303,
+      "step": 16446
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2035866273462813e-06,
+      "loss": 1.1661,
+      "step": 16447
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2026111936298687e-06,
+      "loss": 1.0296,
+      "step": 16448
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.201635949133153e-06,
+      "loss": 1.2752,
+      "step": 16449
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.200660893879797e-06,
+      "loss": 1.1399,
+      "step": 16450
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1996860278934628e-06,
+      "loss": 0.9599,
+      "step": 16451
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.198711351197811e-06,
+      "loss": 1.0886,
+      "step": 16452
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1977368638164908e-06,
+      "loss": 1.1,
+      "step": 16453
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.196762565773153e-06,
+      "loss": 0.8366,
+      "step": 16454
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1957884570914357e-06,
+      "loss": 1.1291,
+      "step": 16455
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1948145377949835e-06,
+      "loss": 1.1002,
+      "step": 16456
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1938408079074293e-06,
+      "loss": 1.129,
+      "step": 16457
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.192867267452401e-06,
+      "loss": 1.0445,
+      "step": 16458
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1918939164535214e-06,
+      "loss": 1.0942,
+      "step": 16459
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.190920754934416e-06,
+      "loss": 1.2378,
+      "step": 16460
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1899477829186986e-06,
+      "loss": 1.1718,
+      "step": 16461
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1889750004299803e-06,
+      "loss": 1.0379,
+      "step": 16462
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1880024074918636e-06,
+      "loss": 1.1049,
+      "step": 16463
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.187030004127958e-06,
+      "loss": 1.0435,
+      "step": 16464
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1860577903618564e-06,
+      "loss": 1.2076,
+      "step": 16465
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1850857662171533e-06,
+      "loss": 1.078,
+      "step": 16466
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.184113931717432e-06,
+      "loss": 1.0349,
+      "step": 16467
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.183142286886283e-06,
+      "loss": 0.9873,
+      "step": 16468
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1821708317472823e-06,
+      "loss": 1.1685,
+      "step": 16469
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1811995663240006e-06,
+      "loss": 1.0265,
+      "step": 16470
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.180228490640014e-06,
+      "loss": 1.0695,
+      "step": 16471
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.179257604718886e-06,
+      "loss": 1.271,
+      "step": 16472
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1782869085841752e-06,
+      "loss": 1.1893,
+      "step": 16473
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.177316402259435e-06,
+      "loss": 1.1467,
+      "step": 16474
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1763460857682237e-06,
+      "loss": 0.9729,
+      "step": 16475
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.175375959134084e-06,
+      "loss": 1.0134,
+      "step": 16476
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1744060223805587e-06,
+      "loss": 0.9817,
+      "step": 16477
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.173436275531182e-06,
+      "loss": 1.1378,
+      "step": 16478
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.172466718609493e-06,
+      "loss": 1.0015,
+      "step": 16479
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1714973516390157e-06,
+      "loss": 1.1823,
+      "step": 16480
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.170528174643275e-06,
+      "loss": 1.1441,
+      "step": 16481
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.169559187645788e-06,
+      "loss": 1.1789,
+      "step": 16482
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1685903906700733e-06,
+      "loss": 0.975,
+      "step": 16483
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1676217837396386e-06,
+      "loss": 1.1547,
+      "step": 16484
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1666533668779886e-06,
+      "loss": 1.1797,
+      "step": 16485
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.165685140108622e-06,
+      "loss": 1.0136,
+      "step": 16486
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1647171034550398e-06,
+      "loss": 1.0827,
+      "step": 16487
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1637492569407306e-06,
+      "loss": 0.899,
+      "step": 16488
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.162781600589182e-06,
+      "loss": 1.0741,
+      "step": 16489
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.161814134423872e-06,
+      "loss": 1.0884,
+      "step": 16490
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.160846858468285e-06,
+      "loss": 0.992,
+      "step": 16491
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1598797727458898e-06,
+      "loss": 1.0529,
+      "step": 16492
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.158912877280156e-06,
+      "loss": 1.0823,
+      "step": 16493
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.157946172094545e-06,
+      "loss": 1.1442,
+      "step": 16494
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.156979657212519e-06,
+      "loss": 1.2545,
+      "step": 16495
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.156013332657533e-06,
+      "loss": 1.2504,
+      "step": 16496
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.155047198453033e-06,
+      "loss": 1.0218,
+      "step": 16497
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1540812546224644e-06,
+      "loss": 0.8595,
+      "step": 16498
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1531155011892733e-06,
+      "loss": 1.0286,
+      "step": 16499
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1521499381768886e-06,
+      "loss": 1.2777,
+      "step": 16500
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1511845656087482e-06,
+      "loss": 1.1318,
+      "step": 16501
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.150219383508275e-06,
+      "loss": 1.116,
+      "step": 16502
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1492543918988906e-06,
+      "loss": 1.1036,
+      "step": 16503
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.148289590804016e-06,
+      "loss": 1.2228,
+      "step": 16504
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.147324980247062e-06,
+      "loss": 1.1301,
+      "step": 16505
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.146360560251437e-06,
+      "loss": 1.0775,
+      "step": 16506
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.145396330840541e-06,
+      "loss": 1.0301,
+      "step": 16507
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.14443229203778e-06,
+      "loss": 0.8573,
+      "step": 16508
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1434684438665455e-06,
+      "loss": 0.974,
+      "step": 16509
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1425047863502223e-06,
+      "loss": 1.0684,
+      "step": 16510
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1415413195122036e-06,
+      "loss": 1.2795,
+      "step": 16511
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.140578043375866e-06,
+      "loss": 0.9728,
+      "step": 16512
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.139614957964585e-06,
+      "loss": 0.9876,
+      "step": 16513
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1386520633017303e-06,
+      "loss": 1.1148,
+      "step": 16514
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.137689359410674e-06,
+      "loss": 1.0548,
+      "step": 16515
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.136726846314774e-06,
+      "loss": 0.971,
+      "step": 16516
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.135764524037388e-06,
+      "loss": 1.1592,
+      "step": 16517
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.134802392601867e-06,
+      "loss": 1.3023,
+      "step": 16518
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.133840452031564e-06,
+      "loss": 1.156,
+      "step": 16519
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.13287870234982e-06,
+      "loss": 1.0201,
+      "step": 16520
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1319171435799723e-06,
+      "loss": 1.1718,
+      "step": 16521
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.130955775745355e-06,
+      "loss": 1.1249,
+      "step": 16522
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1299945988693006e-06,
+      "loss": 1.1501,
+      "step": 16523
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.129033612975132e-06,
+      "loss": 1.1208,
+      "step": 16524
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.128072818086171e-06,
+      "loss": 1.0215,
+      "step": 16525
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1271122142257296e-06,
+      "loss": 1.0948,
+      "step": 16526
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1261518014171235e-06,
+      "loss": 1.0542,
+      "step": 16527
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1251915796836577e-06,
+      "loss": 0.9676,
+      "step": 16528
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.124231549048632e-06,
+      "loss": 0.9831,
+      "step": 16529
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1232717095353437e-06,
+      "loss": 1.0629,
+      "step": 16530
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.122312061167088e-06,
+      "loss": 1.2603,
+      "step": 16531
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.121352603967152e-06,
+      "loss": 1.0708,
+      "step": 16532
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.120393337958817e-06,
+      "loss": 1.1339,
+      "step": 16533
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.119434263165361e-06,
+      "loss": 1.1581,
+      "step": 16534
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.118475379610061e-06,
+      "loss": 1.2381,
+      "step": 16535
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1175166873161857e-06,
+      "loss": 1.2121,
+      "step": 16536
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1165581863069985e-06,
+      "loss": 1.0904,
+      "step": 16537
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1155998766057574e-06,
+      "loss": 1.0631,
+      "step": 16538
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1146417582357226e-06,
+      "loss": 0.9845,
+      "step": 16539
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1136838312201423e-06,
+      "loss": 1.0109,
+      "step": 16540
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1127260955822627e-06,
+      "loss": 1.0958,
+      "step": 16541
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.111768551345321e-06,
+      "loss": 1.1907,
+      "step": 16542
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.110811198532563e-06,
+      "loss": 1.0424,
+      "step": 16543
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.109854037167215e-06,
+      "loss": 0.9338,
+      "step": 16544
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.108897067272505e-06,
+      "loss": 1.1191,
+      "step": 16545
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1079402888716547e-06,
+      "loss": 1.1748,
+      "step": 16546
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.106983701987886e-06,
+      "loss": 1.1278,
+      "step": 16547
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1060273066444104e-06,
+      "loss": 1.0794,
+      "step": 16548
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1050711028644356e-06,
+      "loss": 1.0962,
+      "step": 16549
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1041150906711662e-06,
+      "loss": 1.1239,
+      "step": 16550
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1031592700878033e-06,
+      "loss": 1.1886,
+      "step": 16551
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1022036411375424e-06,
+      "loss": 1.2433,
+      "step": 16552
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.10124820384357e-06,
+      "loss": 0.9845,
+      "step": 16553
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1002929582290764e-06,
+      "loss": 1.1409,
+      "step": 16554
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.099337904317241e-06,
+      "loss": 1.0749,
+      "step": 16555
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.098383042131239e-06,
+      "loss": 1.0553,
+      "step": 16556
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.09742837169424e-06,
+      "loss": 0.9596,
+      "step": 16557
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0964738930294173e-06,
+      "loss": 1.0495,
+      "step": 16558
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0955196061599294e-06,
+      "loss": 1.0767,
+      "step": 16559
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0945655111089345e-06,
+      "loss": 1.0383,
+      "step": 16560
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.093611607899583e-06,
+      "loss": 0.8444,
+      "step": 16561
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.092657896555028e-06,
+      "loss": 1.14,
+      "step": 16562
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0917043770984124e-06,
+      "loss": 1.1376,
+      "step": 16563
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0907510495528725e-06,
+      "loss": 1.0389,
+      "step": 16564
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0897979139415426e-06,
+      "loss": 1.1026,
+      "step": 16565
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0888449702875568e-06,
+      "loss": 1.0799,
+      "step": 16566
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.087892218614038e-06,
+      "loss": 1.1705,
+      "step": 16567
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.086939658944106e-06,
+      "loss": 1.1252,
+      "step": 16568
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0859872913008748e-06,
+      "loss": 1.1847,
+      "step": 16569
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.08503511570746e-06,
+      "loss": 1.0985,
+      "step": 16570
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.084083132186966e-06,
+      "loss": 1.1362,
+      "step": 16571
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.083131340762494e-06,
+      "loss": 1.1117,
+      "step": 16572
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0821797414571397e-06,
+      "loss": 0.995,
+      "step": 16573
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.081228334294e-06,
+      "loss": 1.0338,
+      "step": 16574
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.080277119296161e-06,
+      "loss": 0.9345,
+      "step": 16575
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0793260964867047e-06,
+      "loss": 1.0333,
+      "step": 16576
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.078375265888707e-06,
+      "loss": 1.0539,
+      "step": 16577
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0774246275252484e-06,
+      "loss": 1.2578,
+      "step": 16578
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0764741814193946e-06,
+      "loss": 1.1265,
+      "step": 16579
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0755239275942097e-06,
+      "loss": 1.0447,
+      "step": 16580
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.074573866072751e-06,
+      "loss": 1.1246,
+      "step": 16581
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0736239968780803e-06,
+      "loss": 1.0511,
+      "step": 16582
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.072674320033241e-06,
+      "loss": 1.003,
+      "step": 16583
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0717248355612864e-06,
+      "loss": 1.0235,
+      "step": 16584
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.070775543485253e-06,
+      "loss": 1.1269,
+      "step": 16585
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.069826443828176e-06,
+      "loss": 0.941,
+      "step": 16586
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.068877536613091e-06,
+      "loss": 1.1305,
+      "step": 16587
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0679288218630245e-06,
+      "loss": 1.1683,
+      "step": 16588
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.066980299600998e-06,
+      "loss": 1.1551,
+      "step": 16589
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0660319698500264e-06,
+      "loss": 1.2491,
+      "step": 16590
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.065083832633129e-06,
+      "loss": 1.2704,
+      "step": 16591
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.064135887973311e-06,
+      "loss": 1.073,
+      "step": 16592
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0631881358935747e-06,
+      "loss": 1.0663,
+      "step": 16593
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.062240576416923e-06,
+      "loss": 1.2345,
+      "step": 16594
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0612932095663473e-06,
+      "loss": 1.1532,
+      "step": 16595
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0603460353648398e-06,
+      "loss": 1.2156,
+      "step": 16596
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0593990538353816e-06,
+      "loss": 0.9649,
+      "step": 16597
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0584522650009587e-06,
+      "loss": 1.2857,
+      "step": 16598
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.057505668884543e-06,
+      "loss": 1.0849,
+      "step": 16599
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.056559265509108e-06,
+      "loss": 0.8608,
+      "step": 16600
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0556130548976148e-06,
+      "loss": 1.1697,
+      "step": 16601
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0546670370730326e-06,
+      "loss": 1.1288,
+      "step": 16602
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.053721212058315e-06,
+      "loss": 0.8895,
+      "step": 16603
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.052775579876415e-06,
+      "loss": 1.0647,
+      "step": 16604
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0518301405502763e-06,
+      "loss": 1.1355,
+      "step": 16605
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.050884894102848e-06,
+      "loss": 1.1234,
+      "step": 16606
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0499398405570658e-06,
+      "loss": 1.0569,
+      "step": 16607
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0489949799358646e-06,
+      "loss": 1.2694,
+      "step": 16608
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0480503122621674e-06,
+      "loss": 1.1111,
+      "step": 16609
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.047105837558907e-06,
+      "loss": 1.1023,
+      "step": 16610
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.046161555849e-06,
+      "loss": 1.1208,
+      "step": 16611
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0452174671553593e-06,
+      "loss": 1.1289,
+      "step": 16612
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0442735715008955e-06,
+      "loss": 1.1669,
+      "step": 16613
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.043329868908517e-06,
+      "loss": 1.2172,
+      "step": 16614
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0423863594011217e-06,
+      "loss": 1.0623,
+      "step": 16615
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0414430430016085e-06,
+      "loss": 1.1554,
+      "step": 16616
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0404999197328644e-06,
+      "loss": 1.1835,
+      "step": 16617
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0395569896177814e-06,
+      "loss": 1.2259,
+      "step": 16618
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.03861425267924e-06,
+      "loss": 1.1265,
+      "step": 16619
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0376717089401166e-06,
+      "loss": 0.9715,
+      "step": 16620
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0367293584232817e-06,
+      "loss": 1.0507,
+      "step": 16621
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.035787201151609e-06,
+      "loss": 0.9846,
+      "step": 16622
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.034845237147959e-06,
+      "loss": 1.0761,
+      "step": 16623
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.033903466435191e-06,
+      "loss": 1.0941,
+      "step": 16624
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0329618890361547e-06,
+      "loss": 1.0863,
+      "step": 16625
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0320205049737063e-06,
+      "loss": 1.0994,
+      "step": 16626
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0310793142706876e-06,
+      "loss": 1.1465,
+      "step": 16627
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0301383169499376e-06,
+      "loss": 1.1522,
+      "step": 16628
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.02919751303429e-06,
+      "loss": 1.069,
+      "step": 16629
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0282569025465805e-06,
+      "loss": 1.1501,
+      "step": 16630
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0273164855096305e-06,
+      "loss": 1.2063,
+      "step": 16631
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0263762619462637e-06,
+      "loss": 1.1535,
+      "step": 16632
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.025436231879293e-06,
+      "loss": 1.0837,
+      "step": 16633
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.024496395331534e-06,
+      "loss": 1.1741,
+      "step": 16634
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0235567523257937e-06,
+      "loss": 1.0571,
+      "step": 16635
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0226173028848727e-06,
+      "loss": 1.1443,
+      "step": 16636
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0216780470315657e-06,
+      "loss": 1.112,
+      "step": 16637
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0207389847886727e-06,
+      "loss": 1.0541,
+      "step": 16638
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.019800116178978e-06,
+      "loss": 0.9805,
+      "step": 16639
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0188614412252625e-06,
+      "loss": 1.126,
+      "step": 16640
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0179229599503114e-06,
+      "loss": 1.0723,
+      "step": 16641
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.016984672376896e-06,
+      "loss": 1.0115,
+      "step": 16642
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.016046578527785e-06,
+      "loss": 1.205,
+      "step": 16643
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0151086784257403e-06,
+      "loss": 1.0768,
+      "step": 16644
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.014170972093529e-06,
+      "loss": 1.2121,
+      "step": 16645
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0132334595539015e-06,
+      "loss": 1.095,
+      "step": 16646
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0122961408296105e-06,
+      "loss": 1.1994,
+      "step": 16647
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.011359015943398e-06,
+      "loss": 0.9334,
+      "step": 16648
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0104220849180113e-06,
+      "loss": 1.1508,
+      "step": 16649
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0094853477761845e-06,
+      "loss": 1.0579,
+      "step": 16650
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0085488045406488e-06,
+      "loss": 1.1352,
+      "step": 16651
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0076124552341293e-06,
+      "loss": 1.1287,
+      "step": 16652
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.006676299879353e-06,
+      "loss": 1.1068,
+      "step": 16653
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0057403384990347e-06,
+      "loss": 1.1721,
+      "step": 16654
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0048045711158894e-06,
+      "loss": 1.0624,
+      "step": 16655
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0038689977526206e-06,
+      "loss": 1.0793,
+      "step": 16656
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.002933618431938e-06,
+      "loss": 1.036,
+      "step": 16657
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0019984331765384e-06,
+      "loss": 1.0419,
+      "step": 16658
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0010634420091157e-06,
+      "loss": 1.1504,
+      "step": 16659
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.000128644952357e-06,
+      "loss": 1.0304,
+      "step": 16660
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.999194042028951e-06,
+      "loss": 1.1152,
+      "step": 16661
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9982596332615765e-06,
+      "loss": 1.0854,
+      "step": 16662
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9973254186729084e-06,
+      "loss": 1.2124,
+      "step": 16663
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9963913982856144e-06,
+      "loss": 1.0904,
+      "step": 16664
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9954575721223667e-06,
+      "loss": 1.0419,
+      "step": 16665
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9945239402058203e-06,
+      "loss": 1.0948,
+      "step": 16666
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9935905025586368e-06,
+      "loss": 1.1208,
+      "step": 16667
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9926572592034667e-06,
+      "loss": 1.1931,
+      "step": 16668
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.991724210162952e-06,
+      "loss": 1.052,
+      "step": 16669
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9907913554597434e-06,
+      "loss": 0.8552,
+      "step": 16670
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9898586951164734e-06,
+      "loss": 0.925,
+      "step": 16671
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9889262291557764e-06,
+      "loss": 1.0932,
+      "step": 16672
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9879939576002773e-06,
+      "loss": 0.9578,
+      "step": 16673
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9870618804726048e-06,
+      "loss": 1.0866,
+      "step": 16674
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9861299977953764e-06,
+      "loss": 1.1932,
+      "step": 16675
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.985198309591204e-06,
+      "loss": 1.1475,
+      "step": 16676
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.984266815882695e-06,
+      "loss": 1.121,
+      "step": 16677
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9833355166924607e-06,
+      "loss": 1.1199,
+      "step": 16678
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9824044120430964e-06,
+      "loss": 1.1326,
+      "step": 16679
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.981473501957195e-06,
+      "loss": 1.2436,
+      "step": 16680
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9805427864573532e-06,
+      "loss": 1.2873,
+      "step": 16681
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9796122655661543e-06,
+      "loss": 1.1793,
+      "step": 16682
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9786819393061783e-06,
+      "loss": 1.2452,
+      "step": 16683
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.977751807699999e-06,
+      "loss": 1.1628,
+      "step": 16684
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9768218707701924e-06,
+      "loss": 0.9399,
+      "step": 16685
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.975892128539325e-06,
+      "loss": 1.1715,
+      "step": 16686
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9749625810299577e-06,
+      "loss": 1.1974,
+      "step": 16687
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9740332282646448e-06,
+      "loss": 1.1324,
+      "step": 16688
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.973104070265944e-06,
+      "loss": 1.102,
+      "step": 16689
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.972175107056401e-06,
+      "loss": 1.1338,
+      "step": 16690
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.97124633865856e-06,
+      "loss": 1.1484,
+      "step": 16691
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9703177650949558e-06,
+      "loss": 1.2234,
+      "step": 16692
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9693893863881274e-06,
+      "loss": 1.0091,
+      "step": 16693
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.968461202560602e-06,
+      "loss": 1.0859,
+      "step": 16694
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9675332136349023e-06,
+      "loss": 1.0692,
+      "step": 16695
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.966605419633547e-06,
+      "loss": 0.9855,
+      "step": 16696
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.965677820579055e-06,
+      "loss": 1.1398,
+      "step": 16697
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9647504164939345e-06,
+      "loss": 1.1191,
+      "step": 16698
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9638232074006904e-06,
+      "loss": 1.1409,
+      "step": 16699
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9628961933218216e-06,
+      "loss": 1.0243,
+      "step": 16700
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.961969374279827e-06,
+      "loss": 1.1037,
+      "step": 16701
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9610427502971987e-06,
+      "loss": 0.9619,
+      "step": 16702
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9601163213964194e-06,
+      "loss": 1.0148,
+      "step": 16703
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9591900875999703e-06,
+      "loss": 1.0765,
+      "step": 16704
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.958264048930334e-06,
+      "loss": 1.1804,
+      "step": 16705
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9573382054099786e-06,
+      "loss": 1.0911,
+      "step": 16706
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9564125570613723e-06,
+      "loss": 1.0142,
+      "step": 16707
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9554871039069754e-06,
+      "loss": 1.2025,
+      "step": 16708
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9545618459692516e-06,
+      "loss": 1.2454,
+      "step": 16709
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9536367832706503e-06,
+      "loss": 1.0548,
+      "step": 16710
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9527119158336215e-06,
+      "loss": 1.1003,
+      "step": 16711
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.951787243680605e-06,
+      "loss": 1.1835,
+      "step": 16712
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9508627668340463e-06,
+      "loss": 1.0519,
+      "step": 16713
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.949938485316376e-06,
+      "loss": 1.0054,
+      "step": 16714
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9490143991500245e-06,
+      "loss": 1.3162,
+      "step": 16715
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9480905083574144e-06,
+      "loss": 1.0788,
+      "step": 16716
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.94716681296097e-06,
+      "loss": 1.0007,
+      "step": 16717
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.946243312983104e-06,
+      "loss": 1.131,
+      "step": 16718
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9453200084462275e-06,
+      "loss": 0.9932,
+      "step": 16719
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9443968993727447e-06,
+      "loss": 1.1124,
+      "step": 16720
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9434739857850603e-06,
+      "loss": 1.0593,
+      "step": 16721
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.94255126770557e-06,
+      "loss": 1.1733,
+      "step": 16722
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.941628745156664e-06,
+      "loss": 0.9279,
+      "step": 16723
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.940706418160726e-06,
+      "loss": 1.1431,
+      "step": 16724
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.939784286740145e-06,
+      "loss": 1.1233,
+      "step": 16725
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9388623509172945e-06,
+      "loss": 1.0922,
+      "step": 16726
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.937940610714546e-06,
+      "loss": 1.0039,
+      "step": 16727
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.937019066154272e-06,
+      "loss": 1.0507,
+      "step": 16728
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9360977172588326e-06,
+      "loss": 1.1341,
+      "step": 16729
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.935176564050586e-06,
+      "loss": 1.1291,
+      "step": 16730
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9342556065518834e-06,
+      "loss": 1.0412,
+      "step": 16731
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9333348447850796e-06,
+      "loss": 1.0053,
+      "step": 16732
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9324142787725176e-06,
+      "loss": 0.9064,
+      "step": 16733
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9314939085365335e-06,
+      "loss": 1.1248,
+      "step": 16734
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.930573734099461e-06,
+      "loss": 1.1338,
+      "step": 16735
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.929653755483636e-06,
+      "loss": 1.0648,
+      "step": 16736
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9287339727113796e-06,
+      "loss": 0.97,
+      "step": 16737
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9278143858050137e-06,
+      "loss": 1.0247,
+      "step": 16738
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.92689499478685e-06,
+      "loss": 1.0684,
+      "step": 16739
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.925975799679205e-06,
+      "loss": 1.0,
+      "step": 16740
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9250568005043835e-06,
+      "loss": 1.0941,
+      "step": 16741
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9241379972846852e-06,
+      "loss": 1.1128,
+      "step": 16742
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9232193900424045e-06,
+      "loss": 1.1649,
+      "step": 16743
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9223009787998394e-06,
+      "loss": 1.0797,
+      "step": 16744
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9213827635792737e-06,
+      "loss": 1.0404,
+      "step": 16745
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.92046474440299e-06,
+      "loss": 1.1434,
+      "step": 16746
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.919546921293264e-06,
+      "loss": 1.1302,
+      "step": 16747
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.918629294272372e-06,
+      "loss": 1.0283,
+      "step": 16748
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.917711863362581e-06,
+      "loss": 1.1829,
+      "step": 16749
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9167946285861515e-06,
+      "loss": 1.0309,
+      "step": 16750
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.915877589965347e-06,
+      "loss": 1.1241,
+      "step": 16751
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9149607475224164e-06,
+      "loss": 1.0778,
+      "step": 16752
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9140441012796128e-06,
+      "loss": 1.0616,
+      "step": 16753
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.91312765125918e-06,
+      "loss": 1.1756,
+      "step": 16754
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9122113974833558e-06,
+      "loss": 1.0828,
+      "step": 16755
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9112953399743737e-06,
+      "loss": 1.048,
+      "step": 16756
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.910379478754468e-06,
+      "loss": 1.1669,
+      "step": 16757
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.909463813845862e-06,
+      "loss": 1.0475,
+      "step": 16758
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.908548345270775e-06,
+      "loss": 1.1447,
+      "step": 16759
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9076330730514214e-06,
+      "loss": 1.0084,
+      "step": 16760
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.906717997210017e-06,
+      "loss": 1.0404,
+      "step": 16761
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9058031177687652e-06,
+      "loss": 1.115,
+      "step": 16762
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9048884347498674e-06,
+      "loss": 1.1549,
+      "step": 16763
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9039739481755182e-06,
+      "loss": 1.1513,
+      "step": 16764
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9030596580679139e-06,
+      "loss": 1.1011,
+      "step": 16765
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9021455644492403e-06,
+      "loss": 0.9994,
+      "step": 16766
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9012316673416754e-06,
+      "loss": 0.9478,
+      "step": 16767
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9003179667674032e-06,
+      "loss": 1.106,
+      "step": 16768
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.899404462748593e-06,
+      "loss": 1.1162,
+      "step": 16769
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8984911553074148e-06,
+      "loss": 1.0511,
+      "step": 16770
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8975780444660274e-06,
+      "loss": 1.1134,
+      "step": 16771
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8966651302465943e-06,
+      "loss": 1.1483,
+      "step": 16772
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.895752412671269e-06,
+      "loss": 1.1529,
+      "step": 16773
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.894839891762198e-06,
+      "loss": 1.083,
+      "step": 16774
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8939275675415236e-06,
+      "loss": 1.1437,
+      "step": 16775
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8930154400313904e-06,
+      "loss": 1.1192,
+      "step": 16776
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.892103509253932e-06,
+      "loss": 1.1642,
+      "step": 16777
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.891191775231276e-06,
+      "loss": 1.1601,
+      "step": 16778
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8902802379855456e-06,
+      "loss": 1.0535,
+      "step": 16779
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.889368897538868e-06,
+      "loss": 1.0475,
+      "step": 16780
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.888457753913353e-06,
+      "loss": 1.0785,
+      "step": 16781
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8875468071311143e-06,
+      "loss": 1.1496,
+      "step": 16782
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.886636057214254e-06,
+      "loss": 1.053,
+      "step": 16783
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8857255041848776e-06,
+      "loss": 1.1092,
+      "step": 16784
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8848151480650812e-06,
+      "loss": 0.9439,
+      "step": 16785
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8839049888769546e-06,
+      "loss": 1.1496,
+      "step": 16786
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8829950266425833e-06,
+      "loss": 1.147,
+      "step": 16787
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8820852613840534e-06,
+      "loss": 1.1378,
+      "step": 16788
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8811756931234403e-06,
+      "loss": 1.0945,
+      "step": 16789
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8802663218828177e-06,
+      "loss": 1.1416,
+      "step": 16790
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8793571476842487e-06,
+      "loss": 1.0728,
+      "step": 16791
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8784481705498014e-06,
+      "loss": 1.1105,
+      "step": 16792
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8775393905015337e-06,
+      "loss": 1.109,
+      "step": 16793
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8766308075614969e-06,
+      "loss": 1.1639,
+      "step": 16794
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.875722421751739e-06,
+      "loss": 0.9103,
+      "step": 16795
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8748142330943064e-06,
+      "loss": 1.094,
+      "step": 16796
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8739062416112385e-06,
+      "loss": 1.0826,
+      "step": 16797
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8729984473245677e-06,
+      "loss": 1.1069,
+      "step": 16798
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8720908502563207e-06,
+      "loss": 1.1066,
+      "step": 16799
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8711834504285287e-06,
+      "loss": 1.1431,
+      "step": 16800
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8702762478632075e-06,
+      "loss": 1.0351,
+      "step": 16801
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.869369242582374e-06,
+      "loss": 1.0914,
+      "step": 16802
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8684624346080349e-06,
+      "loss": 0.9818,
+      "step": 16803
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8675558239622005e-06,
+      "loss": 1.0117,
+      "step": 16804
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8666494106668709e-06,
+      "loss": 0.974,
+      "step": 16805
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8657431947440396e-06,
+      "loss": 1.0579,
+      "step": 16806
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.864837176215697e-06,
+      "loss": 1.1597,
+      "step": 16807
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8639313551038329e-06,
+      "loss": 1.2074,
+      "step": 16808
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.863025731430429e-06,
+      "loss": 1.0371,
+      "step": 16809
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8621203052174596e-06,
+      "loss": 1.1275,
+      "step": 16810
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8612150764868952e-06,
+      "loss": 1.2385,
+      "step": 16811
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8603100452607092e-06,
+      "loss": 1.1008,
+      "step": 16812
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8594052115608596e-06,
+      "loss": 1.0933,
+      "step": 16813
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8585005754093021e-06,
+      "loss": 1.0417,
+      "step": 16814
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.857596136827996e-06,
+      "loss": 1.1741,
+      "step": 16815
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8566918958388846e-06,
+      "loss": 1.1612,
+      "step": 16816
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8557878524639128e-06,
+      "loss": 1.122,
+      "step": 16817
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8548840067250162e-06,
+      "loss": 0.9342,
+      "step": 16818
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.853980358644133e-06,
+      "loss": 1.1445,
+      "step": 16819
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.85307690824319e-06,
+      "loss": 1.2896,
+      "step": 16820
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8521736555441105e-06,
+      "loss": 1.1248,
+      "step": 16821
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.851270600568813e-06,
+      "loss": 1.0727,
+      "step": 16822
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8503677433392154e-06,
+      "loss": 1.0931,
+      "step": 16823
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8494650838772266e-06,
+      "loss": 1.0903,
+      "step": 16824
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8485626222047492e-06,
+      "loss": 0.9599,
+      "step": 16825
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8476603583436814e-06,
+      "loss": 1.081,
+      "step": 16826
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8467582923159254e-06,
+      "loss": 1.1974,
+      "step": 16827
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8458564241433674e-06,
+      "loss": 1.1714,
+      "step": 16828
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8449547538478929e-06,
+      "loss": 1.106,
+      "step": 16829
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8440532814513812e-06,
+      "loss": 1.1239,
+      "step": 16830
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8431520069757135e-06,
+      "loss": 1.1134,
+      "step": 16831
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8422509304427583e-06,
+      "loss": 1.1559,
+      "step": 16832
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8413500518743788e-06,
+      "loss": 0.9897,
+      "step": 16833
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.840449371292442e-06,
+      "loss": 1.1227,
+      "step": 16834
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8395488887188007e-06,
+      "loss": 1.0155,
+      "step": 16835
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8386486041753114e-06,
+      "loss": 1.1738,
+      "step": 16836
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8377485176838194e-06,
+      "loss": 1.1651,
+      "step": 16837
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8368486292661657e-06,
+      "loss": 1.1316,
+      "step": 16838
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8359489389441865e-06,
+      "loss": 1.1371,
+      "step": 16839
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8350494467397196e-06,
+      "loss": 1.1047,
+      "step": 16840
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8341501526745897e-06,
+      "loss": 1.0919,
+      "step": 16841
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8332510567706208e-06,
+      "loss": 1.0618,
+      "step": 16842
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8323521590496296e-06,
+      "loss": 0.9661,
+      "step": 16843
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8314534595334332e-06,
+      "loss": 1.1908,
+      "step": 16844
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8305549582438387e-06,
+      "loss": 1.1149,
+      "step": 16845
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8296566552026495e-06,
+      "loss": 1.1925,
+      "step": 16846
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.828758550431663e-06,
+      "loss": 1.0275,
+      "step": 16847
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.827860643952678e-06,
+      "loss": 1.0158,
+      "step": 16848
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8269629357874818e-06,
+      "loss": 1.0515,
+      "step": 16849
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8260654259578592e-06,
+      "loss": 1.0863,
+      "step": 16850
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8251681144855882e-06,
+      "loss": 1.1147,
+      "step": 16851
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8242710013924469e-06,
+      "loss": 1.1286,
+      "step": 16852
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.823374086700206e-06,
+      "loss": 1.102,
+      "step": 16853
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8224773704306254e-06,
+      "loss": 0.9896,
+      "step": 16854
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8215808526054734e-06,
+      "loss": 1.056,
+      "step": 16855
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8206845332465017e-06,
+      "loss": 1.1923,
+      "step": 16856
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8197884123754616e-06,
+      "loss": 1.1868,
+      "step": 16857
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8188924900140982e-06,
+      "loss": 1.1485,
+      "step": 16858
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8179967661841557e-06,
+      "loss": 1.0725,
+      "step": 16859
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8171012409073696e-06,
+      "loss": 1.1496,
+      "step": 16860
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.816205914205471e-06,
+      "loss": 1.0297,
+      "step": 16861
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8153107861001851e-06,
+      "loss": 1.0738,
+      "step": 16862
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8144158566132386e-06,
+      "loss": 1.1154,
+      "step": 16863
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8135211257663464e-06,
+      "loss": 0.9828,
+      "step": 16864
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8126265935812215e-06,
+      "loss": 1.256,
+      "step": 16865
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8117322600795672e-06,
+      "loss": 1.1885,
+      "step": 16866
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.810838125283093e-06,
+      "loss": 1.1829,
+      "step": 16867
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.809944189213495e-06,
+      "loss": 1.0823,
+      "step": 16868
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8090504518924645e-06,
+      "loss": 1.1132,
+      "step": 16869
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8081569133416888e-06,
+      "loss": 1.1946,
+      "step": 16870
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8072635735828547e-06,
+      "loss": 1.0715,
+      "step": 16871
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8063704326376408e-06,
+      "loss": 1.1951,
+      "step": 16872
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8054774905277206e-06,
+      "loss": 1.1655,
+      "step": 16873
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8045847472747592e-06,
+      "loss": 0.9738,
+      "step": 16874
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8036922029004266e-06,
+      "loss": 1.2754,
+      "step": 16875
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8027998574263794e-06,
+      "loss": 1.0784,
+      "step": 16876
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8019077108742734e-06,
+      "loss": 1.0348,
+      "step": 16877
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8010157632657544e-06,
+      "loss": 1.1948,
+      "step": 16878
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.800124014622473e-06,
+      "loss": 1.1287,
+      "step": 16879
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7992324649660664e-06,
+      "loss": 1.1143,
+      "step": 16880
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7983411143181717e-06,
+      "loss": 1.1192,
+      "step": 16881
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7974499627004138e-06,
+      "loss": 1.1441,
+      "step": 16882
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7965590101344254e-06,
+      "loss": 1.0768,
+      "step": 16883
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7956682566418249e-06,
+      "loss": 1.1931,
+      "step": 16884
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7947777022442281e-06,
+      "loss": 1.0052,
+      "step": 16885
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7938873469632422e-06,
+      "loss": 1.1225,
+      "step": 16886
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7929971908204802e-06,
+      "loss": 1.0955,
+      "step": 16887
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7921072338375411e-06,
+      "loss": 1.2248,
+      "step": 16888
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.79121747603602e-06,
+      "loss": 1.1487,
+      "step": 16889
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7903279174375065e-06,
+      "loss": 1.1018,
+      "step": 16890
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7894385580635943e-06,
+      "loss": 1.0257,
+      "step": 16891
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7885493979358616e-06,
+      "loss": 1.1383,
+      "step": 16892
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7876604370758855e-06,
+      "loss": 1.0831,
+      "step": 16893
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7867716755052378e-06,
+      "loss": 1.1325,
+      "step": 16894
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.785883113245488e-06,
+      "loss": 1.1931,
+      "step": 16895
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7849947503181996e-06,
+      "loss": 1.0261,
+      "step": 16896
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7841065867449258e-06,
+      "loss": 0.9687,
+      "step": 16897
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.783218622547226e-06,
+      "loss": 1.0419,
+      "step": 16898
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7823308577466459e-06,
+      "loss": 0.9954,
+      "step": 16899
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7814432923647274e-06,
+      "loss": 1.0542,
+      "step": 16900
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7805559264230088e-06,
+      "loss": 1.0833,
+      "step": 16901
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7796687599430273e-06,
+      "loss": 1.0664,
+      "step": 16902
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7787817929463103e-06,
+      "loss": 0.9656,
+      "step": 16903
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7778950254543815e-06,
+      "loss": 1.1723,
+      "step": 16904
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.777008457488757e-06,
+      "loss": 1.1658,
+      "step": 16905
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7761220890709574e-06,
+      "loss": 1.1189,
+      "step": 16906
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.775235920222489e-06,
+      "loss": 1.0695,
+      "step": 16907
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7743499509648565e-06,
+      "loss": 1.0591,
+      "step": 16908
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7734641813195574e-06,
+      "loss": 1.2247,
+      "step": 16909
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7725786113080924e-06,
+      "loss": 1.0958,
+      "step": 16910
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7716932409519472e-06,
+      "loss": 1.1625,
+      "step": 16911
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7708080702726094e-06,
+      "loss": 1.1601,
+      "step": 16912
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7699230992915562e-06,
+      "loss": 1.1571,
+      "step": 16913
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.769038328030267e-06,
+      "loss": 1.0953,
+      "step": 16914
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7681537565102124e-06,
+      "loss": 0.9474,
+      "step": 16915
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7672693847528544e-06,
+      "loss": 1.0312,
+      "step": 16916
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7663852127796588e-06,
+      "loss": 1.2056,
+      "step": 16917
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7655012406120776e-06,
+      "loss": 0.9718,
+      "step": 16918
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7646174682715677e-06,
+      "loss": 1.1536,
+      "step": 16919
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7637338957795714e-06,
+      "loss": 1.2174,
+      "step": 16920
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7628505231575321e-06,
+      "loss": 1.016,
+      "step": 16921
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.761967350426883e-06,
+      "loss": 0.9814,
+      "step": 16922
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7610843776090626e-06,
+      "loss": 1.1731,
+      "step": 16923
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7602016047254943e-06,
+      "loss": 1.0493,
+      "step": 16924
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7593190317975994e-06,
+      "loss": 1.1968,
+      "step": 16925
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.758436658846796e-06,
+      "loss": 1.1285,
+      "step": 16926
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7575544858944982e-06,
+      "loss": 1.0303,
+      "step": 16927
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7566725129621143e-06,
+      "loss": 1.0289,
+      "step": 16928
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.755790740071045e-06,
+      "loss": 1.1021,
+      "step": 16929
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7549091672426866e-06,
+      "loss": 1.0519,
+      "step": 16930
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7540277944984362e-06,
+      "loss": 1.0725,
+      "step": 16931
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7531466218596826e-06,
+      "loss": 1.1421,
+      "step": 16932
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7522656493478063e-06,
+      "loss": 1.0971,
+      "step": 16933
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7513848769841846e-06,
+      "loss": 1.1325,
+      "step": 16934
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.750504304790196e-06,
+      "loss": 1.1897,
+      "step": 16935
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7496239327872078e-06,
+      "loss": 1.1919,
+      "step": 16936
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7487437609965796e-06,
+      "loss": 1.042,
+      "step": 16937
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7478637894396777e-06,
+      "loss": 1.1174,
+      "step": 16938
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7469840181378527e-06,
+      "loss": 1.0804,
+      "step": 16939
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7461044471124545e-06,
+      "loss": 0.8068,
+      "step": 16940
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7452250763848244e-06,
+      "loss": 1.1045,
+      "step": 16941
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.744345905976309e-06,
+      "loss": 1.1209,
+      "step": 16942
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.743466935908238e-06,
+      "loss": 0.9856,
+      "step": 16943
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.742588166201944e-06,
+      "loss": 1.1056,
+      "step": 16944
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7417095968787478e-06,
+      "loss": 1.0528,
+      "step": 16945
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.740831227959976e-06,
+      "loss": 1.0998,
+      "step": 16946
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.73995305946694e-06,
+      "loss": 1.1479,
+      "step": 16947
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.739075091420952e-06,
+      "loss": 0.9955,
+      "step": 16948
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7381973238433148e-06,
+      "loss": 1.1542,
+      "step": 16949
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7373197567553335e-06,
+      "loss": 1.1097,
+      "step": 16950
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7364423901783024e-06,
+      "loss": 1.2342,
+      "step": 16951
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7355652241335119e-06,
+      "loss": 1.1012,
+      "step": 16952
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7346882586422465e-06,
+      "loss": 1.1374,
+      "step": 16953
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7338114937257921e-06,
+      "loss": 1.1867,
+      "step": 16954
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.732934929405422e-06,
+      "loss": 1.0239,
+      "step": 16955
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7320585657024091e-06,
+      "loss": 1.2686,
+      "step": 16956
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7311824026380176e-06,
+      "loss": 0.932,
+      "step": 16957
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7303064402335135e-06,
+      "loss": 1.1435,
+      "step": 16958
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.729430678510151e-06,
+      "loss": 1.0952,
+      "step": 16959
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7285551174891846e-06,
+      "loss": 1.0538,
+      "step": 16960
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.727679757191857e-06,
+      "loss": 1.1012,
+      "step": 16961
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7268045976394154e-06,
+      "loss": 0.9693,
+      "step": 16962
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7259296388530967e-06,
+      "loss": 1.0697,
+      "step": 16963
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7250548808541324e-06,
+      "loss": 1.1394,
+      "step": 16964
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7241803236637478e-06,
+      "loss": 0.8715,
+      "step": 16965
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7233059673031705e-06,
+      "loss": 1.1513,
+      "step": 16966
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7224318117936168e-06,
+      "loss": 1.0001,
+      "step": 16967
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7215578571562986e-06,
+      "loss": 0.9819,
+      "step": 16968
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7206841034124245e-06,
+      "loss": 1.1419,
+      "step": 16969
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7198105505831997e-06,
+      "loss": 0.9143,
+      "step": 16970
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7189371986898217e-06,
+      "loss": 1.0545,
+      "step": 16971
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.718064047753485e-06,
+      "loss": 1.103,
+      "step": 16972
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7171910977953744e-06,
+      "loss": 1.0165,
+      "step": 16973
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7163183488366797e-06,
+      "loss": 1.0722,
+      "step": 16974
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7154458008985775e-06,
+      "loss": 0.8512,
+      "step": 16975
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7145734540022418e-06,
+      "loss": 1.0487,
+      "step": 16976
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7137013081688392e-06,
+      "loss": 1.1626,
+      "step": 16977
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7128293634195393e-06,
+      "loss": 1.1153,
+      "step": 16978
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7119576197754985e-06,
+      "loss": 1.2106,
+      "step": 16979
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.711086077257872e-06,
+      "loss": 1.0917,
+      "step": 16980
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7102147358878073e-06,
+      "loss": 1.0988,
+      "step": 16981
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7093435956864545e-06,
+      "loss": 0.876,
+      "step": 16982
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7084726566749499e-06,
+      "loss": 1.0574,
+      "step": 16983
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7076019188744274e-06,
+      "loss": 1.0846,
+      "step": 16984
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7067313823060216e-06,
+      "loss": 1.1439,
+      "step": 16985
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7058610469908566e-06,
+      "loss": 1.0488,
+      "step": 16986
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7049909129500508e-06,
+      "loss": 0.97,
+      "step": 16987
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7041209802047177e-06,
+      "loss": 1.0801,
+      "step": 16988
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7032512487759745e-06,
+      "loss": 0.9995,
+      "step": 16989
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7023817186849235e-06,
+      "loss": 1.0509,
+      "step": 16990
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7015123899526652e-06,
+      "loss": 1.0491,
+      "step": 16991
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7006432626002945e-06,
+      "loss": 1.1893,
+      "step": 16992
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6997743366489072e-06,
+      "loss": 1.1003,
+      "step": 16993
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6989056121195858e-06,
+      "loss": 1.0562,
+      "step": 16994
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.698037089033413e-06,
+      "loss": 0.9955,
+      "step": 16995
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6971687674114612e-06,
+      "loss": 1.0012,
+      "step": 16996
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.696300647274809e-06,
+      "loss": 1.1532,
+      "step": 16997
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6954327286445205e-06,
+      "loss": 1.1201,
+      "step": 16998
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6945650115416556e-06,
+      "loss": 1.1278,
+      "step": 16999
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.693697495987271e-06,
+      "loss": 0.9266,
+      "step": 17000
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6928301820024195e-06,
+      "loss": 1.2752,
+      "step": 17001
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6919630696081513e-06,
+      "loss": 1.0222,
+      "step": 17002
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6910961588255071e-06,
+      "loss": 1.1702,
+      "step": 17003
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6902294496755233e-06,
+      "loss": 1.0502,
+      "step": 17004
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6893629421792302e-06,
+      "loss": 1.103,
+      "step": 17005
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6884966363576593e-06,
+      "loss": 1.0495,
+      "step": 17006
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.687630532231833e-06,
+      "loss": 1.13,
+      "step": 17007
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.686764629822767e-06,
+      "loss": 1.0658,
+      "step": 17008
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6858989291514737e-06,
+      "loss": 1.1619,
+      "step": 17009
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6850334302389638e-06,
+      "loss": 1.1251,
+      "step": 17010
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6841681331062398e-06,
+      "loss": 1.0742,
+      "step": 17011
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6833030377742988e-06,
+      "loss": 1.0803,
+      "step": 17012
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6824381442641325e-06,
+      "loss": 1.1516,
+      "step": 17013
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6815734525967332e-06,
+      "loss": 1.1369,
+      "step": 17014
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.680708962793084e-06,
+      "loss": 1.1917,
+      "step": 17015
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6798446748741603e-06,
+      "loss": 1.0661,
+      "step": 17016
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6789805888609368e-06,
+      "loss": 1.0204,
+      "step": 17017
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.678116704774384e-06,
+      "loss": 1.0849,
+      "step": 17018
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6772530226354666e-06,
+      "loss": 1.2189,
+      "step": 17019
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.676389542465141e-06,
+      "loss": 0.9542,
+      "step": 17020
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6755262642843607e-06,
+      "loss": 1.0363,
+      "step": 17021
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6746631881140775e-06,
+      "loss": 1.1984,
+      "step": 17022
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6738003139752358e-06,
+      "loss": 1.075,
+      "step": 17023
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6729376418887722e-06,
+      "loss": 1.2481,
+      "step": 17024
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6720751718756244e-06,
+      "loss": 1.0692,
+      "step": 17025
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6712129039567215e-06,
+      "loss": 1.0638,
+      "step": 17026
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6703508381529865e-06,
+      "loss": 1.13,
+      "step": 17027
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6694889744853381e-06,
+      "loss": 1.1406,
+      "step": 17028
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6686273129746954e-06,
+      "loss": 1.2561,
+      "step": 17029
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6677658536419661e-06,
+      "loss": 0.9525,
+      "step": 17030
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6669045965080555e-06,
+      "loss": 1.0553,
+      "step": 17031
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6660435415938615e-06,
+      "loss": 1.1222,
+      "step": 17032
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6651826889202827e-06,
+      "loss": 1.1807,
+      "step": 17033
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6643220385082092e-06,
+      "loss": 0.9763,
+      "step": 17034
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6634615903785255e-06,
+      "loss": 1.0213,
+      "step": 17035
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6626013445521094e-06,
+      "loss": 0.9782,
+      "step": 17036
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6617413010498419e-06,
+      "loss": 1.1949,
+      "step": 17037
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6608814598925904e-06,
+      "loss": 1.0168,
+      "step": 17038
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.660021821101222e-06,
+      "loss": 1.1888,
+      "step": 17039
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6591623846965942e-06,
+      "loss": 1.1468,
+      "step": 17040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6583031506995684e-06,
+      "loss": 1.1809,
+      "step": 17041
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6574441191309941e-06,
+      "loss": 1.1922,
+      "step": 17042
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6565852900117153e-06,
+      "loss": 1.167,
+      "step": 17043
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6557266633625724e-06,
+      "loss": 1.1612,
+      "step": 17044
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6548682392044057e-06,
+      "loss": 1.0394,
+      "step": 17045
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6540100175580454e-06,
+      "loss": 1.2515,
+      "step": 17046
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6531519984443167e-06,
+      "loss": 1.0052,
+      "step": 17047
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6522941818840398e-06,
+      "loss": 1.0721,
+      "step": 17048
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6514365678980349e-06,
+      "loss": 1.0622,
+      "step": 17049
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6505791565071139e-06,
+      "loss": 1.052,
+      "step": 17050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6497219477320804e-06,
+      "loss": 1.1026,
+      "step": 17051
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6488649415937364e-06,
+      "loss": 1.0801,
+      "step": 17052
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.648008138112882e-06,
+      "loss": 1.1284,
+      "step": 17053
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6471515373103086e-06,
+      "loss": 1.0294,
+      "step": 17054
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6462951392068027e-06,
+      "loss": 1.1304,
+      "step": 17055
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6454389438231433e-06,
+      "loss": 1.1461,
+      "step": 17056
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6445829511801137e-06,
+      "loss": 1.0747,
+      "step": 17057
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.643727161298484e-06,
+      "loss": 1.1552,
+      "step": 17058
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.642871574199021e-06,
+      "loss": 1.0313,
+      "step": 17059
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6420161899024845e-06,
+      "loss": 1.1513,
+      "step": 17060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.641161008429637e-06,
+      "loss": 1.036,
+      "step": 17061
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6403060298012309e-06,
+      "loss": 1.1849,
+      "step": 17062
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6394512540380114e-06,
+      "loss": 1.0593,
+      "step": 17063
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.63859668116072e-06,
+      "loss": 1.1727,
+      "step": 17064
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6377423111900993e-06,
+      "loss": 0.9064,
+      "step": 17065
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.636888144146881e-06,
+      "loss": 1.2015,
+      "step": 17066
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.636034180051792e-06,
+      "loss": 1.1362,
+      "step": 17067
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.635180418925554e-06,
+      "loss": 1.011,
+      "step": 17068
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6343268607888896e-06,
+      "loss": 1.1421,
+      "step": 17069
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6334735056625096e-06,
+      "loss": 1.2139,
+      "step": 17070
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6326203535671191e-06,
+      "loss": 1.1265,
+      "step": 17071
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6317674045234288e-06,
+      "loss": 1.1493,
+      "step": 17072
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6309146585521329e-06,
+      "loss": 1.04,
+      "step": 17073
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.630062115673925e-06,
+      "loss": 1.0086,
+      "step": 17074
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6292097759094927e-06,
+      "loss": 1.0319,
+      "step": 17075
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6283576392795242e-06,
+      "loss": 1.1867,
+      "step": 17076
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.627505705804696e-06,
+      "loss": 1.2568,
+      "step": 17077
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.626653975505681e-06,
+      "loss": 1.0105,
+      "step": 17078
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6258024484031465e-06,
+      "loss": 1.0882,
+      "step": 17079
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6249511245177619e-06,
+      "loss": 1.1023,
+      "step": 17080
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6241000038701827e-06,
+      "loss": 1.0697,
+      "step": 17081
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6232490864810634e-06,
+      "loss": 1.0811,
+      "step": 17082
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6223983723710523e-06,
+      "loss": 1.1605,
+      "step": 17083
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6215478615607948e-06,
+      "loss": 1.2123,
+      "step": 17084
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6206975540709325e-06,
+      "loss": 1.1128,
+      "step": 17085
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6198474499220985e-06,
+      "loss": 1.2066,
+      "step": 17086
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.618997549134922e-06,
+      "loss": 1.2438,
+      "step": 17087
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6181478517300243e-06,
+      "loss": 1.1754,
+      "step": 17088
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.61729835772803e-06,
+      "loss": 1.1248,
+      "step": 17089
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6164490671495525e-06,
+      "loss": 1.012,
+      "step": 17090
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.615599980015201e-06,
+      "loss": 1.1506,
+      "step": 17091
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.614751096345578e-06,
+      "loss": 1.0916,
+      "step": 17092
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6139024161612882e-06,
+      "loss": 1.0563,
+      "step": 17093
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6130539394829237e-06,
+      "loss": 0.9797,
+      "step": 17094
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.612205666331076e-06,
+      "loss": 0.9851,
+      "step": 17095
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6113575967263262e-06,
+      "loss": 1.169,
+      "step": 17096
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.610509730689259e-06,
+      "loss": 1.1745,
+      "step": 17097
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6096620682404496e-06,
+      "loss": 1.1376,
+      "step": 17098
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6088146094004652e-06,
+      "loss": 1.058,
+      "step": 17099
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.607967354189871e-06,
+      "loss": 1.1375,
+      "step": 17100
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6071203026292315e-06,
+      "loss": 1.1078,
+      "step": 17101
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.606273454739099e-06,
+      "loss": 0.963,
+      "step": 17102
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.605426810540026e-06,
+      "loss": 1.1513,
+      "step": 17103
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.604580370052553e-06,
+      "loss": 1.0331,
+      "step": 17104
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.603734133297228e-06,
+      "loss": 1.0936,
+      "step": 17105
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6028881002945817e-06,
+      "loss": 1.1107,
+      "step": 17106
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6020422710651474e-06,
+      "loss": 1.2149,
+      "step": 17107
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.601196645629447e-06,
+      "loss": 1.0685,
+      "step": 17108
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6003512240080054e-06,
+      "loss": 1.0583,
+      "step": 17109
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5995060062213385e-06,
+      "loss": 1.1594,
+      "step": 17110
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.598660992289952e-06,
+      "loss": 1.0675,
+      "step": 17111
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5978161822343596e-06,
+      "loss": 1.0666,
+      "step": 17112
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.596971576075058e-06,
+      "loss": 1.0279,
+      "step": 17113
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5961271738325434e-06,
+      "loss": 0.9998,
+      "step": 17114
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5952829755273057e-06,
+      "loss": 1.1897,
+      "step": 17115
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5944389811798344e-06,
+      "loss": 1.046,
+      "step": 17116
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5935951908106095e-06,
+      "loss": 1.1465,
+      "step": 17117
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5927516044401071e-06,
+      "loss": 1.0727,
+      "step": 17118
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.591908222088796e-06,
+      "loss": 1.141,
+      "step": 17119
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.591065043777148e-06,
+      "loss": 1.1095,
+      "step": 17120
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.590222069525621e-06,
+      "loss": 1.188,
+      "step": 17121
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.589379299354672e-06,
+      "loss": 1.0336,
+      "step": 17122
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.58853673328475e-06,
+      "loss": 1.1699,
+      "step": 17123
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.587694371336308e-06,
+      "loss": 1.1122,
+      "step": 17124
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5868522135297827e-06,
+      "loss": 0.9681,
+      "step": 17125
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.586010259885612e-06,
+      "loss": 1.1603,
+      "step": 17126
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5851685104242254e-06,
+      "loss": 1.1539,
+      "step": 17127
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5843269651660554e-06,
+      "loss": 1.0681,
+      "step": 17128
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.583485624131519e-06,
+      "loss": 1.1558,
+      "step": 17129
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5826444873410352e-06,
+      "loss": 1.021,
+      "step": 17130
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5818035548150124e-06,
+      "loss": 0.987,
+      "step": 17131
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5809628265738619e-06,
+      "loss": 0.9277,
+      "step": 17132
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5801223026379854e-06,
+      "loss": 1.2321,
+      "step": 17133
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5792819830277783e-06,
+      "loss": 0.9887,
+      "step": 17134
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5784418677636304e-06,
+      "loss": 1.1897,
+      "step": 17135
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.577601956865934e-06,
+      "loss": 1.1811,
+      "step": 17136
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5767622503550683e-06,
+      "loss": 1.2535,
+      "step": 17137
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5759227482514106e-06,
+      "loss": 1.0694,
+      "step": 17138
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5750834505753322e-06,
+      "loss": 0.9826,
+      "step": 17139
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5742443573472032e-06,
+      "loss": 1.0749,
+      "step": 17140
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5734054685873856e-06,
+      "loss": 1.0524,
+      "step": 17141
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.572566784316234e-06,
+      "loss": 1.1468,
+      "step": 17142
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5717283045541009e-06,
+      "loss": 1.0516,
+      "step": 17143
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.570890029321338e-06,
+      "loss": 1.0826,
+      "step": 17144
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5700519586382834e-06,
+      "loss": 1.0941,
+      "step": 17145
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5692140925252776e-06,
+      "loss": 1.0791,
+      "step": 17146
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5683764310026496e-06,
+      "loss": 1.0831,
+      "step": 17147
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5675389740907299e-06,
+      "loss": 1.1978,
+      "step": 17148
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5667017218098412e-06,
+      "loss": 0.9545,
+      "step": 17149
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5658646741803008e-06,
+      "loss": 1.1463,
+      "step": 17150
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5650278312224176e-06,
+      "loss": 1.1079,
+      "step": 17151
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5641911929565056e-06,
+      "loss": 1.1518,
+      "step": 17152
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5633547594028642e-06,
+      "loss": 1.1835,
+      "step": 17153
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5625185305817892e-06,
+      "loss": 1.1306,
+      "step": 17154
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5616825065135778e-06,
+      "loss": 1.0555,
+      "step": 17155
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5608466872185157e-06,
+      "loss": 1.198,
+      "step": 17156
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.560011072716886e-06,
+      "loss": 1.1297,
+      "step": 17157
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5591756630289645e-06,
+      "loss": 1.1423,
+      "step": 17158
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5583404581750283e-06,
+      "loss": 1.0558,
+      "step": 17159
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5575054581753435e-06,
+      "loss": 1.0569,
+      "step": 17160
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5566706630501715e-06,
+      "loss": 1.1209,
+      "step": 17161
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5558360728197707e-06,
+      "loss": 1.1242,
+      "step": 17162
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.555001687504396e-06,
+      "loss": 1.1433,
+      "step": 17163
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5541675071242956e-06,
+      "loss": 1.0961,
+      "step": 17164
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.553333531699711e-06,
+      "loss": 1.0656,
+      "step": 17165
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5524997612508785e-06,
+      "loss": 1.2707,
+      "step": 17166
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.551666195798034e-06,
+      "loss": 1.0401,
+      "step": 17167
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5508328353614065e-06,
+      "loss": 1.2515,
+      "step": 17168
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5499996799612193e-06,
+      "loss": 1.1689,
+      "step": 17169
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5491667296176882e-06,
+      "loss": 1.0069,
+      "step": 17170
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5483339843510258e-06,
+      "loss": 1.0066,
+      "step": 17171
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.547501444181445e-06,
+      "loss": 1.1314,
+      "step": 17172
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5466691091291452e-06,
+      "loss": 1.1184,
+      "step": 17173
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.545836979214327e-06,
+      "loss": 1.0906,
+      "step": 17174
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5450050544571792e-06,
+      "loss": 1.1796,
+      "step": 17175
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5441733348778965e-06,
+      "loss": 1.0941,
+      "step": 17176
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5433418204966588e-06,
+      "loss": 1.0903,
+      "step": 17177
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5425105113336448e-06,
+      "loss": 1.0108,
+      "step": 17178
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5416794074090258e-06,
+      "loss": 1.0039,
+      "step": 17179
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5408485087429747e-06,
+      "loss": 1.0833,
+      "step": 17180
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5400178153556533e-06,
+      "loss": 1.09,
+      "step": 17181
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5391873272672197e-06,
+      "loss": 1.1161,
+      "step": 17182
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5383570444978245e-06,
+      "loss": 1.0792,
+      "step": 17183
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5375269670676218e-06,
+      "loss": 1.2853,
+      "step": 17184
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5366970949967518e-06,
+      "loss": 1.1722,
+      "step": 17185
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5358674283053543e-06,
+      "loss": 1.0617,
+      "step": 17186
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5350379670135596e-06,
+      "loss": 1.1539,
+      "step": 17187
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5342087111415015e-06,
+      "loss": 1.1314,
+      "step": 17188
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.533379660709301e-06,
+      "loss": 1.1562,
+      "step": 17189
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5325508157370761e-06,
+      "loss": 1.0834,
+      "step": 17190
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5317221762449398e-06,
+      "loss": 1.0581,
+      "step": 17191
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5308937422530024e-06,
+      "loss": 1.1688,
+      "step": 17192
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5300655137813692e-06,
+      "loss": 1.1628,
+      "step": 17193
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5292374908501328e-06,
+      "loss": 1.2295,
+      "step": 17194
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.528409673479394e-06,
+      "loss": 1.2253,
+      "step": 17195
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.527582061689239e-06,
+      "loss": 1.0132,
+      "step": 17196
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5267546554997493e-06,
+      "loss": 1.098,
+      "step": 17197
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5259274549310043e-06,
+      "loss": 1.0692,
+      "step": 17198
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5251004600030804e-06,
+      "loss": 1.2143,
+      "step": 17199
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5242736707360438e-06,
+      "loss": 0.9444,
+      "step": 17200
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5234470871499607e-06,
+      "loss": 1.054,
+      "step": 17201
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5226207092648849e-06,
+      "loss": 1.0718,
+      "step": 17202
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5217945371008746e-06,
+      "loss": 1.2152,
+      "step": 17203
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5209685706779786e-06,
+      "loss": 1.0332,
+      "step": 17204
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5201428100162397e-06,
+      "loss": 1.2265,
+      "step": 17205
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5193172551356939e-06,
+      "loss": 1.1629,
+      "step": 17206
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5184919060563796e-06,
+      "loss": 1.167,
+      "step": 17207
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5176667627983233e-06,
+      "loss": 1.114,
+      "step": 17208
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5168418253815498e-06,
+      "loss": 1.1682,
+      "step": 17209
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5160170938260742e-06,
+      "loss": 1.0077,
+      "step": 17210
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5151925681519152e-06,
+      "loss": 0.9887,
+      "step": 17211
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5143682483790812e-06,
+      "loss": 0.9856,
+      "step": 17212
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5135441345275726e-06,
+      "loss": 1.143,
+      "step": 17213
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.512720226617389e-06,
+      "loss": 0.9614,
+      "step": 17214
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5118965246685269e-06,
+      "loss": 0.9745,
+      "step": 17215
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5110730287009744e-06,
+      "loss": 1.0898,
+      "step": 17216
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5102497387347138e-06,
+      "loss": 1.1533,
+      "step": 17217
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5094266547897219e-06,
+      "loss": 1.0816,
+      "step": 17218
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5086037768859784e-06,
+      "loss": 1.0801,
+      "step": 17219
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5077811050434477e-06,
+      "loss": 1.1315,
+      "step": 17220
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5069586392820968e-06,
+      "loss": 1.0484,
+      "step": 17221
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5061363796218787e-06,
+      "loss": 1.1475,
+      "step": 17222
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5053143260827542e-06,
+      "loss": 1.0843,
+      "step": 17223
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5044924786846681e-06,
+      "loss": 1.0506,
+      "step": 17224
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5036708374475662e-06,
+      "loss": 1.0453,
+      "step": 17225
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.502849402391383e-06,
+      "loss": 1.1672,
+      "step": 17226
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5020281735360588e-06,
+      "loss": 1.1309,
+      "step": 17227
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5012071509015192e-06,
+      "loss": 1.3319,
+      "step": 17228
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5003863345076875e-06,
+      "loss": 1.2682,
+      "step": 17229
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4995657243744798e-06,
+      "loss": 1.0881,
+      "step": 17230
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4987453205218172e-06,
+      "loss": 1.003,
+      "step": 17231
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4979251229696034e-06,
+      "loss": 1.1051,
+      "step": 17232
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4971051317377427e-06,
+      "loss": 1.1138,
+      "step": 17233
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4962853468461336e-06,
+      "loss": 1.0297,
+      "step": 17234
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.495465768314671e-06,
+      "loss": 1.1529,
+      "step": 17235
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.494646396163245e-06,
+      "loss": 1.1884,
+      "step": 17236
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4938272304117362e-06,
+      "loss": 1.0432,
+      "step": 17237
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4930082710800232e-06,
+      "loss": 1.0893,
+      "step": 17238
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4921895181879841e-06,
+      "loss": 0.9612,
+      "step": 17239
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4913709717554836e-06,
+      "loss": 1.1701,
+      "step": 17240
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4905526318023856e-06,
+      "loss": 1.0796,
+      "step": 17241
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4897344983485518e-06,
+      "loss": 1.0424,
+      "step": 17242
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4889165714138331e-06,
+      "loss": 0.9925,
+      "step": 17243
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4880988510180794e-06,
+      "loss": 0.9745,
+      "step": 17244
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4872813371811324e-06,
+      "loss": 1.0191,
+      "step": 17245
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4864640299228339e-06,
+      "loss": 1.089,
+      "step": 17246
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4856469292630161e-06,
+      "loss": 1.1906,
+      "step": 17247
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4848300352215073e-06,
+      "loss": 1.0807,
+      "step": 17248
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4840133478181296e-06,
+      "loss": 1.1441,
+      "step": 17249
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4831968670727038e-06,
+      "loss": 1.1924,
+      "step": 17250
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4823805930050439e-06,
+      "loss": 1.0601,
+      "step": 17251
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4815645256349587e-06,
+      "loss": 1.0247,
+      "step": 17252
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.48074866498225e-06,
+      "loss": 1.1759,
+      "step": 17253
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4799330110667142e-06,
+      "loss": 1.0769,
+      "step": 17254
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4791175639081513e-06,
+      "loss": 1.021,
+      "step": 17255
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.478302323526345e-06,
+      "loss": 1.2697,
+      "step": 17256
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.477487289941081e-06,
+      "loss": 1.2055,
+      "step": 17257
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.476672463172133e-06,
+      "loss": 1.1825,
+      "step": 17258
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.475857843239281e-06,
+      "loss": 1.2126,
+      "step": 17259
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4750434301622907e-06,
+      "loss": 1.2471,
+      "step": 17260
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4742292239609257e-06,
+      "loss": 1.1486,
+      "step": 17261
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4734152246549417e-06,
+      "loss": 1.1474,
+      "step": 17262
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4726014322640958e-06,
+      "loss": 1.172,
+      "step": 17263
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.471787846808136e-06,
+      "loss": 1.1111,
+      "step": 17264
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.470974468306804e-06,
+      "loss": 1.0554,
+      "step": 17265
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4701612967798363e-06,
+      "loss": 1.0304,
+      "step": 17266
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4693483322469703e-06,
+      "loss": 1.0747,
+      "step": 17267
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.468535574727934e-06,
+      "loss": 1.2087,
+      "step": 17268
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4677230242424478e-06,
+      "loss": 1.1313,
+      "step": 17269
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4669106808102307e-06,
+      "loss": 1.1125,
+      "step": 17270
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4660985444509968e-06,
+      "loss": 1.1345,
+      "step": 17271
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4652866151844548e-06,
+      "loss": 0.9429,
+      "step": 17272
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4644748930303077e-06,
+      "loss": 1.0849,
+      "step": 17273
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4636633780082499e-06,
+      "loss": 1.0475,
+      "step": 17274
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.462852070137979e-06,
+      "loss": 1.0165,
+      "step": 17275
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4620409694391824e-06,
+      "loss": 0.9662,
+      "step": 17276
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4612300759315423e-06,
+      "loss": 1.1315,
+      "step": 17277
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.460419389634734e-06,
+      "loss": 1.1247,
+      "step": 17278
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4596089105684352e-06,
+      "loss": 1.2148,
+      "step": 17279
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4587986387523123e-06,
+      "loss": 1.0441,
+      "step": 17280
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4579885742060252e-06,
+      "loss": 1.1903,
+      "step": 17281
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4571787169492369e-06,
+      "loss": 1.1689,
+      "step": 17282
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.456369067001596e-06,
+      "loss": 1.0151,
+      "step": 17283
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.455559624382754e-06,
+      "loss": 1.0259,
+      "step": 17284
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4547503891123483e-06,
+      "loss": 1.1798,
+      "step": 17285
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4539413612100228e-06,
+      "loss": 1.0568,
+      "step": 17286
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4531325406954066e-06,
+      "loss": 1.092,
+      "step": 17287
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4523239275881284e-06,
+      "loss": 1.0527,
+      "step": 17288
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4515155219078091e-06,
+      "loss": 1.0482,
+      "step": 17289
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4507073236740698e-06,
+      "loss": 0.9197,
+      "step": 17290
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4498993329065203e-06,
+      "loss": 0.9932,
+      "step": 17291
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4490915496247704e-06,
+      "loss": 1.045,
+      "step": 17292
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4482839738484178e-06,
+      "loss": 1.0603,
+      "step": 17293
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4474766055970668e-06,
+      "loss": 1.2048,
+      "step": 17294
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4466694448903063e-06,
+      "loss": 1.147,
+      "step": 17295
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4458624917477238e-06,
+      "loss": 1.0817,
+      "step": 17296
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4450557461888993e-06,
+      "loss": 1.1149,
+      "step": 17297
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4442492082334148e-06,
+      "loss": 1.0881,
+      "step": 17298
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4434428779008403e-06,
+      "loss": 1.1642,
+      "step": 17299
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4426367552107435e-06,
+      "loss": 1.2666,
+      "step": 17300
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4418308401826842e-06,
+      "loss": 1.1197,
+      "step": 17301
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4410251328362245e-06,
+      "loss": 1.1937,
+      "step": 17302
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4402196331909136e-06,
+      "loss": 1.0261,
+      "step": 17303
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4394143412662987e-06,
+      "loss": 1.1934,
+      "step": 17304
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4386092570819188e-06,
+      "loss": 1.1451,
+      "step": 17305
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4378043806573173e-06,
+      "loss": 1.2168,
+      "step": 17306
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4369997120120228e-06,
+      "loss": 1.1258,
+      "step": 17307
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4361952511655618e-06,
+      "loss": 1.1669,
+      "step": 17308
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4353909981374559e-06,
+      "loss": 1.1663,
+      "step": 17309
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4345869529472235e-06,
+      "loss": 1.0239,
+      "step": 17310
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4337831156143767e-06,
+      "loss": 1.2682,
+      "step": 17311
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4329794861584212e-06,
+      "loss": 1.0379,
+      "step": 17312
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.432176064598857e-06,
+      "loss": 1.1683,
+      "step": 17313
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4313728509551838e-06,
+      "loss": 1.1116,
+      "step": 17314
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.430569845246893e-06,
+      "loss": 1.1883,
+      "step": 17315
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.42976704749347e-06,
+      "loss": 1.0795,
+      "step": 17316
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4289644577143947e-06,
+      "loss": 1.108,
+      "step": 17317
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.428162075929148e-06,
+      "loss": 1.1844,
+      "step": 17318
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4273599021571982e-06,
+      "loss": 1.1582,
+      "step": 17319
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4265579364180126e-06,
+      "loss": 1.068,
+      "step": 17320
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4257561787310492e-06,
+      "loss": 1.1119,
+      "step": 17321
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4249546291157712e-06,
+      "loss": 1.1585,
+      "step": 17322
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4241532875916253e-06,
+      "loss": 1.0814,
+      "step": 17323
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4233521541780583e-06,
+      "loss": 0.9449,
+      "step": 17324
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4225512288945088e-06,
+      "loss": 1.0781,
+      "step": 17325
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4217505117604192e-06,
+      "loss": 0.9846,
+      "step": 17326
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.420950002795216e-06,
+      "loss": 1.1885,
+      "step": 17327
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4201497020183241e-06,
+      "loss": 1.0567,
+      "step": 17328
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4193496094491688e-06,
+      "loss": 1.1049,
+      "step": 17329
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4185497251071633e-06,
+      "loss": 1.0234,
+      "step": 17330
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.417750049011719e-06,
+      "loss": 0.942,
+      "step": 17331
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4169505811822393e-06,
+      "loss": 1.105,
+      "step": 17332
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4161513216381273e-06,
+      "loss": 1.0848,
+      "step": 17333
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.415352270398781e-06,
+      "loss": 1.1658,
+      "step": 17334
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4145534274835882e-06,
+      "loss": 1.1809,
+      "step": 17335
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4137547929119343e-06,
+      "loss": 0.9451,
+      "step": 17336
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4129563667031987e-06,
+      "loss": 1.1416,
+      "step": 17337
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.412158148876761e-06,
+      "loss": 0.9899,
+      "step": 17338
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4113601394519894e-06,
+      "loss": 1.1416,
+      "step": 17339
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.410562338448248e-06,
+      "loss": 1.1636,
+      "step": 17340
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4097647458848961e-06,
+      "loss": 1.2844,
+      "step": 17341
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4089673617812927e-06,
+      "loss": 1.1501,
+      "step": 17342
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4081701861567864e-06,
+      "loss": 1.1022,
+      "step": 17343
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4073732190307232e-06,
+      "loss": 1.2298,
+      "step": 17344
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4065764604224386e-06,
+      "loss": 1.0655,
+      "step": 17345
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4057799103512725e-06,
+      "loss": 1.0509,
+      "step": 17346
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4049835688365543e-06,
+      "loss": 1.0743,
+      "step": 17347
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4041874358976081e-06,
+      "loss": 0.9227,
+      "step": 17348
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.40339151155375e-06,
+      "loss": 1.0301,
+      "step": 17349
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.402595795824302e-06,
+      "loss": 1.1374,
+      "step": 17350
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4018002887285687e-06,
+      "loss": 1.0554,
+      "step": 17351
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.401004990285857e-06,
+      "loss": 1.1393,
+      "step": 17352
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4002099005154624e-06,
+      "loss": 1.1188,
+      "step": 17353
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3994150194366851e-06,
+      "loss": 1.3052,
+      "step": 17354
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3986203470688109e-06,
+      "loss": 1.0618,
+      "step": 17355
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3978258834311264e-06,
+      "loss": 1.0673,
+      "step": 17356
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3970316285429065e-06,
+      "loss": 1.0479,
+      "step": 17357
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.396237582423432e-06,
+      "loss": 1.0512,
+      "step": 17358
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3954437450919679e-06,
+      "loss": 1.1548,
+      "step": 17359
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3946501165677783e-06,
+      "loss": 0.9788,
+      "step": 17360
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3938566968701217e-06,
+      "loss": 0.9409,
+      "step": 17361
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3930634860182546e-06,
+      "loss": 1.0314,
+      "step": 17362
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.392270484031426e-06,
+      "loss": 1.1607,
+      "step": 17363
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3914776909288774e-06,
+      "loss": 1.0914,
+      "step": 17364
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3906851067298466e-06,
+      "loss": 1.102,
+      "step": 17365
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3898927314535705e-06,
+      "loss": 1.0703,
+      "step": 17366
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3891005651192769e-06,
+      "loss": 0.9789,
+      "step": 17367
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3883086077461872e-06,
+      "loss": 1.1527,
+      "step": 17368
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3875168593535237e-06,
+      "loss": 1.1901,
+      "step": 17369
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3867253199604969e-06,
+      "loss": 1.133,
+      "step": 17370
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3859339895863167e-06,
+      "loss": 1.0919,
+      "step": 17371
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3851428682501821e-06,
+      "loss": 1.0952,
+      "step": 17372
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3843519559712981e-06,
+      "loss": 1.0991,
+      "step": 17373
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3835612527688535e-06,
+      "loss": 1.2389,
+      "step": 17374
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3827707586620375e-06,
+      "loss": 1.0687,
+      "step": 17375
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3819804736700315e-06,
+      "loss": 1.0699,
+      "step": 17376
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3811903978120156e-06,
+      "loss": 1.0993,
+      "step": 17377
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3804005311071633e-06,
+      "loss": 1.0845,
+      "step": 17378
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3796108735746395e-06,
+      "loss": 1.1692,
+      "step": 17379
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3788214252336063e-06,
+      "loss": 1.0164,
+      "step": 17380
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3780321861032252e-06,
+      "loss": 1.1667,
+      "step": 17381
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3772431562026467e-06,
+      "loss": 1.0821,
+      "step": 17382
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3764543355510186e-06,
+      "loss": 1.1688,
+      "step": 17383
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3756657241674798e-06,
+      "loss": 1.1546,
+      "step": 17384
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3748773220711741e-06,
+      "loss": 0.9025,
+      "step": 17385
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3740891292812287e-06,
+      "loss": 1.0331,
+      "step": 17386
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3733011458167734e-06,
+      "loss": 0.8936,
+      "step": 17387
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3725133716969263e-06,
+      "loss": 1.0665,
+      "step": 17388
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3717258069408102e-06,
+      "loss": 1.0942,
+      "step": 17389
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3709384515675327e-06,
+      "loss": 1.204,
+      "step": 17390
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3701513055962024e-06,
+      "loss": 1.0221,
+      "step": 17391
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3693643690459179e-06,
+      "loss": 1.0924,
+      "step": 17392
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.368577641935781e-06,
+      "loss": 1.1762,
+      "step": 17393
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3677911242848807e-06,
+      "loss": 0.9935,
+      "step": 17394
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.367004816112303e-06,
+      "loss": 1.1147,
+      "step": 17395
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3662187174371278e-06,
+      "loss": 1.0145,
+      "step": 17396
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3654328282784358e-06,
+      "loss": 1.1202,
+      "step": 17397
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3646471486552947e-06,
+      "loss": 1.0024,
+      "step": 17398
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.363861678586773e-06,
+      "loss": 1.0838,
+      "step": 17399
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3630764180919276e-06,
+      "loss": 1.0805,
+      "step": 17400
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3622913671898186e-06,
+      "loss": 1.3096,
+      "step": 17401
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3615065258994963e-06,
+      "loss": 1.0611,
+      "step": 17402
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3607218942400057e-06,
+      "loss": 1.021,
+      "step": 17403
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3599374722303848e-06,
+      "loss": 1.1126,
+      "step": 17404
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3591532598896751e-06,
+      "loss": 1.1222,
+      "step": 17405
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3583692572369034e-06,
+      "loss": 1.1607,
+      "step": 17406
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3575854642910957e-06,
+      "loss": 1.1919,
+      "step": 17407
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3568018810712702e-06,
+      "loss": 1.2066,
+      "step": 17408
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.356018507596447e-06,
+      "loss": 1.1106,
+      "step": 17409
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3552353438856324e-06,
+      "loss": 1.057,
+      "step": 17410
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3544523899578343e-06,
+      "loss": 1.1366,
+      "step": 17411
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3536696458320476e-06,
+      "loss": 1.0262,
+      "step": 17412
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3528871115272735e-06,
+      "loss": 0.9569,
+      "step": 17413
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3521047870624982e-06,
+      "loss": 1.0157,
+      "step": 17414
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3513226724567053e-06,
+      "loss": 1.1816,
+      "step": 17415
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3505407677288763e-06,
+      "loss": 1.0086,
+      "step": 17416
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.349759072897988e-06,
+      "loss": 1.0462,
+      "step": 17417
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.348977587983008e-06,
+      "loss": 1.0934,
+      "step": 17418
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3481963130028997e-06,
+      "loss": 1.0698,
+      "step": 17419
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.347415247976621e-06,
+      "loss": 0.9773,
+      "step": 17420
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3466343929231295e-06,
+      "loss": 0.9548,
+      "step": 17421
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3458537478613731e-06,
+      "loss": 1.2244,
+      "step": 17422
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3450733128102954e-06,
+      "loss": 1.0186,
+      "step": 17423
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3442930877888327e-06,
+      "loss": 1.0573,
+      "step": 17424
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.343513072815924e-06,
+      "loss": 1.148,
+      "step": 17425
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3427332679104943e-06,
+      "loss": 1.085,
+      "step": 17426
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3419536730914683e-06,
+      "loss": 1.1587,
+      "step": 17427
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3411742883777624e-06,
+      "loss": 1.0254,
+      "step": 17428
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3403951137882942e-06,
+      "loss": 1.1001,
+      "step": 17429
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.339616149341969e-06,
+      "loss": 1.1487,
+      "step": 17430
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3388373950576906e-06,
+      "loss": 1.1899,
+      "step": 17431
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3380588509543558e-06,
+      "loss": 1.0331,
+      "step": 17432
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3372805170508595e-06,
+      "loss": 1.0981,
+      "step": 17433
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3365023933660903e-06,
+      "loss": 1.1716,
+      "step": 17434
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3357244799189295e-06,
+      "loss": 0.9734,
+      "step": 17435
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.334946776728252e-06,
+      "loss": 0.974,
+      "step": 17436
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.334169283812936e-06,
+      "loss": 1.1378,
+      "step": 17437
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3333920011918455e-06,
+      "loss": 1.0506,
+      "step": 17438
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.332614928883844e-06,
+      "loss": 1.088,
+      "step": 17439
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3318380669077857e-06,
+      "loss": 1.0458,
+      "step": 17440
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3310614152825274e-06,
+      "loss": 1.1697,
+      "step": 17441
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3302849740269152e-06,
+      "loss": 1.1187,
+      "step": 17442
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3295087431597898e-06,
+      "loss": 1.1416,
+      "step": 17443
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3287327226999846e-06,
+      "loss": 1.1201,
+      "step": 17444
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3279569126663384e-06,
+      "loss": 1.3001,
+      "step": 17445
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3271813130776745e-06,
+      "loss": 1.1514,
+      "step": 17446
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3264059239528137e-06,
+      "loss": 1.102,
+      "step": 17447
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.325630745310571e-06,
+      "loss": 1.1372,
+      "step": 17448
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3248557771697613e-06,
+      "loss": 1.1118,
+      "step": 17449
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3240810195491904e-06,
+      "loss": 1.139,
+      "step": 17450
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3233064724676548e-06,
+      "loss": 1.0998,
+      "step": 17451
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3225321359439568e-06,
+      "loss": 1.1362,
+      "step": 17452
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3217580099968841e-06,
+      "loss": 1.1659,
+      "step": 17453
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3209840946452236e-06,
+      "loss": 1.0749,
+      "step": 17454
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3202103899077512e-06,
+      "loss": 1.1101,
+      "step": 17455
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.31943689580325e-06,
+      "loss": 1.0543,
+      "step": 17456
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3186636123504858e-06,
+      "loss": 1.0606,
+      "step": 17457
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.317890539568225e-06,
+      "loss": 1.0845,
+      "step": 17458
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3171176774752258e-06,
+      "loss": 1.2138,
+      "step": 17459
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3163450260902466e-06,
+      "loss": 1.0964,
+      "step": 17460
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3155725854320367e-06,
+      "loss": 1.1326,
+      "step": 17461
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.31480035551934e-06,
+      "loss": 1.0934,
+      "step": 17462
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3140283363708938e-06,
+      "loss": 1.1212,
+      "step": 17463
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3132565280054376e-06,
+      "loss": 1.1666,
+      "step": 17464
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3124849304416986e-06,
+      "loss": 1.0678,
+      "step": 17465
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.311713543698402e-06,
+      "loss": 1.2177,
+      "step": 17466
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3109423677942635e-06,
+      "loss": 1.2053,
+      "step": 17467
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3101714027480027e-06,
+      "loss": 1.2039,
+      "step": 17468
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.309400648578326e-06,
+      "loss": 1.0394,
+      "step": 17469
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3086301053039373e-06,
+      "loss": 1.0583,
+      "step": 17470
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.307859772943534e-06,
+      "loss": 1.0943,
+      "step": 17471
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3070896515158128e-06,
+      "loss": 1.0875,
+      "step": 17472
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3063197410394612e-06,
+      "loss": 0.9542,
+      "step": 17473
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3055500415331613e-06,
+      "loss": 1.2272,
+      "step": 17474
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.304780553015591e-06,
+      "loss": 1.184,
+      "step": 17475
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.304011275505428e-06,
+      "loss": 1.0904,
+      "step": 17476
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.303242209021336e-06,
+      "loss": 1.2028,
+      "step": 17477
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.30247335358198e-06,
+      "loss": 1.1788,
+      "step": 17478
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3017047092060143e-06,
+      "loss": 1.1298,
+      "step": 17479
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.300936275912098e-06,
+      "loss": 1.2377,
+      "step": 17480
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3001680537188744e-06,
+      "loss": 1.2086,
+      "step": 17481
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2994000426449872e-06,
+      "loss": 1.1975,
+      "step": 17482
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2986322427090703e-06,
+      "loss": 1.1513,
+      "step": 17483
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2978646539297623e-06,
+      "loss": 1.1406,
+      "step": 17484
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2970972763256872e-06,
+      "loss": 1.1444,
+      "step": 17485
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2963301099154656e-06,
+      "loss": 1.0404,
+      "step": 17486
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.295563154717715e-06,
+      "loss": 1.1263,
+      "step": 17487
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.294796410751049e-06,
+      "loss": 0.9857,
+      "step": 17488
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2940298780340732e-06,
+      "loss": 0.8969,
+      "step": 17489
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.293263556585389e-06,
+      "loss": 1.2172,
+      "step": 17490
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2924974464235905e-06,
+      "loss": 1.1994,
+      "step": 17491
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2917315475672743e-06,
+      "loss": 1.128,
+      "step": 17492
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2909658600350227e-06,
+      "loss": 0.9259,
+      "step": 17493
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.290200383845417e-06,
+      "loss": 0.9958,
+      "step": 17494
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2894351190170306e-06,
+      "loss": 1.1034,
+      "step": 17495
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2886700655684404e-06,
+      "loss": 0.9851,
+      "step": 17496
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2879052235182077e-06,
+      "loss": 1.016,
+      "step": 17497
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2871405928848924e-06,
+      "loss": 1.0126,
+      "step": 17498
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2863761736870527e-06,
+      "loss": 1.0703,
+      "step": 17499
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2856119659432365e-06,
+      "loss": 1.1911,
+      "step": 17500
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2848479696719906e-06,
+      "loss": 1.1218,
+      "step": 17501
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.284084184891855e-06,
+      "loss": 1.1306,
+      "step": 17502
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2833206116213615e-06,
+      "loss": 1.1529,
+      "step": 17503
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2825572498790438e-06,
+      "loss": 1.1874,
+      "step": 17504
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2817940996834243e-06,
+      "loss": 1.0804,
+      "step": 17505
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2810311610530223e-06,
+      "loss": 1.1886,
+      "step": 17506
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.280268434006351e-06,
+      "loss": 1.036,
+      "step": 17507
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.279505918561923e-06,
+      "loss": 0.9618,
+      "step": 17508
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2787436147382404e-06,
+      "loss": 0.941,
+      "step": 17509
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2779815225538028e-06,
+      "loss": 1.0647,
+      "step": 17510
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2772196420270999e-06,
+      "loss": 1.1476,
+      "step": 17511
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2764579731766257e-06,
+      "loss": 1.0286,
+      "step": 17512
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2756965160208624e-06,
+      "loss": 1.1098,
+      "step": 17513
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2749352705782858e-06,
+      "loss": 1.0972,
+      "step": 17514
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2741742368673693e-06,
+      "loss": 1.1862,
+      "step": 17515
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2734134149065846e-06,
+      "loss": 1.09,
+      "step": 17516
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2726528047143927e-06,
+      "loss": 1.1431,
+      "step": 17517
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.271892406309251e-06,
+      "loss": 1.1419,
+      "step": 17518
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2711322197096098e-06,
+      "loss": 1.0589,
+      "step": 17519
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2703722449339206e-06,
+      "loss": 1.1316,
+      "step": 17520
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.269612482000625e-06,
+      "loss": 1.1924,
+      "step": 17521
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2688529309281583e-06,
+      "loss": 0.8448,
+      "step": 17522
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2680935917349524e-06,
+      "loss": 0.9938,
+      "step": 17523
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2673344644394369e-06,
+      "loss": 1.1501,
+      "step": 17524
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2665755490600318e-06,
+      "loss": 1.2609,
+      "step": 17525
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2658168456151543e-06,
+      "loss": 1.0546,
+      "step": 17526
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2650583541232142e-06,
+      "loss": 1.1458,
+      "step": 17527
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2643000746026213e-06,
+      "loss": 1.0602,
+      "step": 17528
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.263542007071774e-06,
+      "loss": 1.0993,
+      "step": 17529
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2627841515490703e-06,
+      "loss": 1.094,
+      "step": 17530
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.262026508052897e-06,
+      "loss": 1.1111,
+      "step": 17531
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2612690766016455e-06,
+      "loss": 1.151,
+      "step": 17532
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2605118572136932e-06,
+      "loss": 1.0967,
+      "step": 17533
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2597548499074175e-06,
+      "loss": 1.2052,
+      "step": 17534
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2589980547011837e-06,
+      "loss": 1.129,
+      "step": 17535
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.258241471613364e-06,
+      "loss": 1.0233,
+      "step": 17536
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2574851006623145e-06,
+      "loss": 1.2294,
+      "step": 17537
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2567289418663898e-06,
+      "loss": 1.2038,
+      "step": 17538
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2559729952439414e-06,
+      "loss": 0.9761,
+      "step": 17539
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.255217260813314e-06,
+      "loss": 1.1539,
+      "step": 17540
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.254461738592846e-06,
+      "loss": 1.0301,
+      "step": 17541
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2537064286008693e-06,
+      "loss": 1.0338,
+      "step": 17542
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2529513308557184e-06,
+      "loss": 1.1697,
+      "step": 17543
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2521964453757141e-06,
+      "loss": 1.1721,
+      "step": 17544
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.251441772179176e-06,
+      "loss": 1.1208,
+      "step": 17545
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2506873112844154e-06,
+      "loss": 1.141,
+      "step": 17546
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.249933062709745e-06,
+      "loss": 1.1722,
+      "step": 17547
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.249179026473466e-06,
+      "loss": 1.1838,
+      "step": 17548
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.248425202593877e-06,
+      "loss": 1.127,
+      "step": 17549
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.247671591089269e-06,
+      "loss": 1.0649,
+      "step": 17550
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.246918191977934e-06,
+      "loss": 1.0363,
+      "step": 17551
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2461650052781515e-06,
+      "loss": 1.0731,
+      "step": 17552
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.245412031008202e-06,
+      "loss": 1.1345,
+      "step": 17553
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2446592691863524e-06,
+      "loss": 1.0422,
+      "step": 17554
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2439067198308762e-06,
+      "loss": 1.0066,
+      "step": 17555
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2431543829600334e-06,
+      "loss": 1.136,
+      "step": 17556
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2424022585920814e-06,
+      "loss": 1.1112,
+      "step": 17557
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.241650346745269e-06,
+      "loss": 1.1494,
+      "step": 17558
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2408986474378471e-06,
+      "loss": 1.1927,
+      "step": 17559
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2401471606880554e-06,
+      "loss": 1.0463,
+      "step": 17560
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2393958865141308e-06,
+      "loss": 0.9683,
+      "step": 17561
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.238644824934301e-06,
+      "loss": 1.1631,
+      "step": 17562
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2378939759667974e-06,
+      "loss": 0.9886,
+      "step": 17563
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2371433396298393e-06,
+      "loss": 1.2074,
+      "step": 17564
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.236392915941641e-06,
+      "loss": 1.0158,
+      "step": 17565
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2356427049204122e-06,
+      "loss": 1.2891,
+      "step": 17566
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.234892706584362e-06,
+      "loss": 1.008,
+      "step": 17567
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2341429209516885e-06,
+      "loss": 1.0865,
+      "step": 17568
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2333933480405868e-06,
+      "loss": 1.1431,
+      "step": 17569
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2326439878692441e-06,
+      "loss": 1.1475,
+      "step": 17570
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2318948404558517e-06,
+      "loss": 1.0723,
+      "step": 17571
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2311459058185838e-06,
+      "loss": 1.2034,
+      "step": 17572
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2303971839756167e-06,
+      "loss": 1.0193,
+      "step": 17573
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2296486749451187e-06,
+      "loss": 1.1163,
+      "step": 17574
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.228900378745256e-06,
+      "loss": 1.075,
+      "step": 17575
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.228152295394186e-06,
+      "loss": 1.1929,
+      "step": 17576
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2274044249100624e-06,
+      "loss": 1.0973,
+      "step": 17577
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2266567673110318e-06,
+      "loss": 1.2017,
+      "step": 17578
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.225909322615243e-06,
+      "loss": 1.0095,
+      "step": 17579
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2251620908408312e-06,
+      "loss": 1.1238,
+      "step": 17580
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.224415072005929e-06,
+      "loss": 0.986,
+      "step": 17581
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2236682661286636e-06,
+      "loss": 1.2234,
+      "step": 17582
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2229216732271587e-06,
+      "loss": 0.9812,
+      "step": 17583
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.222175293319534e-06,
+      "loss": 1.1264,
+      "step": 17584
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2214291264239009e-06,
+      "loss": 1.1787,
+      "step": 17585
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2206831725583645e-06,
+      "loss": 1.0427,
+      "step": 17586
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2199374317410294e-06,
+      "loss": 1.0329,
+      "step": 17587
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2191919039899925e-06,
+      "loss": 1.2614,
+      "step": 17588
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2184465893233432e-06,
+      "loss": 1.1406,
+      "step": 17589
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2177014877591687e-06,
+      "loss": 1.1707,
+      "step": 17590
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.216956599315553e-06,
+      "loss": 1.0673,
+      "step": 17591
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2162119240105707e-06,
+      "loss": 0.9261,
+      "step": 17592
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.215467461862293e-06,
+      "loss": 1.227,
+      "step": 17593
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2147232128887842e-06,
+      "loss": 1.119,
+      "step": 17594
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2139791771081078e-06,
+      "loss": 1.1846,
+      "step": 17595
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.213235354538318e-06,
+      "loss": 1.0088,
+      "step": 17596
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.212491745197465e-06,
+      "loss": 1.0021,
+      "step": 17597
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.211748349103592e-06,
+      "loss": 1.1243,
+      "step": 17598
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2110051662747434e-06,
+      "loss": 1.0344,
+      "step": 17599
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2102621967289519e-06,
+      "loss": 1.0028,
+      "step": 17600
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2095194404842481e-06,
+      "loss": 1.1157,
+      "step": 17601
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2087768975586522e-06,
+      "loss": 0.9005,
+      "step": 17602
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2080345679701899e-06,
+      "loss": 1.1184,
+      "step": 17603
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2072924517368722e-06,
+      "loss": 1.1339,
+      "step": 17604
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.206550548876708e-06,
+      "loss": 1.2035,
+      "step": 17605
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2058088594077e-06,
+      "loss": 1.0574,
+      "step": 17606
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2050673833478498e-06,
+      "loss": 0.9382,
+      "step": 17607
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.204326120715149e-06,
+      "loss": 1.1056,
+      "step": 17608
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2035850715275865e-06,
+      "loss": 0.9708,
+      "step": 17609
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2028442358031433e-06,
+      "loss": 1.1346,
+      "step": 17610
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2021036135597996e-06,
+      "loss": 1.1787,
+      "step": 17611
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2013632048155288e-06,
+      "loss": 1.248,
+      "step": 17612
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2006230095882965e-06,
+      "loss": 1.1467,
+      "step": 17613
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1998830278960638e-06,
+      "loss": 1.0793,
+      "step": 17614
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.199143259756792e-06,
+      "loss": 1.1316,
+      "step": 17615
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1984037051884312e-06,
+      "loss": 1.0983,
+      "step": 17616
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.197664364208928e-06,
+      "loss": 1.0628,
+      "step": 17617
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1969252368362216e-06,
+      "loss": 1.3459,
+      "step": 17618
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1961863230882531e-06,
+      "loss": 1.1792,
+      "step": 17619
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1954476229829504e-06,
+      "loss": 1.1416,
+      "step": 17620
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1947091365382412e-06,
+      "loss": 1.0843,
+      "step": 17621
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1939708637720438e-06,
+      "loss": 1.0083,
+      "step": 17622
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1932328047022789e-06,
+      "loss": 1.0296,
+      "step": 17623
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1924949593468526e-06,
+      "loss": 1.0782,
+      "step": 17624
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1917573277236705e-06,
+      "loss": 1.119,
+      "step": 17625
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1910199098506358e-06,
+      "loss": 1.0646,
+      "step": 17626
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1902827057456424e-06,
+      "loss": 1.1441,
+      "step": 17627
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1895457154265778e-06,
+      "loss": 1.0464,
+      "step": 17628
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1888089389113267e-06,
+      "loss": 1.0063,
+      "step": 17629
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1880723762177714e-06,
+      "loss": 1.0189,
+      "step": 17630
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1873360273637858e-06,
+      "loss": 1.1724,
+      "step": 17631
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1865998923672373e-06,
+      "loss": 1.055,
+      "step": 17632
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1858639712459875e-06,
+      "loss": 1.1229,
+      "step": 17633
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1851282640178996e-06,
+      "loss": 1.1273,
+      "step": 17634
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1843927707008262e-06,
+      "loss": 1.161,
+      "step": 17635
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.183657491312613e-06,
+      "loss": 1.1001,
+      "step": 17636
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1829224258711036e-06,
+      "loss": 1.1185,
+      "step": 17637
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1821875743941392e-06,
+      "loss": 1.1121,
+      "step": 17638
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1814529368995498e-06,
+      "loss": 1.034,
+      "step": 17639
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1807185134051624e-06,
+      "loss": 1.0376,
+      "step": 17640
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1799843039287984e-06,
+      "loss": 0.9942,
+      "step": 17641
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.179250308488279e-06,
+      "loss": 1.0482,
+      "step": 17642
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1785165271014144e-06,
+      "loss": 1.0297,
+      "step": 17643
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1777829597860092e-06,
+      "loss": 0.9378,
+      "step": 17644
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1770496065598647e-06,
+      "loss": 1.0273,
+      "step": 17645
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.176316467440781e-06,
+      "loss": 1.2106,
+      "step": 17646
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1755835424465478e-06,
+      "loss": 1.0798,
+      "step": 17647
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1748508315949502e-06,
+      "loss": 1.0743,
+      "step": 17648
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1741183349037665e-06,
+      "loss": 1.0892,
+      "step": 17649
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.173386052390777e-06,
+      "loss": 1.1797,
+      "step": 17650
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1726539840737494e-06,
+      "loss": 1.009,
+      "step": 17651
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1719221299704497e-06,
+      "loss": 1.2195,
+      "step": 17652
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1711904900986358e-06,
+      "loss": 1.026,
+      "step": 17653
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.170459064476066e-06,
+      "loss": 0.9521,
+      "step": 17654
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.169727853120488e-06,
+      "loss": 0.9951,
+      "step": 17655
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1689968560496456e-06,
+      "loss": 1.196,
+      "step": 17656
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1682660732812767e-06,
+      "loss": 0.8811,
+      "step": 17657
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1675355048331195e-06,
+      "loss": 1.0349,
+      "step": 17658
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1668051507228995e-06,
+      "loss": 1.1805,
+      "step": 17659
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1660750109683415e-06,
+      "loss": 1.1062,
+      "step": 17660
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1653450855871618e-06,
+      "loss": 1.0856,
+      "step": 17661
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1646153745970768e-06,
+      "loss": 0.9302,
+      "step": 17662
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1638858780157925e-06,
+      "loss": 1.1807,
+      "step": 17663
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1631565958610115e-06,
+      "loss": 1.0837,
+      "step": 17664
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1624275281504294e-06,
+      "loss": 1.2611,
+      "step": 17665
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1616986749017412e-06,
+      "loss": 1.0549,
+      "step": 17666
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.160970036132636e-06,
+      "loss": 1.0517,
+      "step": 17667
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.160241611860794e-06,
+      "loss": 1.1445,
+      "step": 17668
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.159513402103888e-06,
+      "loss": 0.9745,
+      "step": 17669
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1587854068795957e-06,
+      "loss": 1.1154,
+      "step": 17670
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1580576262055798e-06,
+      "loss": 1.1417,
+      "step": 17671
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.157330060099503e-06,
+      "loss": 1.0973,
+      "step": 17672
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1566027085790177e-06,
+      "loss": 1.1254,
+      "step": 17673
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1558755716617786e-06,
+      "loss": 1.18,
+      "step": 17674
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1551486493654307e-06,
+      "loss": 1.0957,
+      "step": 17675
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.154421941707613e-06,
+      "loss": 1.0178,
+      "step": 17676
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1536954487059582e-06,
+      "loss": 1.0491,
+      "step": 17677
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1529691703781022e-06,
+      "loss": 1.1701,
+      "step": 17678
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.152243106741665e-06,
+      "loss": 1.0514,
+      "step": 17679
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1515172578142675e-06,
+      "loss": 0.966,
+      "step": 17680
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.150791623613522e-06,
+      "loss": 1.0835,
+      "step": 17681
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.150066204157041e-06,
+      "loss": 1.0963,
+      "step": 17682
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1493409994624272e-06,
+      "loss": 1.075,
+      "step": 17683
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1486160095472776e-06,
+      "loss": 1.1057,
+      "step": 17684
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1478912344291848e-06,
+      "loss": 0.9928,
+      "step": 17685
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1471666741257404e-06,
+      "loss": 1.1445,
+      "step": 17686
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1464423286545256e-06,
+      "loss": 1.0404,
+      "step": 17687
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1457181980331188e-06,
+      "loss": 1.0026,
+      "step": 17688
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1449942822790894e-06,
+      "loss": 1.1641,
+      "step": 17689
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1442705814100085e-06,
+      "loss": 1.0935,
+      "step": 17690
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.143547095443438e-06,
+      "loss": 1.1367,
+      "step": 17691
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1428238243969325e-06,
+      "loss": 1.0409,
+      "step": 17692
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1421007682880436e-06,
+      "loss": 1.1129,
+      "step": 17693
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1413779271343206e-06,
+      "loss": 1.0248,
+      "step": 17694
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1406553009533028e-06,
+      "loss": 1.0615,
+      "step": 17695
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1399328897625262e-06,
+      "loss": 1.2326,
+      "step": 17696
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.139210693579521e-06,
+      "loss": 1.1686,
+      "step": 17697
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1384887124218147e-06,
+      "loss": 0.9896,
+      "step": 17698
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1377669463069263e-06,
+      "loss": 1.0079,
+      "step": 17699
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1370453952523719e-06,
+      "loss": 1.1451,
+      "step": 17700
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1363240592756586e-06,
+      "loss": 1.0839,
+      "step": 17701
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1356029383942958e-06,
+      "loss": 1.3127,
+      "step": 17702
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1348820326257792e-06,
+      "loss": 1.1467,
+      "step": 17703
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1341613419876052e-06,
+      "loss": 1.1305,
+      "step": 17704
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1334408664972596e-06,
+      "loss": 1.0817,
+      "step": 17705
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1327206061722308e-06,
+      "loss": 1.0531,
+      "step": 17706
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1320005610299944e-06,
+      "loss": 1.02,
+      "step": 17707
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1312807310880257e-06,
+      "loss": 1.1063,
+      "step": 17708
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1305611163637886e-06,
+      "loss": 1.0214,
+      "step": 17709
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1298417168747522e-06,
+      "loss": 1.0935,
+      "step": 17710
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1291225326383703e-06,
+      "loss": 0.964,
+      "step": 17711
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1284035636720948e-06,
+      "loss": 1.0232,
+      "step": 17712
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1276848099933758e-06,
+      "loss": 1.2671,
+      "step": 17713
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1269662716196538e-06,
+      "loss": 1.0517,
+      "step": 17714
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1262479485683663e-06,
+      "loss": 1.1378,
+      "step": 17715
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1255298408569427e-06,
+      "loss": 1.0534,
+      "step": 17716
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.124811948502812e-06,
+      "loss": 1.1248,
+      "step": 17717
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1240942715233948e-06,
+      "loss": 1.2844,
+      "step": 17718
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1233768099361065e-06,
+      "loss": 1.1461,
+      "step": 17719
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.122659563758356e-06,
+      "loss": 1.1278,
+      "step": 17720
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1219425330075518e-06,
+      "loss": 0.9932,
+      "step": 17721
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1212257177010944e-06,
+      "loss": 1.1175,
+      "step": 17722
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1205091178563753e-06,
+      "loss": 1.0842,
+      "step": 17723
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1197927334907854e-06,
+      "loss": 1.0626,
+      "step": 17724
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1190765646217128e-06,
+      "loss": 1.0454,
+      "step": 17725
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1183606112665334e-06,
+      "loss": 1.0525,
+      "step": 17726
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1176448734426214e-06,
+      "loss": 1.1432,
+      "step": 17727
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.116929351167345e-06,
+      "loss": 1.0529,
+      "step": 17728
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1162140444580705e-06,
+      "loss": 1.0365,
+      "step": 17729
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1154989533321558e-06,
+      "loss": 1.1475,
+      "step": 17730
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1147840778069518e-06,
+      "loss": 1.0832,
+      "step": 17731
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1140694178998069e-06,
+      "loss": 1.0861,
+      "step": 17732
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.113354973628067e-06,
+      "loss": 0.9494,
+      "step": 17733
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1126407450090658e-06,
+      "loss": 1.0417,
+      "step": 17734
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.111926732060138e-06,
+      "loss": 1.0301,
+      "step": 17735
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1112129347986068e-06,
+      "loss": 1.1668,
+      "step": 17736
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1104993532417996e-06,
+      "loss": 1.0755,
+      "step": 17737
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1097859874070294e-06,
+      "loss": 1.0794,
+      "step": 17738
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1090728373116077e-06,
+      "loss": 1.1375,
+      "step": 17739
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1083599029728399e-06,
+      "loss": 1.0195,
+      "step": 17740
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1076471844080283e-06,
+      "loss": 1.2076,
+      "step": 17741
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1069346816344695e-06,
+      "loss": 1.1069,
+      "step": 17742
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1062223946694527e-06,
+      "loss": 1.014,
+      "step": 17743
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.10551032353026e-06,
+      "loss": 1.1479,
+      "step": 17744
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1047984682341762e-06,
+      "loss": 1.1213,
+      "step": 17745
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.104086828798473e-06,
+      "loss": 1.2091,
+      "step": 17746
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1033754052404211e-06,
+      "loss": 1.1531,
+      "step": 17747
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1026641975772823e-06,
+      "loss": 1.2346,
+      "step": 17748
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1019532058263194e-06,
+      "loss": 1.1577,
+      "step": 17749
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1012424300047818e-06,
+      "loss": 1.1421,
+      "step": 17750
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1005318701299216e-06,
+      "loss": 0.9841,
+      "step": 17751
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0998215262189794e-06,
+      "loss": 1.0714,
+      "step": 17752
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0991113982891955e-06,
+      "loss": 1.0803,
+      "step": 17753
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0984014863578018e-06,
+      "loss": 1.1359,
+      "step": 17754
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0976917904420247e-06,
+      "loss": 1.0945,
+      "step": 17755
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.096982310559086e-06,
+      "loss": 1.1764,
+      "step": 17756
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.096273046726205e-06,
+      "loss": 1.0988,
+      "step": 17757
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0955639989605925e-06,
+      "loss": 1.1439,
+      "step": 17758
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.094855167279456e-06,
+      "loss": 1.1051,
+      "step": 17759
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0941465516999917e-06,
+      "loss": 1.0768,
+      "step": 17760
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.093438152239402e-06,
+      "loss": 1.1384,
+      "step": 17761
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0927299689148752e-06,
+      "loss": 1.031,
+      "step": 17762
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0920220017435968e-06,
+      "loss": 1.1078,
+      "step": 17763
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0913142507427455e-06,
+      "loss": 1.0161,
+      "step": 17764
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0906067159294987e-06,
+      "loss": 1.0749,
+      "step": 17765
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0898993973210259e-06,
+      "loss": 1.1743,
+      "step": 17766
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0891922949344913e-06,
+      "loss": 1.0006,
+      "step": 17767
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0884854087870523e-06,
+      "loss": 1.1261,
+      "step": 17768
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.087778738895866e-06,
+      "loss": 1.0179,
+      "step": 17769
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.087072285278079e-06,
+      "loss": 1.1896,
+      "step": 17770
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0863660479508375e-06,
+      "loss": 1.1212,
+      "step": 17771
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0856600269312745e-06,
+      "loss": 1.1843,
+      "step": 17772
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0849542222365295e-06,
+      "loss": 1.1485,
+      "step": 17773
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0842486338837267e-06,
+      "loss": 0.9599,
+      "step": 17774
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0835432618899887e-06,
+      "loss": 1.1244,
+      "step": 17775
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0828381062724325e-06,
+      "loss": 1.0821,
+      "step": 17776
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0821331670481728e-06,
+      "loss": 1.0449,
+      "step": 17777
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0814284442343136e-06,
+      "loss": 0.9664,
+      "step": 17778
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0807239378479583e-06,
+      "loss": 1.0144,
+      "step": 17779
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0800196479062008e-06,
+      "loss": 1.053,
+      "step": 17780
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0793155744261352e-06,
+      "loss": 1.1274,
+      "step": 17781
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0786117174248457e-06,
+      "loss": 1.1236,
+      "step": 17782
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0779080769194127e-06,
+      "loss": 1.1926,
+      "step": 17783
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0772046529269109e-06,
+      "loss": 1.1238,
+      "step": 17784
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.076501445464413e-06,
+      "loss": 0.9749,
+      "step": 17785
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0757984545489807e-06,
+      "loss": 1.2284,
+      "step": 17786
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0750956801976753e-06,
+      "loss": 1.241,
+      "step": 17787
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0743931224275495e-06,
+      "loss": 0.9305,
+      "step": 17788
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0736907812556541e-06,
+      "loss": 1.1559,
+      "step": 17789
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0729886566990333e-06,
+      "loss": 1.0352,
+      "step": 17790
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0722867487747235e-06,
+      "loss": 1.1049,
+      "step": 17791
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0715850574997566e-06,
+      "loss": 1.1514,
+      "step": 17792
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0708835828911656e-06,
+      "loss": 1.2236,
+      "step": 17793
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0701823249659703e-06,
+      "loss": 1.001,
+      "step": 17794
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.069481283741186e-06,
+      "loss": 1.1022,
+      "step": 17795
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0687804592338292e-06,
+      "loss": 1.195,
+      "step": 17796
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0680798514609048e-06,
+      "loss": 1.0894,
+      "step": 17797
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.067379460439415e-06,
+      "loss": 1.0962,
+      "step": 17798
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.066679286186354e-06,
+      "loss": 1.0387,
+      "step": 17799
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.065979328718716e-06,
+      "loss": 1.1027,
+      "step": 17800
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0652795880534862e-06,
+      "loss": 1.0783,
+      "step": 17801
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0645800642076453e-06,
+      "loss": 1.0673,
+      "step": 17802
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0638807571981647e-06,
+      "loss": 1.1482,
+      "step": 17803
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0631816670420214e-06,
+      "loss": 0.9905,
+      "step": 17804
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0624827937561755e-06,
+      "loss": 1.2872,
+      "step": 17805
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0617841373575888e-06,
+      "loss": 1.1238,
+      "step": 17806
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0610856978632123e-06,
+      "loss": 1.0725,
+      "step": 17807
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.06038747529e-06,
+      "loss": 1.1282,
+      "step": 17808
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.059689469654893e-06,
+      "loss": 1.0365,
+      "step": 17809
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0589916809748301e-06,
+      "loss": 1.1355,
+      "step": 17810
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0582941092667421e-06,
+      "loss": 1.1065,
+      "step": 17811
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0575967545475618e-06,
+      "loss": 1.0649,
+      "step": 17812
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0568996168342094e-06,
+      "loss": 0.9717,
+      "step": 17813
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0562026961436023e-06,
+      "loss": 1.1688,
+      "step": 17814
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0555059924926513e-06,
+      "loss": 1.1671,
+      "step": 17815
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0548095058982678e-06,
+      "loss": 0.8902,
+      "step": 17816
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.05411323637735e-06,
+      "loss": 1.1231,
+      "step": 17817
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0534171839467956e-06,
+      "loss": 1.1189,
+      "step": 17818
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.052721348623492e-06,
+      "loss": 1.001,
+      "step": 17819
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.052025730424332e-06,
+      "loss": 0.9998,
+      "step": 17820
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0513303293661914e-06,
+      "loss": 1.0632,
+      "step": 17821
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.050635145465946e-06,
+      "loss": 1.0276,
+      "step": 17822
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0499401787404651e-06,
+      "loss": 1.0519,
+      "step": 17823
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0492454292066178e-06,
+      "loss": 1.1206,
+      "step": 17824
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0485508968812596e-06,
+      "loss": 1.1059,
+      "step": 17825
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.047856581781246e-06,
+      "loss": 0.9814,
+      "step": 17826
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.047162483923424e-06,
+      "loss": 1.0993,
+      "step": 17827
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0464686033246419e-06,
+      "loss": 1.2176,
+      "step": 17828
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0457749400017336e-06,
+      "loss": 1.021,
+      "step": 17829
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0450814939715358e-06,
+      "loss": 1.1333,
+      "step": 17830
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0443882652508719e-06,
+      "loss": 1.1583,
+      "step": 17831
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0436952538565692e-06,
+      "loss": 1.0068,
+      "step": 17832
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0430024598054412e-06,
+      "loss": 1.145,
+      "step": 17833
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.042309883114304e-06,
+      "loss": 0.9996,
+      "step": 17834
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0416175237999615e-06,
+      "loss": 0.912,
+      "step": 17835
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0409253818792165e-06,
+      "loss": 1.1077,
+      "step": 17836
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0402334573688654e-06,
+      "loss": 1.0384,
+      "step": 17837
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0395417502856985e-06,
+      "loss": 1.047,
+      "step": 17838
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.038850260646499e-06,
+      "loss": 1.0253,
+      "step": 17839
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0381589884680522e-06,
+      "loss": 1.2014,
+      "step": 17840
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0374679337671311e-06,
+      "loss": 1.1198,
+      "step": 17841
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0367770965605062e-06,
+      "loss": 1.0627,
+      "step": 17842
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0360864768649381e-06,
+      "loss": 1.1016,
+      "step": 17843
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0353960746971914e-06,
+      "loss": 1.075,
+      "step": 17844
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0347058900740182e-06,
+      "loss": 1.1944,
+      "step": 17845
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0340159230121672e-06,
+      "loss": 0.9908,
+      "step": 17846
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0333261735283795e-06,
+      "loss": 1.0884,
+      "step": 17847
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0326366416393973e-06,
+      "loss": 1.2036,
+      "step": 17848
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.031947327361953e-06,
+      "loss": 1.1041,
+      "step": 17849
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0312582307127717e-06,
+      "loss": 1.16,
+      "step": 17850
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0305693517085757e-06,
+      "loss": 1.0109,
+      "step": 17851
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0298806903660852e-06,
+      "loss": 1.1247,
+      "step": 17852
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.02919224670201e-06,
+      "loss": 1.0672,
+      "step": 17853
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0285040207330566e-06,
+      "loss": 1.0435,
+      "step": 17854
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0278160124759251e-06,
+      "loss": 1.0974,
+      "step": 17855
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0271282219473132e-06,
+      "loss": 1.1454,
+      "step": 17856
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0264406491639123e-06,
+      "loss": 1.2444,
+      "step": 17857
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0257532941424065e-06,
+      "loss": 1.0673,
+      "step": 17858
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0250661568994746e-06,
+      "loss": 1.0454,
+      "step": 17859
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0243792374517937e-06,
+      "loss": 1.1482,
+      "step": 17860
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0236925358160321e-06,
+      "loss": 1.1233,
+      "step": 17861
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0230060520088548e-06,
+      "loss": 1.0717,
+      "step": 17862
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0223197860469192e-06,
+      "loss": 1.0515,
+      "step": 17863
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.02163373794688e-06,
+      "loss": 1.038,
+      "step": 17864
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.020947907725387e-06,
+      "loss": 1.0895,
+      "step": 17865
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0202622953990816e-06,
+      "loss": 1.0873,
+      "step": 17866
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0195769009845992e-06,
+      "loss": 1.0409,
+      "step": 17867
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0188917244985775e-06,
+      "loss": 1.0569,
+      "step": 17868
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0182067659576412e-06,
+      "loss": 1.2274,
+      "step": 17869
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0175220253784123e-06,
+      "loss": 1.1858,
+      "step": 17870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0168375027775045e-06,
+      "loss": 1.1068,
+      "step": 17871
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0161531981715345e-06,
+      "loss": 1.2117,
+      "step": 17872
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0154691115771066e-06,
+      "loss": 1.1963,
+      "step": 17873
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0147852430108195e-06,
+      "loss": 1.1411,
+      "step": 17874
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0141015924892693e-06,
+      "loss": 1.1244,
+      "step": 17875
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0134181600290483e-06,
+      "loss": 1.1232,
+      "step": 17876
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0127349456467394e-06,
+      "loss": 1.0651,
+      "step": 17877
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0120519493589231e-06,
+      "loss": 1.1752,
+      "step": 17878
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0113691711821716e-06,
+      "loss": 1.1573,
+      "step": 17879
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0106866111330572e-06,
+      "loss": 1.1097,
+      "step": 17880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.010004269228143e-06,
+      "loss": 0.9624,
+      "step": 17881
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.009322145483984e-06,
+      "loss": 1.2418,
+      "step": 17882
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.008640239917138e-06,
+      "loss": 1.011,
+      "step": 17883
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0079585525441504e-06,
+      "loss": 0.9441,
+      "step": 17884
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.007277083381565e-06,
+      "loss": 1.1521,
+      "step": 17885
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0065958324459147e-06,
+      "loss": 1.1288,
+      "step": 17886
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0059147997537378e-06,
+      "loss": 1.1313,
+      "step": 17887
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0052339853215588e-06,
+      "loss": 1.0636,
+      "step": 17888
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0045533891658966e-06,
+      "loss": 1.2598,
+      "step": 17889
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.003873011303268e-06,
+      "loss": 1.3027,
+      "step": 17890
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0031928517501866e-06,
+      "loss": 1.0794,
+      "step": 17891
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0025129105231557e-06,
+      "loss": 1.0952,
+      "step": 17892
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0018331876386767e-06,
+      "loss": 1.0396,
+      "step": 17893
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0011536831132407e-06,
+      "loss": 0.9721,
+      "step": 17894
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0004743969633423e-06,
+      "loss": 1.1674,
+      "step": 17895
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.99795329205464e-07,
+      "loss": 0.9705,
+      "step": 17896
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.991164798560848e-07,
+      "loss": 1.0047,
+      "step": 17897
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.984378489316747e-07,
+      "loss": 1.1883,
+      "step": 17898
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.977594364487087e-07,
+      "loss": 1.0292,
+      "step": 17899
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.970812424236464e-07,
+      "loss": 1.0626,
+      "step": 17900
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.96403266872945e-07,
+      "loss": 1.038,
+      "step": 17901
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.957255098130559e-07,
+      "loss": 1.1294,
+      "step": 17902
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.950479712604299e-07,
+      "loss": 1.1088,
+      "step": 17903
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.943706512315078e-07,
+      "loss": 1.25,
+      "step": 17904
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.936935497427248e-07,
+      "loss": 1.2406,
+      "step": 17905
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.930166668105101e-07,
+      "loss": 1.2098,
+      "step": 17906
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.923400024512952e-07,
+      "loss": 1.1054,
+      "step": 17907
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.91663556681498e-07,
+      "loss": 1.1148,
+      "step": 17908
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.909873295175331e-07,
+      "loss": 1.1631,
+      "step": 17909
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.903113209758098e-07,
+      "loss": 1.2043,
+      "step": 17910
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.896355310727356e-07,
+      "loss": 1.1664,
+      "step": 17911
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.88959959824708e-07,
+      "loss": 1.1863,
+      "step": 17912
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.882846072481223e-07,
+      "loss": 1.1326,
+      "step": 17913
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.876094733593656e-07,
+      "loss": 0.8063,
+      "step": 17914
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.869345581748236e-07,
+      "loss": 1.2273,
+      "step": 17915
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.862598617108709e-07,
+      "loss": 1.0926,
+      "step": 17916
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.85585383983886e-07,
+      "loss": 1.2021,
+      "step": 17917
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.849111250102327e-07,
+      "loss": 1.0887,
+      "step": 17918
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.842370848062721e-07,
+      "loss": 1.0612,
+      "step": 17919
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.83563263388365e-07,
+      "loss": 1.2443,
+      "step": 17920
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.828896607728612e-07,
+      "loss": 1.1087,
+      "step": 17921
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.822162769761056e-07,
+      "loss": 1.2267,
+      "step": 17922
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.815431120144414e-07,
+      "loss": 1.1906,
+      "step": 17923
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.808701659042042e-07,
+      "loss": 0.8552,
+      "step": 17924
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.801974386617241e-07,
+      "loss": 0.9994,
+      "step": 17925
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.795249303033228e-07,
+      "loss": 1.0414,
+      "step": 17926
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.788526408453258e-07,
+      "loss": 1.0305,
+      "step": 17927
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.78180570304046e-07,
+      "loss": 1.1982,
+      "step": 17928
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.7750871869579e-07,
+      "loss": 1.0269,
+      "step": 17929
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.768370860368614e-07,
+      "loss": 1.136,
+      "step": 17930
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.76165672343563e-07,
+      "loss": 0.9985,
+      "step": 17931
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.754944776321862e-07,
+      "loss": 1.0546,
+      "step": 17932
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.748235019190165e-07,
+      "loss": 1.1127,
+      "step": 17933
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.741527452203382e-07,
+      "loss": 1.1707,
+      "step": 17934
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.734822075524298e-07,
+      "loss": 1.0146,
+      "step": 17935
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.72811888931562e-07,
+      "loss": 0.9314,
+      "step": 17936
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.721417893740026e-07,
+      "loss": 1.15,
+      "step": 17937
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.71471908896009e-07,
+      "loss": 1.1266,
+      "step": 17938
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.708022475138422e-07,
+      "loss": 1.0467,
+      "step": 17939
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.70132805243752e-07,
+      "loss": 1.1671,
+      "step": 17940
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.694635821019815e-07,
+      "loss": 0.9776,
+      "step": 17941
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.687945781047714e-07,
+      "loss": 1.0465,
+      "step": 17942
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.681257932683574e-07,
+      "loss": 0.9472,
+      "step": 17943
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.6745722760897e-07,
+      "loss": 1.0782,
+      "step": 17944
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.667888811428316e-07,
+      "loss": 0.947,
+      "step": 17945
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.661207538861584e-07,
+      "loss": 1.0623,
+      "step": 17946
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.654528458551694e-07,
+      "loss": 1.159,
+      "step": 17947
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.647851570660704e-07,
+      "loss": 1.0344,
+      "step": 17948
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.641176875350645e-07,
+      "loss": 1.0369,
+      "step": 17949
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.634504372783459e-07,
+      "loss": 1.1337,
+      "step": 17950
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.627834063121122e-07,
+      "loss": 1.0609,
+      "step": 17951
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.62116594652547e-07,
+      "loss": 1.0468,
+      "step": 17952
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.614500023158335e-07,
+      "loss": 1.1657,
+      "step": 17953
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.60783629318145e-07,
+      "loss": 1.179,
+      "step": 17954
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.601174756756558e-07,
+      "loss": 1.1489,
+      "step": 17955
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.594515414045302e-07,
+      "loss": 1.0622,
+      "step": 17956
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.587858265209283e-07,
+      "loss": 1.0758,
+      "step": 17957
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.581203310410037e-07,
+      "loss": 1.0701,
+      "step": 17958
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.574550549809092e-07,
+      "loss": 1.0595,
+      "step": 17959
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.567899983567875e-07,
+      "loss": 1.105,
+      "step": 17960
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.561251611847766e-07,
+      "loss": 0.8701,
+      "step": 17961
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.554605434810094e-07,
+      "loss": 1.0339,
+      "step": 17962
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.547961452616182e-07,
+      "loss": 0.894,
+      "step": 17963
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.541319665427228e-07,
+      "loss": 1.1804,
+      "step": 17964
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.534680073404423e-07,
+      "loss": 1.1022,
+      "step": 17965
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.528042676708871e-07,
+      "loss": 1.176,
+      "step": 17966
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.521407475501665e-07,
+      "loss": 1.0812,
+      "step": 17967
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.51477446994381e-07,
+      "loss": 1.2325,
+      "step": 17968
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.508143660196256e-07,
+      "loss": 1.1009,
+      "step": 17969
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.50151504641994e-07,
+      "loss": 1.087,
+      "step": 17970
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.494888628775711e-07,
+      "loss": 1.1937,
+      "step": 17971
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.488264407424374e-07,
+      "loss": 1.0814,
+      "step": 17972
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.481642382526656e-07,
+      "loss": 0.9347,
+      "step": 17973
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.475022554243285e-07,
+      "loss": 1.0345,
+      "step": 17974
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.468404922734897e-07,
+      "loss": 1.0993,
+      "step": 17975
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.461789488162076e-07,
+      "loss": 1.1464,
+      "step": 17976
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.455176250685338e-07,
+      "loss": 1.0381,
+      "step": 17977
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.44856521046521e-07,
+      "loss": 1.2138,
+      "step": 17978
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.441956367662108e-07,
+      "loss": 1.1761,
+      "step": 17979
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.435349722436404e-07,
+      "loss": 1.0977,
+      "step": 17980
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.428745274948392e-07,
+      "loss": 1.0988,
+      "step": 17981
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.422143025358399e-07,
+      "loss": 1.106,
+      "step": 17982
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.415542973826608e-07,
+      "loss": 1.1463,
+      "step": 17983
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.408945120513202e-07,
+      "loss": 0.9447,
+      "step": 17984
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.402349465578242e-07,
+      "loss": 1.1232,
+      "step": 17985
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.395756009181856e-07,
+      "loss": 1.1245,
+      "step": 17986
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.389164751484014e-07,
+      "loss": 0.8957,
+      "step": 17987
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.382575692644657e-07,
+      "loss": 1.083,
+      "step": 17988
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.37598883282368e-07,
+      "loss": 1.2615,
+      "step": 17989
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.369404172180952e-07,
+      "loss": 1.142,
+      "step": 17990
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.362821710876257e-07,
+      "loss": 1.1622,
+      "step": 17991
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.356241449069326e-07,
+      "loss": 1.0872,
+      "step": 17992
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.349663386919816e-07,
+      "loss": 1.0245,
+      "step": 17993
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.343087524587402e-07,
+      "loss": 0.9596,
+      "step": 17994
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.336513862231644e-07,
+      "loss": 0.953,
+      "step": 17995
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.329942400012059e-07,
+      "loss": 1.0321,
+      "step": 17996
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.323373138088098e-07,
+      "loss": 1.1409,
+      "step": 17997
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.316806076619234e-07,
+      "loss": 1.1253,
+      "step": 17998
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.310241215764781e-07,
+      "loss": 1.1221,
+      "step": 17999
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.303678555684059e-07,
+      "loss": 1.0289,
+      "step": 18000
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.29711809653634e-07,
+      "loss": 1.0739,
+      "step": 18001
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.290559838480806e-07,
+      "loss": 0.9588,
+      "step": 18002
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.284003781676632e-07,
+      "loss": 1.0376,
+      "step": 18003
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.277449926282899e-07,
+      "loss": 0.922,
+      "step": 18004
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.27089827245865e-07,
+      "loss": 1.0639,
+      "step": 18005
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.264348820362857e-07,
+      "loss": 1.1592,
+      "step": 18006
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.257801570154489e-07,
+      "loss": 1.155,
+      "step": 18007
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.251256521992424e-07,
+      "loss": 1.0844,
+      "step": 18008
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.244713676035455e-07,
+      "loss": 1.002,
+      "step": 18009
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.23817303244241e-07,
+      "loss": 1.0688,
+      "step": 18010
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.231634591371974e-07,
+      "loss": 0.8901,
+      "step": 18011
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.225098352982831e-07,
+      "loss": 1.0895,
+      "step": 18012
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.218564317433576e-07,
+      "loss": 1.0153,
+      "step": 18013
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.212032484882805e-07,
+      "loss": 0.9782,
+      "step": 18014
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.205502855489001e-07,
+      "loss": 1.0212,
+      "step": 18015
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.198975429410628e-07,
+      "loss": 1.1081,
+      "step": 18016
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.192450206806069e-07,
+      "loss": 1.1786,
+      "step": 18017
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.185927187833709e-07,
+      "loss": 1.1499,
+      "step": 18018
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.179406372651823e-07,
+      "loss": 0.989,
+      "step": 18019
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.172887761418637e-07,
+      "loss": 1.1938,
+      "step": 18020
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.166371354292348e-07,
+      "loss": 1.003,
+      "step": 18021
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.159857151431106e-07,
+      "loss": 1.1288,
+      "step": 18022
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.153345152992976e-07,
+      "loss": 1.0848,
+      "step": 18023
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.146835359135997e-07,
+      "loss": 1.1875,
+      "step": 18024
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.140327770018109e-07,
+      "loss": 1.1213,
+      "step": 18025
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.133822385797287e-07,
+      "loss": 1.0657,
+      "step": 18026
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.127319206631358e-07,
+      "loss": 1.1139,
+      "step": 18027
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.120818232678142e-07,
+      "loss": 1.1062,
+      "step": 18028
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.114319464095389e-07,
+      "loss": 1.173,
+      "step": 18029
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.107822901040841e-07,
+      "loss": 1.0594,
+      "step": 18030
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.101328543672117e-07,
+      "loss": 1.1016,
+      "step": 18031
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.094836392146833e-07,
+      "loss": 1.1181,
+      "step": 18032
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.088346446622498e-07,
+      "loss": 1.0697,
+      "step": 18033
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.081858707256663e-07,
+      "loss": 1.0259,
+      "step": 18034
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.075373174206737e-07,
+      "loss": 0.949,
+      "step": 18035
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.068889847630091e-07,
+      "loss": 1.2661,
+      "step": 18036
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.062408727684069e-07,
+      "loss": 1.1877,
+      "step": 18037
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.055929814525954e-07,
+      "loss": 1.1829,
+      "step": 18038
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.049453108312967e-07,
+      "loss": 1.0055,
+      "step": 18039
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.04297860920228e-07,
+      "loss": 1.0978,
+      "step": 18040
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.03650631735099e-07,
+      "loss": 1.1621,
+      "step": 18041
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.030036232916195e-07,
+      "loss": 0.8776,
+      "step": 18042
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.023568356054891e-07,
+      "loss": 1.1037,
+      "step": 18043
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.017102686924028e-07,
+      "loss": 1.0923,
+      "step": 18044
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.010639225680496e-07,
+      "loss": 1.1479,
+      "step": 18045
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.004177972481176e-07,
+      "loss": 1.061,
+      "step": 18046
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.997718927482846e-07,
+      "loss": 1.0196,
+      "step": 18047
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.991262090842245e-07,
+      "loss": 1.0625,
+      "step": 18048
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.984807462716061e-07,
+      "loss": 1.0604,
+      "step": 18049
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.978355043260945e-07,
+      "loss": 1.0277,
+      "step": 18050
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.97190483263346e-07,
+      "loss": 0.9986,
+      "step": 18051
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.965456830990127e-07,
+      "loss": 1.1511,
+      "step": 18052
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.959011038487453e-07,
+      "loss": 1.0718,
+      "step": 18053
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.952567455281847e-07,
+      "loss": 1.1306,
+      "step": 18054
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.94612608152966e-07,
+      "loss": 1.0535,
+      "step": 18055
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.939686917387191e-07,
+      "loss": 0.9859,
+      "step": 18056
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.933249963010748e-07,
+      "loss": 1.0536,
+      "step": 18057
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.926815218556506e-07,
+      "loss": 1.0905,
+      "step": 18058
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.920382684180629e-07,
+      "loss": 1.0875,
+      "step": 18059
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.91395236003918e-07,
+      "loss": 1.0589,
+      "step": 18060
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.907524246288257e-07,
+      "loss": 1.0415,
+      "step": 18061
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.901098343083825e-07,
+      "loss": 0.8309,
+      "step": 18062
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.894674650581825e-07,
+      "loss": 1.1747,
+      "step": 18063
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.888253168938121e-07,
+      "loss": 1.089,
+      "step": 18064
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.88183389830859e-07,
+      "loss": 1.2393,
+      "step": 18065
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.875416838848971e-07,
+      "loss": 1.2058,
+      "step": 18066
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.869001990714999e-07,
+      "loss": 1.0889,
+      "step": 18067
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.862589354062323e-07,
+      "loss": 1.0738,
+      "step": 18068
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.856178929046599e-07,
+      "loss": 1.1669,
+      "step": 18069
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.849770715823369e-07,
+      "loss": 1.0988,
+      "step": 18070
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.843364714548141e-07,
+      "loss": 1.0876,
+      "step": 18071
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.836960925376348e-07,
+      "loss": 1.1087,
+      "step": 18072
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.83055934846343e-07,
+      "loss": 1.169,
+      "step": 18073
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.82415998396472e-07,
+      "loss": 1.1092,
+      "step": 18074
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.817762832035504e-07,
+      "loss": 1.1067,
+      "step": 18075
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.811367892831013e-07,
+      "loss": 1.0057,
+      "step": 18076
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.804975166506458e-07,
+      "loss": 1.0739,
+      "step": 18077
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.798584653216968e-07,
+      "loss": 0.9158,
+      "step": 18078
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.79219635311761e-07,
+      "loss": 1.2386,
+      "step": 18079
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.785810266363404e-07,
+      "loss": 1.2533,
+      "step": 18080
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.779426393109348e-07,
+      "loss": 0.9598,
+      "step": 18081
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.773044733510338e-07,
+      "loss": 1.1758,
+      "step": 18082
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.766665287721232e-07,
+      "loss": 1.0336,
+      "step": 18083
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.76028805589686e-07,
+      "loss": 1.2469,
+      "step": 18084
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.753913038191964e-07,
+      "loss": 1.0233,
+      "step": 18085
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.747540234761265e-07,
+      "loss": 1.04,
+      "step": 18086
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.741169645759396e-07,
+      "loss": 1.0086,
+      "step": 18087
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.734801271340964e-07,
+      "loss": 1.0727,
+      "step": 18088
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.728435111660482e-07,
+      "loss": 1.1152,
+      "step": 18089
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.722071166872481e-07,
+      "loss": 1.0595,
+      "step": 18090
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.71570943713137e-07,
+      "loss": 1.0444,
+      "step": 18091
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.709349922591526e-07,
+      "loss": 1.1526,
+      "step": 18092
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.702992623407292e-07,
+      "loss": 1.1631,
+      "step": 18093
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.696637539732933e-07,
+      "loss": 1.0378,
+      "step": 18094
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.690284671722671e-07,
+      "loss": 0.9531,
+      "step": 18095
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.683934019530649e-07,
+      "loss": 1.0487,
+      "step": 18096
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.67758558331101e-07,
+      "loss": 1.1304,
+      "step": 18097
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.671239363217798e-07,
+      "loss": 1.1045,
+      "step": 18098
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.664895359405023e-07,
+      "loss": 1.0128,
+      "step": 18099
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.658553572026618e-07,
+      "loss": 1.0935,
+      "step": 18100
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.652214001236503e-07,
+      "loss": 1.0441,
+      "step": 18101
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.645876647188511e-07,
+      "loss": 1.0843,
+      "step": 18102
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.639541510036431e-07,
+      "loss": 0.9515,
+      "step": 18103
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.633208589933984e-07,
+      "loss": 1.1575,
+      "step": 18104
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.626877887034879e-07,
+      "loss": 1.0653,
+      "step": 18105
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.620549401492728e-07,
+      "loss": 0.9421,
+      "step": 18106
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.614223133461109e-07,
+      "loss": 1.1678,
+      "step": 18107
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.607899083093507e-07,
+      "loss": 1.1354,
+      "step": 18108
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.601577250543458e-07,
+      "loss": 1.0769,
+      "step": 18109
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.595257635964316e-07,
+      "loss": 1.1683,
+      "step": 18110
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.58894023950948e-07,
+      "loss": 1.1378,
+      "step": 18111
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.582625061332206e-07,
+      "loss": 1.0712,
+      "step": 18112
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.576312101585782e-07,
+      "loss": 1.2196,
+      "step": 18113
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.570001360423408e-07,
+      "loss": 1.234,
+      "step": 18114
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.563692837998217e-07,
+      "loss": 1.0962,
+      "step": 18115
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.557386534463264e-07,
+      "loss": 0.9913,
+      "step": 18116
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.551082449971648e-07,
+      "loss": 1.0633,
+      "step": 18117
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.544780584676315e-07,
+      "loss": 0.8351,
+      "step": 18118
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.538480938730198e-07,
+      "loss": 1.2889,
+      "step": 18119
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.532183512286152e-07,
+      "loss": 1.1774,
+      "step": 18120
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.525888305497032e-07,
+      "loss": 1.2909,
+      "step": 18121
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.519595318515583e-07,
+      "loss": 1.0589,
+      "step": 18122
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.513304551494528e-07,
+      "loss": 1.263,
+      "step": 18123
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.507016004586499e-07,
+      "loss": 1.0241,
+      "step": 18124
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.50072967794413e-07,
+      "loss": 1.0788,
+      "step": 18125
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.494445571719978e-07,
+      "loss": 1.1379,
+      "step": 18126
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.488163686066508e-07,
+      "loss": 1.0702,
+      "step": 18127
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.481884021136178e-07,
+      "loss": 1.1313,
+      "step": 18128
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.475606577081386e-07,
+      "loss": 1.0837,
+      "step": 18129
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.469331354054456e-07,
+      "loss": 1.0997,
+      "step": 18130
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.463058352207676e-07,
+      "loss": 1.2532,
+      "step": 18131
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.456787571693248e-07,
+      "loss": 1.0471,
+      "step": 18132
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.450519012663383e-07,
+      "loss": 0.8938,
+      "step": 18133
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.444252675270192e-07,
+      "loss": 1.2223,
+      "step": 18134
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.437988559665733e-07,
+      "loss": 1.0646,
+      "step": 18135
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.431726666001994e-07,
+      "loss": 1.1496,
+      "step": 18136
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.425466994430975e-07,
+      "loss": 1.212,
+      "step": 18137
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.419209545104556e-07,
+      "loss": 1.1549,
+      "step": 18138
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.412954318174582e-07,
+      "loss": 1.1537,
+      "step": 18139
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.406701313792875e-07,
+      "loss": 1.1304,
+      "step": 18140
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.400450532111159e-07,
+      "loss": 1.0843,
+      "step": 18141
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.394201973281125e-07,
+      "loss": 1.0912,
+      "step": 18142
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.387955637454392e-07,
+      "loss": 1.1602,
+      "step": 18143
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.381711524782566e-07,
+      "loss": 1.085,
+      "step": 18144
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.375469635417166e-07,
+      "loss": 1.1089,
+      "step": 18145
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.369229969509651e-07,
+      "loss": 1.0798,
+      "step": 18146
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.362992527211445e-07,
+      "loss": 1.028,
+      "step": 18147
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.356757308673924e-07,
+      "loss": 1.1444,
+      "step": 18148
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.350524314048392e-07,
+      "loss": 1.1191,
+      "step": 18149
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.344293543486104e-07,
+      "loss": 0.9442,
+      "step": 18150
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.33806499713824e-07,
+      "loss": 1.0076,
+      "step": 18151
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.331838675156001e-07,
+      "loss": 1.2587,
+      "step": 18152
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.325614577690433e-07,
+      "loss": 1.0918,
+      "step": 18153
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.319392704892604e-07,
+      "loss": 0.9974,
+      "step": 18154
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.313173056913471e-07,
+      "loss": 1.0475,
+      "step": 18155
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.306955633904012e-07,
+      "loss": 1.0805,
+      "step": 18156
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.300740436015076e-07,
+      "loss": 1.1741,
+      "step": 18157
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.294527463397484e-07,
+      "loss": 1.0816,
+      "step": 18158
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.288316716202005e-07,
+      "loss": 1.0455,
+      "step": 18159
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.282108194579386e-07,
+      "loss": 1.1908,
+      "step": 18160
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.27590189868026e-07,
+      "loss": 1.1139,
+      "step": 18161
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.269697828655244e-07,
+      "loss": 1.1106,
+      "step": 18162
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.263495984654879e-07,
+      "loss": 1.1263,
+      "step": 18163
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.257296366829693e-07,
+      "loss": 1.144,
+      "step": 18164
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.25109897533013e-07,
+      "loss": 1.0521,
+      "step": 18165
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.244903810306538e-07,
+      "loss": 0.9907,
+      "step": 18166
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.238710871909317e-07,
+      "loss": 1.1951,
+      "step": 18167
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.232520160288704e-07,
+      "loss": 0.9319,
+      "step": 18168
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.226331675594957e-07,
+      "loss": 1.147,
+      "step": 18169
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.220145417978254e-07,
+      "loss": 1.0999,
+      "step": 18170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.213961387588698e-07,
+      "loss": 0.8559,
+      "step": 18171
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.207779584576358e-07,
+      "loss": 1.0528,
+      "step": 18172
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.201600009091259e-07,
+      "loss": 1.0143,
+      "step": 18173
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.19542266128337e-07,
+      "loss": 0.997,
+      "step": 18174
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.189247541302592e-07,
+      "loss": 1.057,
+      "step": 18175
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.183074649298739e-07,
+      "loss": 1.1775,
+      "step": 18176
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.17690398542167e-07,
+      "loss": 1.0375,
+      "step": 18177
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.170735549821085e-07,
+      "loss": 1.1692,
+      "step": 18178
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.164569342646677e-07,
+      "loss": 1.1427,
+      "step": 18179
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.158405364048116e-07,
+      "loss": 1.0774,
+      "step": 18180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.15224361417496e-07,
+      "loss": 1.1918,
+      "step": 18181
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.146084093176732e-07,
+      "loss": 0.9736,
+      "step": 18182
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.139926801202891e-07,
+      "loss": 1.3311,
+      "step": 18183
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.133771738402896e-07,
+      "loss": 1.162,
+      "step": 18184
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.127618904926094e-07,
+      "loss": 0.9997,
+      "step": 18185
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.121468300921786e-07,
+      "loss": 1.0327,
+      "step": 18186
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.115319926539222e-07,
+      "loss": 0.9875,
+      "step": 18187
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.109173781927649e-07,
+      "loss": 1.1111,
+      "step": 18188
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.103029867236178e-07,
+      "loss": 1.127,
+      "step": 18189
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.096888182613916e-07,
+      "loss": 1.2381,
+      "step": 18190
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.090748728209885e-07,
+      "loss": 1.2648,
+      "step": 18191
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.084611504173101e-07,
+      "loss": 1.0031,
+      "step": 18192
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.078476510652478e-07,
+      "loss": 1.0945,
+      "step": 18193
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.072343747796907e-07,
+      "loss": 1.119,
+      "step": 18194
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.066213215755181e-07,
+      "loss": 1.04,
+      "step": 18195
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.060084914676114e-07,
+      "loss": 1.2146,
+      "step": 18196
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.053958844708398e-07,
+      "loss": 1.182,
+      "step": 18197
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.047835006000692e-07,
+      "loss": 1.061,
+      "step": 18198
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.041713398701601e-07,
+      "loss": 1.0923,
+      "step": 18199
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.035594022959692e-07,
+      "loss": 1.0464,
+      "step": 18200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.02947687892347e-07,
+      "loss": 1.1281,
+      "step": 18201
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.023361966741361e-07,
+      "loss": 1.1883,
+      "step": 18202
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.017249286561757e-07,
+      "loss": 1.138,
+      "step": 18203
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.011138838533006e-07,
+      "loss": 1.1864,
+      "step": 18204
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.005030622803389e-07,
+      "loss": 1.0066,
+      "step": 18205
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.998924639521132e-07,
+      "loss": 1.0298,
+      "step": 18206
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.992820888834396e-07,
+      "loss": 1.0302,
+      "step": 18207
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.986719370891327e-07,
+      "loss": 1.1825,
+      "step": 18208
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.980620085839985e-07,
+      "loss": 1.174,
+      "step": 18209
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.974523033828363e-07,
+      "loss": 1.0317,
+      "step": 18210
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.96842821500442e-07,
+      "loss": 1.0205,
+      "step": 18211
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.962335629516083e-07,
+      "loss": 1.0759,
+      "step": 18212
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.956245277511187e-07,
+      "loss": 1.2134,
+      "step": 18213
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.950157159137528e-07,
+      "loss": 0.9463,
+      "step": 18214
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.94407127454283e-07,
+      "loss": 0.9952,
+      "step": 18215
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.93798762387481e-07,
+      "loss": 1.1588,
+      "step": 18216
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.931906207281082e-07,
+      "loss": 1.0827,
+      "step": 18217
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.925827024909239e-07,
+      "loss": 1.0289,
+      "step": 18218
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.919750076906762e-07,
+      "loss": 1.0352,
+      "step": 18219
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.91367536342118e-07,
+      "loss": 1.1093,
+      "step": 18220
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.907602884599874e-07,
+      "loss": 1.1078,
+      "step": 18221
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.901532640590226e-07,
+      "loss": 1.1152,
+      "step": 18222
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.895464631539495e-07,
+      "loss": 1.2194,
+      "step": 18223
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.889398857594998e-07,
+      "loss": 1.0797,
+      "step": 18224
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.883335318903895e-07,
+      "loss": 1.0494,
+      "step": 18225
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.877274015613323e-07,
+      "loss": 1.0892,
+      "step": 18226
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.871214947870398e-07,
+      "loss": 1.0573,
+      "step": 18227
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.865158115822158e-07,
+      "loss": 1.133,
+      "step": 18228
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.859103519615563e-07,
+      "loss": 1.0345,
+      "step": 18229
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.85305115939754e-07,
+      "loss": 1.1438,
+      "step": 18230
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.847001035314983e-07,
+      "loss": 1.1198,
+      "step": 18231
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.840953147514696e-07,
+      "loss": 1.0667,
+      "step": 18232
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.834907496143452e-07,
+      "loss": 1.0543,
+      "step": 18233
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.828864081347931e-07,
+      "loss": 1.0698,
+      "step": 18234
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.822822903274829e-07,
+      "loss": 1.1759,
+      "step": 18235
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.81678396207074e-07,
+      "loss": 1.0711,
+      "step": 18236
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.810747257882201e-07,
+      "loss": 1.1719,
+      "step": 18237
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.804712790855684e-07,
+      "loss": 1.2793,
+      "step": 18238
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.798680561137672e-07,
+      "loss": 0.8659,
+      "step": 18239
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.792650568874538e-07,
+      "loss": 1.1408,
+      "step": 18240
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.786622814212597e-07,
+      "loss": 1.1985,
+      "step": 18241
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.780597297298109e-07,
+      "loss": 1.1305,
+      "step": 18242
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.774574018277336e-07,
+      "loss": 1.124,
+      "step": 18243
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.768552977296418e-07,
+      "loss": 1.081,
+      "step": 18244
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.762534174501491e-07,
+      "loss": 0.9155,
+      "step": 18245
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.756517610038571e-07,
+      "loss": 1.1463,
+      "step": 18246
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.750503284053712e-07,
+      "loss": 1.1416,
+      "step": 18247
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.744491196692839e-07,
+      "loss": 1.1502,
+      "step": 18248
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.738481348101857e-07,
+      "loss": 1.0996,
+      "step": 18249
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.732473738426582e-07,
+      "loss": 1.1439,
+      "step": 18250
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.726468367812823e-07,
+      "loss": 1.1209,
+      "step": 18251
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.720465236406316e-07,
+      "loss": 1.0004,
+      "step": 18252
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.714464344352746e-07,
+      "loss": 1.0068,
+      "step": 18253
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.708465691797718e-07,
+      "loss": 1.1344,
+      "step": 18254
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.702469278886793e-07,
+      "loss": 1.1717,
+      "step": 18255
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.696475105765522e-07,
+      "loss": 0.9941,
+      "step": 18256
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.690483172579333e-07,
+      "loss": 1.1456,
+      "step": 18257
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.684493479473654e-07,
+      "loss": 1.12,
+      "step": 18258
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.678506026593813e-07,
+      "loss": 1.2482,
+      "step": 18259
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.672520814085138e-07,
+      "loss": 1.082,
+      "step": 18260
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.666537842092858e-07,
+      "loss": 0.9567,
+      "step": 18261
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.660557110762157e-07,
+      "loss": 1.3635,
+      "step": 18262
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.654578620238162e-07,
+      "loss": 1.1628,
+      "step": 18263
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.648602370665981e-07,
+      "loss": 0.8544,
+      "step": 18264
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.642628362190618e-07,
+      "loss": 1.0273,
+      "step": 18265
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.636656594957048e-07,
+      "loss": 1.2847,
+      "step": 18266
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.630687069110199e-07,
+      "loss": 1.0279,
+      "step": 18267
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.624719784794932e-07,
+      "loss": 0.9407,
+      "step": 18268
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.618754742156042e-07,
+      "loss": 0.9679,
+      "step": 18269
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.612791941338282e-07,
+      "loss": 1.0589,
+      "step": 18270
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.606831382486368e-07,
+      "loss": 1.1576,
+      "step": 18271
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.600873065744952e-07,
+      "loss": 1.1025,
+      "step": 18272
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.594916991258605e-07,
+      "loss": 1.118,
+      "step": 18273
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.588963159171847e-07,
+      "loss": 1.0644,
+      "step": 18274
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.583011569629206e-07,
+      "loss": 1.229,
+      "step": 18275
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.577062222775078e-07,
+      "loss": 0.9522,
+      "step": 18276
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.571115118753847e-07,
+      "loss": 1.0887,
+      "step": 18277
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.565170257709808e-07,
+      "loss": 1.1711,
+      "step": 18278
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.55922763978727e-07,
+      "loss": 1.1925,
+      "step": 18279
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.553287265130415e-07,
+      "loss": 0.9963,
+      "step": 18280
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.547349133883409e-07,
+      "loss": 1.1157,
+      "step": 18281
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.541413246190321e-07,
+      "loss": 1.0781,
+      "step": 18282
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.535479602195239e-07,
+      "loss": 1.173,
+      "step": 18283
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.529548202042137e-07,
+      "loss": 1.1111,
+      "step": 18284
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.523619045874964e-07,
+      "loss": 1.0528,
+      "step": 18285
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.517692133837562e-07,
+      "loss": 1.0111,
+      "step": 18286
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.511767466073827e-07,
+      "loss": 1.1131,
+      "step": 18287
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.505845042727478e-07,
+      "loss": 1.0213,
+      "step": 18288
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.499924863942265e-07,
+      "loss": 1.1187,
+      "step": 18289
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.49400692986183e-07,
+      "loss": 1.1224,
+      "step": 18290
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.488091240629803e-07,
+      "loss": 1.0727,
+      "step": 18291
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.482177796389744e-07,
+      "loss": 1.1296,
+      "step": 18292
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.476266597285131e-07,
+      "loss": 1.0667,
+      "step": 18293
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.470357643459425e-07,
+      "loss": 1.1078,
+      "step": 18294
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.464450935056023e-07,
+      "loss": 1.2465,
+      "step": 18295
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.458546472218264e-07,
+      "loss": 1.1734,
+      "step": 18296
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.452644255089425e-07,
+      "loss": 1.113,
+      "step": 18297
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.446744283812723e-07,
+      "loss": 1.0488,
+      "step": 18298
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.440846558531367e-07,
+      "loss": 1.1924,
+      "step": 18299
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.434951079388463e-07,
+      "loss": 1.1051,
+      "step": 18300
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.429057846527066e-07,
+      "loss": 1.059,
+      "step": 18301
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.423166860090169e-07,
+      "loss": 1.0799,
+      "step": 18302
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.417278120220784e-07,
+      "loss": 1.0811,
+      "step": 18303
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.411391627061781e-07,
+      "loss": 1.0574,
+      "step": 18304
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.405507380756005e-07,
+      "loss": 1.1011,
+      "step": 18305
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.399625381446241e-07,
+      "loss": 0.9891,
+      "step": 18306
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.393745629275251e-07,
+      "loss": 1.1127,
+      "step": 18307
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.38786812438571e-07,
+      "loss": 1.1435,
+      "step": 18308
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.38199286692024e-07,
+      "loss": 1.1488,
+      "step": 18309
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.376119857021435e-07,
+      "loss": 1.0915,
+      "step": 18310
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.370249094831806e-07,
+      "loss": 0.9624,
+      "step": 18311
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.364380580493813e-07,
+      "loss": 0.9103,
+      "step": 18312
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.358514314149856e-07,
+      "loss": 1.2294,
+      "step": 18313
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.352650295942321e-07,
+      "loss": 1.0267,
+      "step": 18314
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.346788526013504e-07,
+      "loss": 1.144,
+      "step": 18315
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.340929004505637e-07,
+      "loss": 1.0247,
+      "step": 18316
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.335071731560917e-07,
+      "loss": 1.0342,
+      "step": 18317
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.329216707321496e-07,
+      "loss": 1.0226,
+      "step": 18318
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.323363931929461e-07,
+      "loss": 1.1013,
+      "step": 18319
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.317513405526821e-07,
+      "loss": 1.0494,
+      "step": 18320
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.311665128255552e-07,
+      "loss": 1.0892,
+      "step": 18321
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.305819100257594e-07,
+      "loss": 1.1788,
+      "step": 18322
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.299975321674812e-07,
+      "loss": 1.1312,
+      "step": 18323
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.294133792648994e-07,
+      "loss": 1.0784,
+      "step": 18324
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.288294513321902e-07,
+      "loss": 1.0864,
+      "step": 18325
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.28245748383527e-07,
+      "loss": 1.1012,
+      "step": 18326
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.276622704330716e-07,
+      "loss": 1.2141,
+      "step": 18327
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.270790174949838e-07,
+      "loss": 1.1485,
+      "step": 18328
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.264959895834157e-07,
+      "loss": 1.1466,
+      "step": 18329
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.259131867125202e-07,
+      "loss": 1.061,
+      "step": 18330
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.253306088964363e-07,
+      "loss": 0.901,
+      "step": 18331
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.247482561493036e-07,
+      "loss": 1.0547,
+      "step": 18332
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.241661284852508e-07,
+      "loss": 1.1517,
+      "step": 18333
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.235842259184079e-07,
+      "loss": 1.1877,
+      "step": 18334
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.230025484628955e-07,
+      "loss": 1.1906,
+      "step": 18335
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.224210961328281e-07,
+      "loss": 1.0899,
+      "step": 18336
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.218398689423178e-07,
+      "loss": 1.0596,
+      "step": 18337
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.212588669054655e-07,
+      "loss": 1.1688,
+      "step": 18338
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.206780900363741e-07,
+      "loss": 1.0664,
+      "step": 18339
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.20097538349136e-07,
+      "loss": 1.0119,
+      "step": 18340
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.195172118578397e-07,
+      "loss": 1.0473,
+      "step": 18341
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.189371105765664e-07,
+      "loss": 1.034,
+      "step": 18342
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.183572345193957e-07,
+      "loss": 1.0702,
+      "step": 18343
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.177775837003987e-07,
+      "loss": 1.2099,
+      "step": 18344
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.171981581336429e-07,
+      "loss": 0.9011,
+      "step": 18345
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.16618957833185e-07,
+      "loss": 1.1393,
+      "step": 18346
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.16039982813086e-07,
+      "loss": 1.0672,
+      "step": 18347
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.154612330873933e-07,
+      "loss": 1.1435,
+      "step": 18348
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.148827086701493e-07,
+      "loss": 1.1604,
+      "step": 18349
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.143044095753982e-07,
+      "loss": 0.9975,
+      "step": 18350
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.137263358171698e-07,
+      "loss": 1.0762,
+      "step": 18351
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.131484874094941e-07,
+      "loss": 1.1249,
+      "step": 18352
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.125708643663909e-07,
+      "loss": 1.0906,
+      "step": 18353
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.119934667018802e-07,
+      "loss": 0.9267,
+      "step": 18354
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.11416294429974e-07,
+      "loss": 1.0656,
+      "step": 18355
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.108393475646779e-07,
+      "loss": 0.9043,
+      "step": 18356
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.102626261199907e-07,
+      "loss": 0.9975,
+      "step": 18357
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.096861301099101e-07,
+      "loss": 1.0672,
+      "step": 18358
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.091098595484258e-07,
+      "loss": 0.9237,
+      "step": 18359
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.085338144495224e-07,
+      "loss": 1.0425,
+      "step": 18360
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.079579948271764e-07,
+      "loss": 1.0808,
+      "step": 18361
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.073824006953633e-07,
+      "loss": 0.8484,
+      "step": 18362
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.06807032068052e-07,
+      "loss": 1.0947,
+      "step": 18363
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.062318889592035e-07,
+      "loss": 1.0528,
+      "step": 18364
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.056569713827732e-07,
+      "loss": 1.2068,
+      "step": 18365
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.050822793527157e-07,
+      "loss": 1.0404,
+      "step": 18366
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.045078128829763e-07,
+      "loss": 1.0645,
+      "step": 18367
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.039335719874951e-07,
+      "loss": 1.1666,
+      "step": 18368
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.033595566802065e-07,
+      "loss": 1.046,
+      "step": 18369
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.027857669750415e-07,
+      "loss": 1.0177,
+      "step": 18370
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.022122028859246e-07,
+      "loss": 1.1731,
+      "step": 18371
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.016388644267736e-07,
+      "loss": 1.1829,
+      "step": 18372
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.010657516115005e-07,
+      "loss": 1.147,
+      "step": 18373
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.004928644540154e-07,
+      "loss": 1.0943,
+      "step": 18374
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.999202029682195e-07,
+      "loss": 1.1409,
+      "step": 18375
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.993477671680104e-07,
+      "loss": 1.129,
+      "step": 18376
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.987755570672772e-07,
+      "loss": 1.0575,
+      "step": 18377
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.982035726799086e-07,
+      "loss": 1.2845,
+      "step": 18378
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.976318140197835e-07,
+      "loss": 1.0382,
+      "step": 18379
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.970602811007765e-07,
+      "loss": 1.133,
+      "step": 18380
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.964889739367575e-07,
+      "loss": 1.1361,
+      "step": 18381
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.95917892541591e-07,
+      "loss": 0.9479,
+      "step": 18382
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.953470369291349e-07,
+      "loss": 1.1252,
+      "step": 18383
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.947764071132423e-07,
+      "loss": 1.0315,
+      "step": 18384
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.942060031077591e-07,
+      "loss": 1.0328,
+      "step": 18385
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.936358249265307e-07,
+      "loss": 1.0592,
+      "step": 18386
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.930658725833917e-07,
+      "loss": 1.2614,
+      "step": 18387
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.924961460921731e-07,
+      "loss": 0.9536,
+      "step": 18388
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.919266454666984e-07,
+      "loss": 0.8632,
+      "step": 18389
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.913573707207921e-07,
+      "loss": 1.1686,
+      "step": 18390
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.907883218682676e-07,
+      "loss": 0.9744,
+      "step": 18391
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.902194989229316e-07,
+      "loss": 1.0578,
+      "step": 18392
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.896509018985886e-07,
+      "loss": 1.1409,
+      "step": 18393
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.890825308090388e-07,
+      "loss": 1.0293,
+      "step": 18394
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.885143856680731e-07,
+      "loss": 1.0868,
+      "step": 18395
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.879464664894786e-07,
+      "loss": 1.0534,
+      "step": 18396
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.873787732870386e-07,
+      "loss": 1.1567,
+      "step": 18397
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.868113060745285e-07,
+      "loss": 1.0431,
+      "step": 18398
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.862440648657187e-07,
+      "loss": 0.9942,
+      "step": 18399
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.856770496743737e-07,
+      "loss": 0.9656,
+      "step": 18400
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.851102605142545e-07,
+      "loss": 1.1635,
+      "step": 18401
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.84543697399116e-07,
+      "loss": 1.106,
+      "step": 18402
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.839773603427069e-07,
+      "loss": 1.0535,
+      "step": 18403
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.834112493587675e-07,
+      "loss": 1.1182,
+      "step": 18404
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.828453644610388e-07,
+      "loss": 0.9842,
+      "step": 18405
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.822797056632536e-07,
+      "loss": 1.1149,
+      "step": 18406
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.81714272979137e-07,
+      "loss": 1.0958,
+      "step": 18407
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.811490664224096e-07,
+      "loss": 1.1293,
+      "step": 18408
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.805840860067903e-07,
+      "loss": 1.0085,
+      "step": 18409
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.800193317459869e-07,
+      "loss": 1.1482,
+      "step": 18410
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.794548036537063e-07,
+      "loss": 1.234,
+      "step": 18411
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.788905017436442e-07,
+      "loss": 1.1617,
+      "step": 18412
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.783264260294997e-07,
+      "loss": 1.1591,
+      "step": 18413
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.777625765249585e-07,
+      "loss": 1.1914,
+      "step": 18414
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.771989532437039e-07,
+      "loss": 1.0316,
+      "step": 18415
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.76635556199412e-07,
+      "loss": 1.1434,
+      "step": 18416
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.760723854057561e-07,
+      "loss": 0.891,
+      "step": 18417
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.755094408764029e-07,
+      "loss": 1.1344,
+      "step": 18418
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.749467226250151e-07,
+      "loss": 1.128,
+      "step": 18419
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.743842306652459e-07,
+      "loss": 1.0476,
+      "step": 18420
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.738219650107436e-07,
+      "loss": 1.1879,
+      "step": 18421
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.732599256751571e-07,
+      "loss": 1.0818,
+      "step": 18422
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.726981126721222e-07,
+      "loss": 0.8412,
+      "step": 18423
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.721365260152746e-07,
+      "loss": 1.1007,
+      "step": 18424
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.71575165718239e-07,
+      "loss": 0.983,
+      "step": 18425
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.710140317946424e-07,
+      "loss": 0.9149,
+      "step": 18426
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.704531242580992e-07,
+      "loss": 1.2679,
+      "step": 18427
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.69892443122222e-07,
+      "loss": 1.0339,
+      "step": 18428
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.693319884006144e-07,
+      "loss": 1.2131,
+      "step": 18429
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.687717601068799e-07,
+      "loss": 1.0789,
+      "step": 18430
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.682117582546132e-07,
+      "loss": 0.9952,
+      "step": 18431
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.676519828574036e-07,
+      "loss": 0.9431,
+      "step": 18432
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.670924339288332e-07,
+      "loss": 1.1903,
+      "step": 18433
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.665331114824836e-07,
+      "loss": 1.0735,
+      "step": 18434
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.659740155319272e-07,
+      "loss": 1.0735,
+      "step": 18435
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.654151460907288e-07,
+      "loss": 0.9845,
+      "step": 18436
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.648565031724552e-07,
+      "loss": 0.9219,
+      "step": 18437
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.642980867906601e-07,
+      "loss": 1.1335,
+      "step": 18438
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.637398969588949e-07,
+      "loss": 1.2032,
+      "step": 18439
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.631819336907042e-07,
+      "loss": 1.0901,
+      "step": 18440
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.626241969996305e-07,
+      "loss": 0.9318,
+      "step": 18441
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.620666868992076e-07,
+      "loss": 0.9342,
+      "step": 18442
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.615094034029645e-07,
+      "loss": 1.066,
+      "step": 18443
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.609523465244239e-07,
+      "loss": 1.0809,
+      "step": 18444
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.603955162771047e-07,
+      "loss": 1.0581,
+      "step": 18445
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.598389126745209e-07,
+      "loss": 1.1009,
+      "step": 18446
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.592825357301769e-07,
+      "loss": 1.2006,
+      "step": 18447
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.587263854575754e-07,
+      "loss": 1.1384,
+      "step": 18448
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.581704618702134e-07,
+      "loss": 1.0961,
+      "step": 18449
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.576147649815812e-07,
+      "loss": 1.0795,
+      "step": 18450
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.570592948051635e-07,
+      "loss": 1.1437,
+      "step": 18451
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.565040513544396e-07,
+      "loss": 1.0019,
+      "step": 18452
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.559490346428843e-07,
+      "loss": 1.1363,
+      "step": 18453
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.553942446839667e-07,
+      "loss": 1.1313,
+      "step": 18454
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.548396814911495e-07,
+      "loss": 1.0033,
+      "step": 18455
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.542853450778885e-07,
+      "loss": 1.1323,
+      "step": 18456
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.537312354576387e-07,
+      "loss": 1.0784,
+      "step": 18457
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.531773526438456e-07,
+      "loss": 1.0561,
+      "step": 18458
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.5262369664995e-07,
+      "loss": 1.0438,
+      "step": 18459
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.520702674893864e-07,
+      "loss": 1.0292,
+      "step": 18460
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.515170651755876e-07,
+      "loss": 1.15,
+      "step": 18461
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.50964089721976e-07,
+      "loss": 1.1099,
+      "step": 18462
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.504113411419722e-07,
+      "loss": 1.0875,
+      "step": 18463
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.498588194489874e-07,
+      "loss": 1.1079,
+      "step": 18464
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.493065246564334e-07,
+      "loss": 1.1244,
+      "step": 18465
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.487544567777093e-07,
+      "loss": 1.1767,
+      "step": 18466
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.482026158262144e-07,
+      "loss": 0.9179,
+      "step": 18467
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.476510018153371e-07,
+      "loss": 1.1047,
+      "step": 18468
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.470996147584684e-07,
+      "loss": 1.0733,
+      "step": 18469
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.465484546689849e-07,
+      "loss": 1.0119,
+      "step": 18470
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.459975215602632e-07,
+      "loss": 1.166,
+      "step": 18471
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.454468154456717e-07,
+      "loss": 1.1687,
+      "step": 18472
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.448963363385763e-07,
+      "loss": 1.1198,
+      "step": 18473
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.443460842523341e-07,
+      "loss": 0.9426,
+      "step": 18474
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.43796059200299e-07,
+      "loss": 1.1067,
+      "step": 18475
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.432462611958168e-07,
+      "loss": 1.1582,
+      "step": 18476
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.426966902522313e-07,
+      "loss": 1.0158,
+      "step": 18477
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.421473463828787e-07,
+      "loss": 1.1978,
+      "step": 18478
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.415982296010892e-07,
+      "loss": 1.1333,
+      "step": 18479
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.410493399201878e-07,
+      "loss": 1.0076,
+      "step": 18480
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.40500677353496e-07,
+      "loss": 1.1428,
+      "step": 18481
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.399522419143278e-07,
+      "loss": 1.2384,
+      "step": 18482
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.394040336159901e-07,
+      "loss": 1.1639,
+      "step": 18483
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.388560524717891e-07,
+      "loss": 1.009,
+      "step": 18484
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.383082984950217e-07,
+      "loss": 1.1045,
+      "step": 18485
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.377607716989809e-07,
+      "loss": 1.1866,
+      "step": 18486
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.372134720969514e-07,
+      "loss": 0.9794,
+      "step": 18487
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.366663997022171e-07,
+      "loss": 1.0998,
+      "step": 18488
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.361195545280518e-07,
+      "loss": 1.1741,
+      "step": 18489
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.355729365877283e-07,
+      "loss": 1.1932,
+      "step": 18490
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.35026545894507e-07,
+      "loss": 1.2283,
+      "step": 18491
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.344803824616519e-07,
+      "loss": 1.0626,
+      "step": 18492
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.339344463024155e-07,
+      "loss": 1.1129,
+      "step": 18493
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.333887374300451e-07,
+      "loss": 1.1412,
+      "step": 18494
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.328432558577824e-07,
+      "loss": 0.9884,
+      "step": 18495
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.322980015988667e-07,
+      "loss": 1.0289,
+      "step": 18496
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.317529746665308e-07,
+      "loss": 1.2239,
+      "step": 18497
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.312081750739973e-07,
+      "loss": 1.1806,
+      "step": 18498
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.306636028344881e-07,
+      "loss": 1.1464,
+      "step": 18499
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.301192579612192e-07,
+      "loss": 1.2045,
+      "step": 18500
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.295751404674022e-07,
+      "loss": 1.1351,
+      "step": 18501
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.290312503662377e-07,
+      "loss": 1.001,
+      "step": 18502
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.284875876709274e-07,
+      "loss": 1.0279,
+      "step": 18503
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.279441523946595e-07,
+      "loss": 1.113,
+      "step": 18504
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.274009445506268e-07,
+      "loss": 1.1314,
+      "step": 18505
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.268579641520101e-07,
+      "loss": 1.0104,
+      "step": 18506
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.263152112119841e-07,
+      "loss": 1.026,
+      "step": 18507
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.257726857437208e-07,
+      "loss": 1.1853,
+      "step": 18508
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.25230387760386e-07,
+      "loss": 1.0397,
+      "step": 18509
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.246883172751405e-07,
+      "loss": 1.2818,
+      "step": 18510
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.241464743011372e-07,
+      "loss": 0.9087,
+      "step": 18511
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.236048588515242e-07,
+      "loss": 1.1191,
+      "step": 18512
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.230634709394479e-07,
+      "loss": 1.1417,
+      "step": 18513
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.225223105780442e-07,
+      "loss": 1.1042,
+      "step": 18514
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.219813777804451e-07,
+      "loss": 1.1018,
+      "step": 18515
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.214406725597776e-07,
+      "loss": 1.1059,
+      "step": 18516
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.209001949291659e-07,
+      "loss": 0.9725,
+      "step": 18517
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.203599449017228e-07,
+      "loss": 1.1138,
+      "step": 18518
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.198199224905587e-07,
+      "loss": 1.1301,
+      "step": 18519
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.192801277087779e-07,
+      "loss": 1.1726,
+      "step": 18520
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.18740560569483e-07,
+      "loss": 1.1992,
+      "step": 18521
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.182012210857647e-07,
+      "loss": 1.0896,
+      "step": 18522
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.176621092707103e-07,
+      "loss": 1.123,
+      "step": 18523
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.17123225137406e-07,
+      "loss": 1.0833,
+      "step": 18524
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.165845686989269e-07,
+      "loss": 1.0052,
+      "step": 18525
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.160461399683437e-07,
+      "loss": 1.0628,
+      "step": 18526
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.155079389587227e-07,
+      "loss": 1.1411,
+      "step": 18527
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.149699656831265e-07,
+      "loss": 1.1038,
+      "step": 18528
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.144322201546093e-07,
+      "loss": 0.9774,
+      "step": 18529
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.138947023862208e-07,
+      "loss": 0.9954,
+      "step": 18530
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.133574123910013e-07,
+      "loss": 1.145,
+      "step": 18531
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.12820350181994e-07,
+      "loss": 1.167,
+      "step": 18532
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.122835157722307e-07,
+      "loss": 1.1483,
+      "step": 18533
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.117469091747386e-07,
+      "loss": 0.9675,
+      "step": 18534
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.112105304025373e-07,
+      "loss": 1.0537,
+      "step": 18535
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.106743794686454e-07,
+      "loss": 1.2069,
+      "step": 18536
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.101384563860746e-07,
+      "loss": 1.2257,
+      "step": 18537
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.09602761167829e-07,
+      "loss": 1.1279,
+      "step": 18538
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.09067293826906e-07,
+      "loss": 1.1931,
+      "step": 18539
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.08532054376303e-07,
+      "loss": 1.0711,
+      "step": 18540
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.079970428290083e-07,
+      "loss": 1.0251,
+      "step": 18541
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.074622591980051e-07,
+      "loss": 1.0011,
+      "step": 18542
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.069277034962684e-07,
+      "loss": 1.1281,
+      "step": 18543
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.063933757367734e-07,
+      "loss": 1.1414,
+      "step": 18544
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.058592759324866e-07,
+      "loss": 1.148,
+      "step": 18545
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.053254040963675e-07,
+      "loss": 0.9304,
+      "step": 18546
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.0479176024137e-07,
+      "loss": 1.1455,
+      "step": 18547
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.042583443804484e-07,
+      "loss": 1.1466,
+      "step": 18548
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.037251565265445e-07,
+      "loss": 1.0586,
+      "step": 18549
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.031921966925979e-07,
+      "loss": 1.0113,
+      "step": 18550
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.026594648915396e-07,
+      "loss": 1.2295,
+      "step": 18551
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.021269611363023e-07,
+      "loss": 1.0733,
+      "step": 18552
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.015946854398047e-07,
+      "loss": 1.2058,
+      "step": 18553
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.010626378149642e-07,
+      "loss": 0.9609,
+      "step": 18554
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.005308182746906e-07,
+      "loss": 1.0772,
+      "step": 18555
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.999992268318933e-07,
+      "loss": 1.1392,
+      "step": 18556
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.994678634994688e-07,
+      "loss": 1.1401,
+      "step": 18557
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.989367282903146e-07,
+      "loss": 1.073,
+      "step": 18558
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.984058212173172e-07,
+      "loss": 1.038,
+      "step": 18559
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.978751422933626e-07,
+      "loss": 0.9057,
+      "step": 18560
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.973446915313275e-07,
+      "loss": 1.1478,
+      "step": 18561
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.968144689440847e-07,
+      "loss": 1.0151,
+      "step": 18562
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.962844745444996e-07,
+      "loss": 1.1349,
+      "step": 18563
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.957547083454362e-07,
+      "loss": 1.1874,
+      "step": 18564
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.952251703597489e-07,
+      "loss": 1.1887,
+      "step": 18565
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.946958606002883e-07,
+      "loss": 1.0303,
+      "step": 18566
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.941667790798977e-07,
+      "loss": 1.04,
+      "step": 18567
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.9363792581142e-07,
+      "loss": 1.3878,
+      "step": 18568
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.93109300807686e-07,
+      "loss": 1.2172,
+      "step": 18569
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.925809040815233e-07,
+      "loss": 1.0324,
+      "step": 18570
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.920527356457573e-07,
+      "loss": 1.0392,
+      "step": 18571
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.915247955132042e-07,
+      "loss": 1.0703,
+      "step": 18572
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.909970836966738e-07,
+      "loss": 0.8963,
+      "step": 18573
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.904696002089727e-07,
+      "loss": 1.2502,
+      "step": 18574
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.899423450629027e-07,
+      "loss": 0.9884,
+      "step": 18575
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.894153182712592e-07,
+      "loss": 0.9788,
+      "step": 18576
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.888885198468297e-07,
+      "loss": 0.8852,
+      "step": 18577
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.883619498023974e-07,
+      "loss": 0.9862,
+      "step": 18578
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.878356081507431e-07,
+      "loss": 1.1291,
+      "step": 18579
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.873094949046387e-07,
+      "loss": 1.1884,
+      "step": 18580
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.86783610076851e-07,
+      "loss": 1.0711,
+      "step": 18581
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.862579536801405e-07,
+      "loss": 1.0642,
+      "step": 18582
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.85732525727264e-07,
+      "loss": 1.1579,
+      "step": 18583
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.852073262309754e-07,
+      "loss": 1.1936,
+      "step": 18584
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.84682355204016e-07,
+      "loss": 1.0163,
+      "step": 18585
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.841576126591264e-07,
+      "loss": 1.0351,
+      "step": 18586
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.8363309860904e-07,
+      "loss": 1.1204,
+      "step": 18587
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.831088130664863e-07,
+      "loss": 1.0351,
+      "step": 18588
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.825847560441889e-07,
+      "loss": 1.1144,
+      "step": 18589
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.820609275548627e-07,
+      "loss": 0.9382,
+      "step": 18590
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.81537327611219e-07,
+      "loss": 1.0709,
+      "step": 18591
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.810139562259676e-07,
+      "loss": 1.1214,
+      "step": 18592
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.80490813411807e-07,
+      "loss": 1.2335,
+      "step": 18593
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.799678991814328e-07,
+      "loss": 1.0933,
+      "step": 18594
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.794452135475326e-07,
+      "loss": 0.9867,
+      "step": 18595
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.78922756522794e-07,
+      "loss": 1.1847,
+      "step": 18596
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.784005281198946e-07,
+      "loss": 1.172,
+      "step": 18597
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.778785283515054e-07,
+      "loss": 1.2332,
+      "step": 18598
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.773567572302929e-07,
+      "loss": 1.0362,
+      "step": 18599
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.768352147689227e-07,
+      "loss": 1.1289,
+      "step": 18600
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.763139009800501e-07,
+      "loss": 1.0331,
+      "step": 18601
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.757928158763249e-07,
+      "loss": 1.1219,
+      "step": 18602
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.752719594703904e-07,
+      "loss": 1.0472,
+      "step": 18603
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.747513317748898e-07,
+      "loss": 1.1484,
+      "step": 18604
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.742309328024565e-07,
+      "loss": 1.1248,
+      "step": 18605
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.737107625657157e-07,
+      "loss": 0.9458,
+      "step": 18606
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.731908210772952e-07,
+      "loss": 1.1115,
+      "step": 18607
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.726711083498093e-07,
+      "loss": 0.9051,
+      "step": 18608
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.721516243958725e-07,
+      "loss": 1.1274,
+      "step": 18609
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.716323692280868e-07,
+      "loss": 1.099,
+      "step": 18610
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.711133428590565e-07,
+      "loss": 0.9896,
+      "step": 18611
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.705945453013773e-07,
+      "loss": 1.186,
+      "step": 18612
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.700759765676378e-07,
+      "loss": 1.1846,
+      "step": 18613
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.695576366704191e-07,
+      "loss": 1.0276,
+      "step": 18614
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.690395256223047e-07,
+      "loss": 1.2365,
+      "step": 18615
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.685216434358664e-07,
+      "loss": 1.1187,
+      "step": 18616
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.680039901236701e-07,
+      "loss": 1.0745,
+      "step": 18617
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.674865656982764e-07,
+      "loss": 1.1624,
+      "step": 18618
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.669693701722456e-07,
+      "loss": 1.0015,
+      "step": 18619
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.664524035581276e-07,
+      "loss": 1.1632,
+      "step": 18620
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.659356658684656e-07,
+      "loss": 1.1668,
+      "step": 18621
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.654191571157996e-07,
+      "loss": 1.1538,
+      "step": 18622
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.649028773126652e-07,
+      "loss": 1.1806,
+      "step": 18623
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.643868264715912e-07,
+      "loss": 1.1872,
+      "step": 18624
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.638710046050988e-07,
+      "loss": 1.1952,
+      "step": 18625
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.633554117257057e-07,
+      "loss": 1.0302,
+      "step": 18626
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.628400478459261e-07,
+      "loss": 1.0578,
+      "step": 18627
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.623249129782649e-07,
+      "loss": 1.3199,
+      "step": 18628
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.618100071352228e-07,
+      "loss": 1.1538,
+      "step": 18629
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.612953303292934e-07,
+      "loss": 1.0591,
+      "step": 18630
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.607808825729699e-07,
+      "loss": 1.1111,
+      "step": 18631
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.602666638787346e-07,
+      "loss": 1.2545,
+      "step": 18632
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.597526742590653e-07,
+      "loss": 1.1633,
+      "step": 18633
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.592389137264354e-07,
+      "loss": 1.0664,
+      "step": 18634
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.587253822933136e-07,
+      "loss": 1.0302,
+      "step": 18635
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.582120799721613e-07,
+      "loss": 1.1139,
+      "step": 18636
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.57699006775434e-07,
+      "loss": 1.0548,
+      "step": 18637
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.571861627155828e-07,
+      "loss": 1.0511,
+      "step": 18638
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.566735478050533e-07,
+      "loss": 0.9906,
+      "step": 18639
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.561611620562857e-07,
+      "loss": 1.1899,
+      "step": 18640
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.556490054817132e-07,
+      "loss": 1.1047,
+      "step": 18641
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.551370780937627e-07,
+      "loss": 0.942,
+      "step": 18642
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.546253799048606e-07,
+      "loss": 1.2074,
+      "step": 18643
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.541139109274229e-07,
+      "loss": 0.8373,
+      "step": 18644
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.536026711738606e-07,
+      "loss": 1.0027,
+      "step": 18645
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.530916606565795e-07,
+      "loss": 1.1064,
+      "step": 18646
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.525808793879839e-07,
+      "loss": 1.0901,
+      "step": 18647
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.520703273804651e-07,
+      "loss": 0.9734,
+      "step": 18648
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.515600046464143e-07,
+      "loss": 1.0632,
+      "step": 18649
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.510499111982148e-07,
+      "loss": 0.9859,
+      "step": 18650
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.505400470482458e-07,
+      "loss": 1.2085,
+      "step": 18651
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.500304122088807e-07,
+      "loss": 1.1093,
+      "step": 18652
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.49521006692485e-07,
+      "loss": 1.1246,
+      "step": 18653
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.490118305114223e-07,
+      "loss": 1.1867,
+      "step": 18654
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.485028836780482e-07,
+      "loss": 1.13,
+      "step": 18655
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.479941662047139e-07,
+      "loss": 0.9813,
+      "step": 18656
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.474856781037618e-07,
+      "loss": 1.0209,
+      "step": 18657
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.469774193875354e-07,
+      "loss": 1.1645,
+      "step": 18658
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.464693900683659e-07,
+      "loss": 1.0501,
+      "step": 18659
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.459615901585836e-07,
+      "loss": 1.1634,
+      "step": 18660
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.454540196705083e-07,
+      "loss": 1.1107,
+      "step": 18661
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.449466786164603e-07,
+      "loss": 1.0695,
+      "step": 18662
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.444395670087499e-07,
+      "loss": 1.0935,
+      "step": 18663
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.439326848596838e-07,
+      "loss": 1.1247,
+      "step": 18664
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.43426032181561e-07,
+      "loss": 1.0624,
+      "step": 18665
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.429196089866762e-07,
+      "loss": 1.0989,
+      "step": 18666
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.424134152873228e-07,
+      "loss": 1.1408,
+      "step": 18667
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.419074510957823e-07,
+      "loss": 1.2182,
+      "step": 18668
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.414017164243324e-07,
+      "loss": 1.078,
+      "step": 18669
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.408962112852445e-07,
+      "loss": 1.1804,
+      "step": 18670
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.403909356907888e-07,
+      "loss": 1.1757,
+      "step": 18671
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.398858896532255e-07,
+      "loss": 0.9953,
+      "step": 18672
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.393810731848103e-07,
+      "loss": 1.0376,
+      "step": 18673
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.388764862977924e-07,
+      "loss": 1.0154,
+      "step": 18674
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.383721290044197e-07,
+      "loss": 0.9882,
+      "step": 18675
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.378680013169291e-07,
+      "loss": 1.009,
+      "step": 18676
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.373641032475552e-07,
+      "loss": 1.142,
+      "step": 18677
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.368604348085249e-07,
+      "loss": 1.0184,
+      "step": 18678
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.363569960120629e-07,
+      "loss": 1.0773,
+      "step": 18679
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.35853786870385e-07,
+      "loss": 0.9666,
+      "step": 18680
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.353508073957037e-07,
+      "loss": 1.1213,
+      "step": 18681
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.348480576002213e-07,
+      "loss": 1.1858,
+      "step": 18682
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.343455374961415e-07,
+      "loss": 0.9961,
+      "step": 18683
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.33843247095659e-07,
+      "loss": 1.1189,
+      "step": 18684
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.333411864109617e-07,
+      "loss": 1.0525,
+      "step": 18685
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.328393554542311e-07,
+      "loss": 1.0088,
+      "step": 18686
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.323377542376495e-07,
+      "loss": 1.1766,
+      "step": 18687
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.318363827733874e-07,
+      "loss": 1.1548,
+      "step": 18688
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.313352410736117e-07,
+      "loss": 1.1263,
+      "step": 18689
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.308343291504814e-07,
+      "loss": 1.1777,
+      "step": 18690
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.303336470161558e-07,
+      "loss": 1.0688,
+      "step": 18691
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.298331946827828e-07,
+      "loss": 1.023,
+      "step": 18692
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.293329721625063e-07,
+      "loss": 0.8866,
+      "step": 18693
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.288329794674673e-07,
+      "loss": 0.981,
+      "step": 18694
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.283332166097987e-07,
+      "loss": 1.1838,
+      "step": 18695
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.278336836016273e-07,
+      "loss": 1.1835,
+      "step": 18696
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.273343804550757e-07,
+      "loss": 1.1,
+      "step": 18697
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.268353071822607e-07,
+      "loss": 1.1027,
+      "step": 18698
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.263364637952939e-07,
+      "loss": 1.2073,
+      "step": 18699
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.258378503062788e-07,
+      "loss": 0.8978,
+      "step": 18700
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.253394667273171e-07,
+      "loss": 0.9199,
+      "step": 18701
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.248413130705021e-07,
+      "loss": 1.2734,
+      "step": 18702
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.243433893479243e-07,
+      "loss": 1.2475,
+      "step": 18703
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.238456955716653e-07,
+      "loss": 1.0508,
+      "step": 18704
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.233482317538007e-07,
+      "loss": 1.0051,
+      "step": 18705
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.228509979064067e-07,
+      "loss": 1.1367,
+      "step": 18706
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.223539940415479e-07,
+      "loss": 1.1478,
+      "step": 18707
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.218572201712846e-07,
+      "loss": 1.088,
+      "step": 18708
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.213606763076706e-07,
+      "loss": 1.0974,
+      "step": 18709
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.208643624627596e-07,
+      "loss": 1.07,
+      "step": 18710
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.203682786485931e-07,
+      "loss": 1.2084,
+      "step": 18711
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.198724248772091e-07,
+      "loss": 1.1336,
+      "step": 18712
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.193768011606404e-07,
+      "loss": 1.2511,
+      "step": 18713
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.188814075109172e-07,
+      "loss": 1.2163,
+      "step": 18714
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.183862439400589e-07,
+      "loss": 1.0193,
+      "step": 18715
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.178913104600814e-07,
+      "loss": 0.9826,
+      "step": 18716
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.17396607082995e-07,
+      "loss": 0.9475,
+      "step": 18717
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.16902133820808e-07,
+      "loss": 1.1178,
+      "step": 18718
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.164078906855163e-07,
+      "loss": 1.2258,
+      "step": 18719
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.159138776891148e-07,
+      "loss": 1.3124,
+      "step": 18720
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.154200948435906e-07,
+      "loss": 1.2503,
+      "step": 18721
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.149265421609284e-07,
+      "loss": 1.1148,
+      "step": 18722
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.144332196531054e-07,
+      "loss": 1.006,
+      "step": 18723
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.139401273320921e-07,
+      "loss": 1.165,
+      "step": 18724
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.134472652098521e-07,
+      "loss": 0.9363,
+      "step": 18725
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.129546332983504e-07,
+      "loss": 0.9589,
+      "step": 18726
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.124622316095384e-07,
+      "loss": 1.0709,
+      "step": 18727
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.119700601553668e-07,
+      "loss": 0.972,
+      "step": 18728
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.114781189477768e-07,
+      "loss": 0.9491,
+      "step": 18729
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.109864079987104e-07,
+      "loss": 0.8736,
+      "step": 18730
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.104949273200965e-07,
+      "loss": 1.1954,
+      "step": 18731
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.100036769238637e-07,
+      "loss": 1.1601,
+      "step": 18732
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.09512656821931e-07,
+      "loss": 1.1293,
+      "step": 18733
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.090218670262171e-07,
+      "loss": 1.1524,
+      "step": 18734
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.085313075486309e-07,
+      "loss": 0.9905,
+      "step": 18735
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.080409784010765e-07,
+      "loss": 1.15,
+      "step": 18736
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.07550879595451e-07,
+      "loss": 0.9919,
+      "step": 18737
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.070610111436502e-07,
+      "loss": 1.1961,
+      "step": 18738
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.065713730575617e-07,
+      "loss": 1.1093,
+      "step": 18739
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.060819653490645e-07,
+      "loss": 1.1445,
+      "step": 18740
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.055927880300394e-07,
+      "loss": 1.099,
+      "step": 18741
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.051038411123554e-07,
+      "loss": 1.1192,
+      "step": 18742
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.046151246078779e-07,
+      "loss": 1.0927,
+      "step": 18743
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.041266385284648e-07,
+      "loss": 1.0132,
+      "step": 18744
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.036383828859726e-07,
+      "loss": 1.0518,
+      "step": 18745
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.031503576922491e-07,
+      "loss": 1.0692,
+      "step": 18746
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.026625629591386e-07,
+      "loss": 1.2063,
+      "step": 18747
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.021749986984748e-07,
+      "loss": 1.1432,
+      "step": 18748
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.016876649220925e-07,
+      "loss": 0.9938,
+      "step": 18749
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.012005616418158e-07,
+      "loss": 1.0835,
+      "step": 18750
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.007136888694685e-07,
+      "loss": 1.0772,
+      "step": 18751
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.002270466168634e-07,
+      "loss": 0.9228,
+      "step": 18752
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.997406348958078e-07,
+      "loss": 1.0305,
+      "step": 18753
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.992544537181099e-07,
+      "loss": 1.131,
+      "step": 18754
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.987685030955658e-07,
+      "loss": 1.1039,
+      "step": 18755
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.982827830399684e-07,
+      "loss": 1.2002,
+      "step": 18756
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.977972935631026e-07,
+      "loss": 0.9873,
+      "step": 18757
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.973120346767523e-07,
+      "loss": 0.9967,
+      "step": 18758
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.968270063926928e-07,
+      "loss": 1.1691,
+      "step": 18759
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.963422087226944e-07,
+      "loss": 1.0946,
+      "step": 18760
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.958576416785199e-07,
+      "loss": 0.9877,
+      "step": 18761
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.953733052719311e-07,
+      "loss": 1.146,
+      "step": 18762
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.948891995146799e-07,
+      "loss": 1.1026,
+      "step": 18763
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.944053244185143e-07,
+      "loss": 1.2461,
+      "step": 18764
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.939216799951751e-07,
+      "loss": 1.0632,
+      "step": 18765
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.934382662564019e-07,
+      "loss": 1.0939,
+      "step": 18766
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.929550832139241e-07,
+      "loss": 0.8643,
+      "step": 18767
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.924721308794666e-07,
+      "loss": 1.0721,
+      "step": 18768
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.919894092647493e-07,
+      "loss": 1.0982,
+      "step": 18769
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.91506918381488e-07,
+      "loss": 1.1592,
+      "step": 18770
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.910246582413902e-07,
+      "loss": 1.2281,
+      "step": 18771
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.905426288561599e-07,
+      "loss": 0.9547,
+      "step": 18772
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.90060830237491e-07,
+      "loss": 0.8265,
+      "step": 18773
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.895792623970808e-07,
+      "loss": 0.9808,
+      "step": 18774
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.890979253466121e-07,
+      "loss": 1.1628,
+      "step": 18775
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.886168190977658e-07,
+      "loss": 0.9164,
+      "step": 18776
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.881359436622168e-07,
+      "loss": 1.1224,
+      "step": 18777
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.876552990516359e-07,
+      "loss": 1.1473,
+      "step": 18778
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.871748852776859e-07,
+      "loss": 0.9955,
+      "step": 18779
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.866947023520241e-07,
+      "loss": 1.1802,
+      "step": 18780
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.862147502863057e-07,
+      "loss": 1.0921,
+      "step": 18781
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.857350290921759e-07,
+      "loss": 1.0863,
+      "step": 18782
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.852555387812774e-07,
+      "loss": 1.1186,
+      "step": 18783
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.847762793652433e-07,
+      "loss": 0.8626,
+      "step": 18784
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.842972508557064e-07,
+      "loss": 1.0872,
+      "step": 18785
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.838184532642909e-07,
+      "loss": 1.1313,
+      "step": 18786
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.833398866026162e-07,
+      "loss": 0.966,
+      "step": 18787
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.82861550882292e-07,
+      "loss": 1.1068,
+      "step": 18788
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.823834461149313e-07,
+      "loss": 1.1967,
+      "step": 18789
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.819055723121346e-07,
+      "loss": 1.1644,
+      "step": 18790
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.814279294854973e-07,
+      "loss": 1.1066,
+      "step": 18791
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.8095051764661e-07,
+      "loss": 1.0914,
+      "step": 18792
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.804733368070602e-07,
+      "loss": 1.1028,
+      "step": 18793
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.799963869784274e-07,
+      "loss": 1.1231,
+      "step": 18794
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.795196681722835e-07,
+      "loss": 0.9126,
+      "step": 18795
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.790431804001982e-07,
+      "loss": 1.2374,
+      "step": 18796
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.785669236737356e-07,
+      "loss": 1.192,
+      "step": 18797
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.78090898004453e-07,
+      "loss": 1.1744,
+      "step": 18798
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.776151034039011e-07,
+      "loss": 1.1645,
+      "step": 18799
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.771395398836243e-07,
+      "loss": 1.0343,
+      "step": 18800
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.766642074551675e-07,
+      "loss": 1.146,
+      "step": 18801
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7618910613006277e-07,
+      "loss": 0.9998,
+      "step": 18802
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.757142359198397e-07,
+      "loss": 1.0001,
+      "step": 18803
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.752395968360213e-07,
+      "loss": 1.3,
+      "step": 18804
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7476518889012725e-07,
+      "loss": 1.2553,
+      "step": 18805
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7429101209366947e-07,
+      "loss": 0.9552,
+      "step": 18806
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.738170664581554e-07,
+      "loss": 1.0897,
+      "step": 18807
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.733433519950836e-07,
+      "loss": 1.1251,
+      "step": 18808
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.728698687159539e-07,
+      "loss": 1.0557,
+      "step": 18809
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.723966166322536e-07,
+      "loss": 1.1032,
+      "step": 18810
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7192359575546797e-07,
+      "loss": 1.1136,
+      "step": 18811
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.714508060970735e-07,
+      "loss": 1.1187,
+      "step": 18812
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.709782476685476e-07,
+      "loss": 1.2044,
+      "step": 18813
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7050592048135555e-07,
+      "loss": 1.0304,
+      "step": 18814
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7003382454696044e-07,
+      "loss": 1.1793,
+      "step": 18815
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.695619598768153e-07,
+      "loss": 1.1795,
+      "step": 18816
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.690903264823754e-07,
+      "loss": 1.0831,
+      "step": 18817
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.686189243750838e-07,
+      "loss": 1.1137,
+      "step": 18818
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.681477535663792e-07,
+      "loss": 1.1934,
+      "step": 18819
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.676768140676957e-07,
+      "loss": 1.1176,
+      "step": 18820
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.672061058904642e-07,
+      "loss": 1.0782,
+      "step": 18821
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.667356290461056e-07,
+      "loss": 1.2101,
+      "step": 18822
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6626538354603623e-07,
+      "loss": 1.0217,
+      "step": 18823
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6579536940166703e-07,
+      "loss": 0.9369,
+      "step": 18824
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.653255866244066e-07,
+      "loss": 1.164,
+      "step": 18825
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6485603522565483e-07,
+      "loss": 1.1431,
+      "step": 18826
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6438671521680246e-07,
+      "loss": 1.0455,
+      "step": 18827
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6391762660924264e-07,
+      "loss": 0.8719,
+      "step": 18828
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.634487694143586e-07,
+      "loss": 1.007,
+      "step": 18829
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.629801436435255e-07,
+      "loss": 1.1257,
+      "step": 18830
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.625117493081166e-07,
+      "loss": 1.1297,
+      "step": 18831
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6204358641949943e-07,
+      "loss": 1.167,
+      "step": 18832
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.615756549890338e-07,
+      "loss": 1.1503,
+      "step": 18833
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6110795502807615e-07,
+      "loss": 1.103,
+      "step": 18834
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.606404865479763e-07,
+      "loss": 1.1388,
+      "step": 18835
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6017324956007524e-07,
+      "loss": 1.2418,
+      "step": 18836
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.597062440757161e-07,
+      "loss": 1.1237,
+      "step": 18837
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.592394701062286e-07,
+      "loss": 1.1981,
+      "step": 18838
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5877292766294156e-07,
+      "loss": 1.0964,
+      "step": 18839
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.583066167571737e-07,
+      "loss": 0.9537,
+      "step": 18840
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.578405374002459e-07,
+      "loss": 1.0654,
+      "step": 18841
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5737468960346475e-07,
+      "loss": 0.9841,
+      "step": 18842
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5690907337813563e-07,
+      "loss": 1.0915,
+      "step": 18843
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5644368873555723e-07,
+      "loss": 1.0981,
+      "step": 18844
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.55978535687025e-07,
+      "loss": 1.2166,
+      "step": 18845
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5551361424382655e-07,
+      "loss": 1.0078,
+      "step": 18846
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.550489244172418e-07,
+      "loss": 1.125,
+      "step": 18847
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5458446621854947e-07,
+      "loss": 0.9783,
+      "step": 18848
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.541202396590205e-07,
+      "loss": 1.2173,
+      "step": 18849
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5365624474991933e-07,
+      "loss": 1.0377,
+      "step": 18850
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5319248150250685e-07,
+      "loss": 1.0223,
+      "step": 18851
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.527289499280352e-07,
+      "loss": 1.1905,
+      "step": 18852
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5226565003775645e-07,
+      "loss": 1.0574,
+      "step": 18853
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.518025818429106e-07,
+      "loss": 0.9897,
+      "step": 18854
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5133974535473633e-07,
+      "loss": 1.0902,
+      "step": 18855
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.508771405844636e-07,
+      "loss": 1.0266,
+      "step": 18856
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5041476754332124e-07,
+      "loss": 1.2771,
+      "step": 18857
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.49952626242528e-07,
+      "loss": 1.0811,
+      "step": 18858
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4949071669329825e-07,
+      "loss": 1.1373,
+      "step": 18859
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.490290389068408e-07,
+      "loss": 1.1009,
+      "step": 18860
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.485675928943611e-07,
+      "loss": 1.1722,
+      "step": 18861
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4810637866705697e-07,
+      "loss": 1.0571,
+      "step": 18862
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4764539623611937e-07,
+      "loss": 1.0416,
+      "step": 18863
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4718464561273487e-07,
+      "loss": 0.971,
+      "step": 18864
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4672412680808574e-07,
+      "loss": 1.1959,
+      "step": 18865
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.462638398333474e-07,
+      "loss": 1.2331,
+      "step": 18866
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4580378469968765e-07,
+      "loss": 1.1652,
+      "step": 18867
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.453439614182731e-07,
+      "loss": 1.1246,
+      "step": 18868
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4488437000026144e-07,
+      "loss": 1.1701,
+      "step": 18869
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4442501045680595e-07,
+      "loss": 1.1187,
+      "step": 18870
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.439658827990523e-07,
+      "loss": 1.0257,
+      "step": 18871
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4350698703814476e-07,
+      "loss": 1.1436,
+      "step": 18872
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4304832318521783e-07,
+      "loss": 1.223,
+      "step": 18873
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.425898912514015e-07,
+      "loss": 0.9074,
+      "step": 18874
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.421316912478202e-07,
+      "loss": 1.0831,
+      "step": 18875
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.416737231855961e-07,
+      "loss": 1.0978,
+      "step": 18876
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4121598707583923e-07,
+      "loss": 1.0575,
+      "step": 18877
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4075848292966073e-07,
+      "loss": 1.1273,
+      "step": 18878
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4030121075815837e-07,
+      "loss": 1.0545,
+      "step": 18879
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3984417057243433e-07,
+      "loss": 1.1383,
+      "step": 18880
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3938736238357535e-07,
+      "loss": 1.153,
+      "step": 18881
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.389307862026693e-07,
+      "loss": 1.2919,
+      "step": 18882
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.384744420407927e-07,
+      "loss": 0.9748,
+      "step": 18883
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.380183299090246e-07,
+      "loss": 1.0721,
+      "step": 18884
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3756244981843055e-07,
+      "loss": 1.0554,
+      "step": 18885
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.371068017800728e-07,
+      "loss": 1.1456,
+      "step": 18886
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.366513858050092e-07,
+      "loss": 1.2143,
+      "step": 18887
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3619620190429204e-07,
+      "loss": 1.2579,
+      "step": 18888
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3574125008896794e-07,
+      "loss": 1.1638,
+      "step": 18889
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.352865303700748e-07,
+      "loss": 0.9796,
+      "step": 18890
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3483204275864943e-07,
+      "loss": 0.9616,
+      "step": 18891
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3437778726571955e-07,
+      "loss": 1.1648,
+      "step": 18892
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3392376390231085e-07,
+      "loss": 1.2372,
+      "step": 18893
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3346997267943894e-07,
+      "loss": 0.9959,
+      "step": 18894
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.33016413608115e-07,
+      "loss": 0.9788,
+      "step": 18895
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.325630866993491e-07,
+      "loss": 1.0742,
+      "step": 18896
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3210999196414027e-07,
+      "loss": 1.1638,
+      "step": 18897
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3165712941348305e-07,
+      "loss": 1.0515,
+      "step": 18898
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3120449905836746e-07,
+      "loss": 1.1249,
+      "step": 18899
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3075210090977814e-07,
+      "loss": 1.0091,
+      "step": 18900
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3029993497869406e-07,
+      "loss": 1.0444,
+      "step": 18901
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2984800127608637e-07,
+      "loss": 1.1849,
+      "step": 18902
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2939629981292197e-07,
+      "loss": 1.0662,
+      "step": 18903
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.289448306001653e-07,
+      "loss": 1.0456,
+      "step": 18904
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2849359364876885e-07,
+      "loss": 1.0633,
+      "step": 18905
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.280425889696849e-07,
+      "loss": 1.1726,
+      "step": 18906
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.275918165738557e-07,
+      "loss": 1.0378,
+      "step": 18907
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.271412764722227e-07,
+      "loss": 1.0208,
+      "step": 18908
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2669096867571813e-07,
+      "loss": 1.0692,
+      "step": 18909
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2624089319526883e-07,
+      "loss": 1.0983,
+      "step": 18910
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2579105004179944e-07,
+      "loss": 0.9503,
+      "step": 18911
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.253414392262234e-07,
+      "loss": 1.1109,
+      "step": 18912
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.248920607594531e-07,
+      "loss": 1.0132,
+      "step": 18913
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2444291465239096e-07,
+      "loss": 1.1569,
+      "step": 18914
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2399400091594154e-07,
+      "loss": 1.1234,
+      "step": 18915
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2354531956099285e-07,
+      "loss": 0.9582,
+      "step": 18916
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.230968705984384e-07,
+      "loss": 1.1094,
+      "step": 18917
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.226486540391572e-07,
+      "loss": 1.0314,
+      "step": 18918
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.222006698940262e-07,
+      "loss": 1.084,
+      "step": 18919
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.217529181739188e-07,
+      "loss": 1.1646,
+      "step": 18920
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.213053988896998e-07,
+      "loss": 1.0663,
+      "step": 18921
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.208581120522293e-07,
+      "loss": 1.0503,
+      "step": 18922
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.204110576723597e-07,
+      "loss": 1.1687,
+      "step": 18923
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.199642357609412e-07,
+      "loss": 1.0624,
+      "step": 18924
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1951764632881864e-07,
+      "loss": 1.0128,
+      "step": 18925
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.190712893868265e-07,
+      "loss": 0.9437,
+      "step": 18926
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.186251649457962e-07,
+      "loss": 1.1571,
+      "step": 18927
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.181792730165568e-07,
+      "loss": 1.1707,
+      "step": 18928
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1773361360992857e-07,
+      "loss": 1.036,
+      "step": 18929
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.17288186736724e-07,
+      "loss": 1.2046,
+      "step": 18930
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1684299240775215e-07,
+      "loss": 1.0004,
+      "step": 18931
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1639803063381887e-07,
+      "loss": 1.1556,
+      "step": 18932
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.159533014257211e-07,
+      "loss": 0.8411,
+      "step": 18933
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1550880479425125e-07,
+      "loss": 0.7552,
+      "step": 18934
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.150645407501952e-07,
+      "loss": 1.1074,
+      "step": 18935
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1462050930433426e-07,
+      "loss": 1.1704,
+      "step": 18936
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1417671046744545e-07,
+      "loss": 0.9943,
+      "step": 18937
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1373314425029565e-07,
+      "loss": 1.1282,
+      "step": 18938
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.132898106636496e-07,
+      "loss": 1.1132,
+      "step": 18939
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1284670971826755e-07,
+      "loss": 1.0869,
+      "step": 18940
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1240384142490097e-07,
+      "loss": 1.2104,
+      "step": 18941
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1196120579429786e-07,
+      "loss": 1.0007,
+      "step": 18942
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1151880283719637e-07,
+      "loss": 1.0979,
+      "step": 18943
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.110766325643367e-07,
+      "loss": 1.0216,
+      "step": 18944
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1063469498644705e-07,
+      "loss": 1.0038,
+      "step": 18945
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.101929901142521e-07,
+      "loss": 1.1112,
+      "step": 18946
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.097515179584699e-07,
+      "loss": 1.0684,
+      "step": 18947
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.093102785298153e-07,
+      "loss": 1.2042,
+      "step": 18948
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0886927183899416e-07,
+      "loss": 1.1649,
+      "step": 18949
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.08428497896709e-07,
+      "loss": 1.1989,
+      "step": 18950
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0798795671365796e-07,
+      "loss": 1.0896,
+      "step": 18951
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0754764830052915e-07,
+      "loss": 0.9985,
+      "step": 18952
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.071075726680096e-07,
+      "loss": 1.0021,
+      "step": 18953
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0666772982677514e-07,
+      "loss": 1.1565,
+      "step": 18954
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0622811978750396e-07,
+      "loss": 1.1765,
+      "step": 18955
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.05788742560862e-07,
+      "loss": 1.1337,
+      "step": 18956
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0534959815751174e-07,
+      "loss": 1.2303,
+      "step": 18957
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0491068658810804e-07,
+      "loss": 1.042,
+      "step": 18958
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0447200786330577e-07,
+      "loss": 1.1863,
+      "step": 18959
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.040335619937474e-07,
+      "loss": 0.9554,
+      "step": 18960
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.035953489900746e-07,
+      "loss": 1.0366,
+      "step": 18961
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.031573688629198e-07,
+      "loss": 1.1243,
+      "step": 18962
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.027196216229123e-07,
+      "loss": 0.9866,
+      "step": 18963
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0228210728067595e-07,
+      "loss": 1.0432,
+      "step": 18964
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.018448258468266e-07,
+      "loss": 1.0888,
+      "step": 18965
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.014077773319758e-07,
+      "loss": 1.1839,
+      "step": 18966
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0097096174673055e-07,
+      "loss": 1.0427,
+      "step": 18967
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.005343791016902e-07,
+      "loss": 1.1168,
+      "step": 18968
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.000980294074497e-07,
+      "loss": 1.2381,
+      "step": 18969
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9966191267459596e-07,
+      "loss": 0.9531,
+      "step": 18970
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.992260289137151e-07,
+      "loss": 1.0427,
+      "step": 18971
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.987903781353841e-07,
+      "loss": 1.0613,
+      "step": 18972
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.983549603501746e-07,
+      "loss": 1.1241,
+      "step": 18973
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.979197755686515e-07,
+      "loss": 1.1931,
+      "step": 18974
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.974848238013773e-07,
+      "loss": 1.0637,
+      "step": 18975
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9705010505890705e-07,
+      "loss": 1.0781,
+      "step": 18976
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9661561935179004e-07,
+      "loss": 1.2255,
+      "step": 18977
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9618136669056783e-07,
+      "loss": 1.0061,
+      "step": 18978
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9574734708578087e-07,
+      "loss": 0.9545,
+      "step": 18979
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9531356054796076e-07,
+      "loss": 1.0943,
+      "step": 18980
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9488000708763464e-07,
+      "loss": 1.0819,
+      "step": 18981
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9444668671532185e-07,
+      "loss": 1.0975,
+      "step": 18982
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.940135994415395e-07,
+      "loss": 1.0883,
+      "step": 18983
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9358074527679704e-07,
+      "loss": 1.0706,
+      "step": 18984
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.931481242315993e-07,
+      "loss": 1.1512,
+      "step": 18985
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.927157363164413e-07,
+      "loss": 1.0554,
+      "step": 18986
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9228358154182134e-07,
+      "loss": 1.0555,
+      "step": 18987
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9185165991822203e-07,
+      "loss": 1.2578,
+      "step": 18988
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.914199714561273e-07,
+      "loss": 1.1086,
+      "step": 18989
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.909885161660099e-07,
+      "loss": 1.1697,
+      "step": 18990
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9055729405834464e-07,
+      "loss": 1.0736,
+      "step": 18991
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9012630514359215e-07,
+      "loss": 0.9843,
+      "step": 18992
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8969554943221413e-07,
+      "loss": 0.9603,
+      "step": 18993
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8926502693465983e-07,
+      "loss": 1.15,
+      "step": 18994
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.88834737661381e-07,
+      "loss": 1.0006,
+      "step": 18995
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.884046816228171e-07,
+      "loss": 1.0677,
+      "step": 18996
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.879748588294041e-07,
+      "loss": 1.0824,
+      "step": 18997
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8754526929157377e-07,
+      "loss": 1.1564,
+      "step": 18998
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8711591301975105e-07,
+      "loss": 1.0738,
+      "step": 18999
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.866867900243543e-07,
+      "loss": 1.0834,
+      "step": 19000
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8625790031579736e-07,
+      "loss": 1.0953,
+      "step": 19001
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.858292439044875e-07,
+      "loss": 1.0802,
+      "step": 19002
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8540082080082863e-07,
+      "loss": 1.0759,
+      "step": 19003
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.849726310152169e-07,
+      "loss": 1.0732,
+      "step": 19004
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8454467455804165e-07,
+      "loss": 1.0816,
+      "step": 19005
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8411695143968917e-07,
+      "loss": 1.2177,
+      "step": 19006
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.836894616705389e-07,
+      "loss": 1.0346,
+      "step": 19007
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.832622052609658e-07,
+      "loss": 1.0353,
+      "step": 19008
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8283518222133717e-07,
+      "loss": 0.8795,
+      "step": 19009
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8240839256201365e-07,
+      "loss": 0.961,
+      "step": 19010
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8198183629335586e-07,
+      "loss": 0.9614,
+      "step": 19011
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8155551342571316e-07,
+      "loss": 1.1453,
+      "step": 19012
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8112942396943185e-07,
+      "loss": 1.1849,
+      "step": 19013
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.807035679348492e-07,
+      "loss": 0.9455,
+      "step": 19014
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8027794533230354e-07,
+      "loss": 1.0552,
+      "step": 19015
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7985255617212113e-07,
+      "loss": 1.1725,
+      "step": 19016
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.794274004646259e-07,
+      "loss": 1.1374,
+      "step": 19017
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7900247822013294e-07,
+      "loss": 1.1395,
+      "step": 19018
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.785777894489562e-07,
+      "loss": 1.0811,
+      "step": 19019
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7815333416140187e-07,
+      "loss": 0.998,
+      "step": 19020
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7772911236776846e-07,
+      "loss": 1.1814,
+      "step": 19021
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7730512407835095e-07,
+      "loss": 1.0963,
+      "step": 19022
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.768813693034401e-07,
+      "loss": 1.1107,
+      "step": 19023
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7645784805331654e-07,
+      "loss": 1.1335,
+      "step": 19024
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.760345603382609e-07,
+      "loss": 1.2275,
+      "step": 19025
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.756115061685406e-07,
+      "loss": 1.1166,
+      "step": 19026
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7518868555442733e-07,
+      "loss": 0.9579,
+      "step": 19027
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.747660985061785e-07,
+      "loss": 0.9374,
+      "step": 19028
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.743437450340492e-07,
+      "loss": 1.0359,
+      "step": 19029
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.73921625148288e-07,
+      "loss": 1.2768,
+      "step": 19030
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7349973885914214e-07,
+      "loss": 0.9479,
+      "step": 19031
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.730780861768457e-07,
+      "loss": 1.0588,
+      "step": 19032
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7265666711163273e-07,
+      "loss": 1.1487,
+      "step": 19033
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7223548167372833e-07,
+      "loss": 1.0615,
+      "step": 19034
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.718145298733566e-07,
+      "loss": 0.7971,
+      "step": 19035
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7139381172072943e-07,
+      "loss": 1.3193,
+      "step": 19036
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.709733272260574e-07,
+      "loss": 1.0745,
+      "step": 19037
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7055307639954574e-07,
+      "loss": 1.0971,
+      "step": 19038
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.701330592513919e-07,
+      "loss": 1.1486,
+      "step": 19039
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.697132757917876e-07,
+      "loss": 1.0133,
+      "step": 19040
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.692937260309193e-07,
+      "loss": 1.1657,
+      "step": 19041
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6887440997897093e-07,
+      "loss": 1.1543,
+      "step": 19042
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.684553276461167e-07,
+      "loss": 1.1424,
+      "step": 19043
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6803647904252613e-07,
+      "loss": 1.1192,
+      "step": 19044
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.676178641783623e-07,
+      "loss": 1.0792,
+      "step": 19045
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6719948306378707e-07,
+      "loss": 1.1035,
+      "step": 19046
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6678133570895004e-07,
+      "loss": 1.1371,
+      "step": 19047
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.663634221240009e-07,
+      "loss": 0.9409,
+      "step": 19048
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.659457423190782e-07,
+      "loss": 1.1913,
+      "step": 19049
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.655282963043205e-07,
+      "loss": 1.1359,
+      "step": 19050
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.651110840898575e-07,
+      "loss": 1.0311,
+      "step": 19051
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6469410568581334e-07,
+      "loss": 1.2083,
+      "step": 19052
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6427736110230537e-07,
+      "loss": 0.9341,
+      "step": 19053
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6386085034945005e-07,
+      "loss": 1.1056,
+      "step": 19054
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6344457343735265e-07,
+      "loss": 1.1621,
+      "step": 19055
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.63028530376115e-07,
+      "loss": 1.1874,
+      "step": 19056
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6261272117583346e-07,
+      "loss": 1.1676,
+      "step": 19057
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.621971458465989e-07,
+      "loss": 1.2119,
+      "step": 19058
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.617818043984955e-07,
+      "loss": 1.0358,
+      "step": 19059
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.61366696841603e-07,
+      "loss": 1.2207,
+      "step": 19060
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.609518231859943e-07,
+      "loss": 1.0602,
+      "step": 19061
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.60537183441737e-07,
+      "loss": 1.122,
+      "step": 19062
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6012277761889536e-07,
+      "loss": 1.1927,
+      "step": 19063
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5970860572752343e-07,
+      "loss": 1.1145,
+      "step": 19064
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.59294667777671e-07,
+      "loss": 1.2147,
+      "step": 19065
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.588809637793855e-07,
+      "loss": 1.1552,
+      "step": 19066
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5846749374270683e-07,
+      "loss": 0.9209,
+      "step": 19067
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.580542576776658e-07,
+      "loss": 1.0153,
+      "step": 19068
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.576412555942921e-07,
+      "loss": 0.9657,
+      "step": 19069
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5722848750260886e-07,
+      "loss": 0.9885,
+      "step": 19070
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5681595341263144e-07,
+      "loss": 1.1589,
+      "step": 19071
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5640365333437065e-07,
+      "loss": 1.078,
+      "step": 19072
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.55991587277833e-07,
+      "loss": 1.1457,
+      "step": 19073
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5557975525301713e-07,
+      "loss": 1.1906,
+      "step": 19074
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.551681572699184e-07,
+      "loss": 1.1112,
+      "step": 19075
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.547567933385232e-07,
+      "loss": 1.1697,
+      "step": 19076
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.543456634688147e-07,
+      "loss": 1.1693,
+      "step": 19077
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.539347676707705e-07,
+      "loss": 1.0989,
+      "step": 19078
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5352410595436146e-07,
+      "loss": 1.1362,
+      "step": 19079
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.53113678329553e-07,
+      "loss": 0.9825,
+      "step": 19080
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.527034848063049e-07,
+      "loss": 0.9556,
+      "step": 19081
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.522935253945725e-07,
+      "loss": 1.2127,
+      "step": 19082
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.518838001043012e-07,
+      "loss": 1.1384,
+      "step": 19083
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5147430894543755e-07,
+      "loss": 0.9522,
+      "step": 19084
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5106505192791684e-07,
+      "loss": 1.066,
+      "step": 19085
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.506560290616712e-07,
+      "loss": 1.0114,
+      "step": 19086
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5024724035662596e-07,
+      "loss": 1.1047,
+      "step": 19087
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.49838685822701e-07,
+      "loss": 1.2343,
+      "step": 19088
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4943036546981056e-07,
+      "loss": 1.1036,
+      "step": 19089
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.490222793078646e-07,
+      "loss": 1.1412,
+      "step": 19090
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.486144273467662e-07,
+      "loss": 1.1368,
+      "step": 19091
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4820680959641086e-07,
+      "loss": 1.0971,
+      "step": 19092
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4779942606669057e-07,
+      "loss": 1.0482,
+      "step": 19093
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.473922767674942e-07,
+      "loss": 1.0263,
+      "step": 19094
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4698536170869825e-07,
+      "loss": 0.9889,
+      "step": 19095
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.465786809001803e-07,
+      "loss": 1.0758,
+      "step": 19096
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.46172234351807e-07,
+      "loss": 1.1288,
+      "step": 19097
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4576602207344265e-07,
+      "loss": 0.9722,
+      "step": 19098
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.453600440749461e-07,
+      "loss": 1.2292,
+      "step": 19099
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4495430036616706e-07,
+      "loss": 0.9659,
+      "step": 19100
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4454879095695226e-07,
+      "loss": 1.0668,
+      "step": 19101
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4414351585714377e-07,
+      "loss": 1.0973,
+      "step": 19102
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4373847507657486e-07,
+      "loss": 0.9681,
+      "step": 19103
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4333366862507544e-07,
+      "loss": 0.9049,
+      "step": 19104
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4292909651246766e-07,
+      "loss": 1.0316,
+      "step": 19105
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.425247587485714e-07,
+      "loss": 1.0384,
+      "step": 19106
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.421206553431966e-07,
+      "loss": 0.9835,
+      "step": 19107
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.417167863061521e-07,
+      "loss": 1.2131,
+      "step": 19108
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.413131516472357e-07,
+      "loss": 1.041,
+      "step": 19109
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.409097513762438e-07,
+      "loss": 1.0754,
+      "step": 19110
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4050658550296766e-07,
+      "loss": 1.1425,
+      "step": 19111
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4010365403718824e-07,
+      "loss": 1.1257,
+      "step": 19112
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.397009569886833e-07,
+      "loss": 1.0563,
+      "step": 19113
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.392984943672273e-07,
+      "loss": 1.0355,
+      "step": 19114
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3889626618258563e-07,
+      "loss": 1.1391,
+      "step": 19115
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.384942724445195e-07,
+      "loss": 1.1061,
+      "step": 19116
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3809251316278326e-07,
+      "loss": 1.2338,
+      "step": 19117
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.37690988347128e-07,
+      "loss": 1.1437,
+      "step": 19118
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3728969800729596e-07,
+      "loss": 1.1514,
+      "step": 19119
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.368886421530271e-07,
+      "loss": 0.9597,
+      "step": 19120
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.364878207940503e-07,
+      "loss": 1.1365,
+      "step": 19121
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.360872339400967e-07,
+      "loss": 1.2195,
+      "step": 19122
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3568688160088514e-07,
+      "loss": 1.1695,
+      "step": 19123
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.352867637861301e-07,
+      "loss": 1.1261,
+      "step": 19124
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3488688050554274e-07,
+      "loss": 0.9837,
+      "step": 19125
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.344872317688275e-07,
+      "loss": 1.166,
+      "step": 19126
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.340878175856821e-07,
+      "loss": 1.1546,
+      "step": 19127
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3368863796579776e-07,
+      "loss": 1.2317,
+      "step": 19128
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.332896929188634e-07,
+      "loss": 1.0409,
+      "step": 19129
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.328909824545601e-07,
+      "loss": 0.9397,
+      "step": 19130
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.324925065825624e-07,
+      "loss": 1.1617,
+      "step": 19131
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.320942653125381e-07,
+      "loss": 1.0715,
+      "step": 19132
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.316962586541561e-07,
+      "loss": 1.1221,
+      "step": 19133
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.312984866170721e-07,
+      "loss": 1.135,
+      "step": 19134
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.309009492109394e-07,
+      "loss": 1.0238,
+      "step": 19135
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.305036464454037e-07,
+      "loss": 1.1585,
+      "step": 19136
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.301065783301083e-07,
+      "loss": 1.271,
+      "step": 19137
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2970974487468887e-07,
+      "loss": 0.942,
+      "step": 19138
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.293131460887733e-07,
+      "loss": 1.0707,
+      "step": 19139
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.289167819819872e-07,
+      "loss": 1.2201,
+      "step": 19140
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.285206525639506e-07,
+      "loss": 1.1222,
+      "step": 19141
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.281247578442737e-07,
+      "loss": 1.2654,
+      "step": 19142
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2772909783256535e-07,
+      "loss": 0.9587,
+      "step": 19143
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2733367253842574e-07,
+      "loss": 1.1259,
+      "step": 19144
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.269384819714538e-07,
+      "loss": 1.0802,
+      "step": 19145
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2654352614123643e-07,
+      "loss": 1.1669,
+      "step": 19146
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.261488050573591e-07,
+      "loss": 1.2576,
+      "step": 19147
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2575431872939877e-07,
+      "loss": 1.1832,
+      "step": 19148
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.253600671669321e-07,
+      "loss": 1.0238,
+      "step": 19149
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.249660503795238e-07,
+      "loss": 1.2298,
+      "step": 19150
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2457226837673714e-07,
+      "loss": 1.1146,
+      "step": 19151
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2417872116812577e-07,
+      "loss": 1.0574,
+      "step": 19152
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2378540876324194e-07,
+      "loss": 1.0724,
+      "step": 19153
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.233923311716303e-07,
+      "loss": 1.124,
+      "step": 19154
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2299948840282755e-07,
+      "loss": 1.0331,
+      "step": 19155
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2260688046636847e-07,
+      "loss": 1.0348,
+      "step": 19156
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2221450737178083e-07,
+      "loss": 1.0413,
+      "step": 19157
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.21822369128586e-07,
+      "loss": 0.9324,
+      "step": 19158
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2143046574629853e-07,
+      "loss": 0.9853,
+      "step": 19159
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2103879723442976e-07,
+      "loss": 0.9638,
+      "step": 19160
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.206473636024865e-07,
+      "loss": 0.9418,
+      "step": 19161
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.202561648599645e-07,
+      "loss": 1.1248,
+      "step": 19162
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.198652010163594e-07,
+      "loss": 1.2039,
+      "step": 19163
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.19474472081156e-07,
+      "loss": 0.9919,
+      "step": 19164
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1908397806383775e-07,
+      "loss": 0.8795,
+      "step": 19165
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1869371897388147e-07,
+      "loss": 1.0888,
+      "step": 19166
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.183036948207574e-07,
+      "loss": 0.9438,
+      "step": 19167
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.17913905613928e-07,
+      "loss": 1.1206,
+      "step": 19168
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1752435136285563e-07,
+      "loss": 0.9935,
+      "step": 19169
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1713503207699283e-07,
+      "loss": 1.1411,
+      "step": 19170
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.167459477657864e-07,
+      "loss": 1.2082,
+      "step": 19171
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.163570984386766e-07,
+      "loss": 0.9425,
+      "step": 19172
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.159684841051025e-07,
+      "loss": 1.2218,
+      "step": 19173
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.155801047744944e-07,
+      "loss": 1.0733,
+      "step": 19174
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1519196045627586e-07,
+      "loss": 0.926,
+      "step": 19175
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.148040511598649e-07,
+      "loss": 0.9339,
+      "step": 19176
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.144163768946784e-07,
+      "loss": 1.0778,
+      "step": 19177
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.140289376701211e-07,
+      "loss": 1.1527,
+      "step": 19178
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.136417334955966e-07,
+      "loss": 1.0733,
+      "step": 19179
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1325476438049953e-07,
+      "loss": 1.0162,
+      "step": 19180
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1286803033422355e-07,
+      "loss": 1.2077,
+      "step": 19181
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.124815313661511e-07,
+      "loss": 1.2053,
+      "step": 19182
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1209526748566146e-07,
+      "loss": 1.2885,
+      "step": 19183
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1170923870212813e-07,
+      "loss": 1.1416,
+      "step": 19184
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.113234450249192e-07,
+      "loss": 1.0623,
+      "step": 19185
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.109378864633983e-07,
+      "loss": 1.1567,
+      "step": 19186
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1055256302692015e-07,
+      "loss": 0.9611,
+      "step": 19187
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.10167474724834e-07,
+      "loss": 0.8417,
+      "step": 19188
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0978262156648784e-07,
+      "loss": 1.052,
+      "step": 19189
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0939800356121985e-07,
+      "loss": 1.149,
+      "step": 19190
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.090136207183625e-07,
+      "loss": 1.2442,
+      "step": 19191
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0862947304724387e-07,
+      "loss": 1.1692,
+      "step": 19192
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.082455605571877e-07,
+      "loss": 1.2073,
+      "step": 19193
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.078618832575098e-07,
+      "loss": 1.1573,
+      "step": 19194
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0747844115751934e-07,
+      "loss": 1.2626,
+      "step": 19195
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.070952342665223e-07,
+      "loss": 1.1226,
+      "step": 19196
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0671226259381904e-07,
+      "loss": 1.1166,
+      "step": 19197
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.06329526148702e-07,
+      "loss": 1.0144,
+      "step": 19198
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.059470249404595e-07,
+      "loss": 1.0561,
+      "step": 19199
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.055647589783717e-07,
+      "loss": 1.0836,
+      "step": 19200
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0518272827171904e-07,
+      "loss": 1.2476,
+      "step": 19201
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0480093282976964e-07,
+      "loss": 0.9267,
+      "step": 19202
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0441937266178834e-07,
+      "loss": 0.9939,
+      "step": 19203
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.040380477770355e-07,
+      "loss": 1.1683,
+      "step": 19204
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.036569581847648e-07,
+      "loss": 1.1869,
+      "step": 19205
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.032761038942234e-07,
+      "loss": 1.1968,
+      "step": 19206
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0289548491465373e-07,
+      "loss": 1.09,
+      "step": 19207
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.025151012552929e-07,
+      "loss": 1.0354,
+      "step": 19208
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.021349529253714e-07,
+      "loss": 1.0861,
+      "step": 19209
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0175503993411405e-07,
+      "loss": 1.0864,
+      "step": 19210
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0137536229074006e-07,
+      "loss": 1.1263,
+      "step": 19211
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.009959200044632e-07,
+      "loss": 1.0388,
+      "step": 19212
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.006167130844928e-07,
+      "loss": 1.0734,
+      "step": 19213
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0023774154002925e-07,
+      "loss": 0.9667,
+      "step": 19214
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.998590053802697e-07,
+      "loss": 1.1565,
+      "step": 19215
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9948050461440446e-07,
+      "loss": 1.1904,
+      "step": 19216
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9910223925162077e-07,
+      "loss": 1.0136,
+      "step": 19217
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.987242093010956e-07,
+      "loss": 1.2785,
+      "step": 19218
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9834641477200164e-07,
+      "loss": 1.1626,
+      "step": 19219
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.979688556735105e-07,
+      "loss": 1.1522,
+      "step": 19220
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9759153201478264e-07,
+      "loss": 1.0679,
+      "step": 19221
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.972144438049751e-07,
+      "loss": 1.0337,
+      "step": 19222
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.968375910532362e-07,
+      "loss": 0.9283,
+      "step": 19223
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.964609737687141e-07,
+      "loss": 1.1037,
+      "step": 19224
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.960845919605471e-07,
+      "loss": 1.0599,
+      "step": 19225
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.957084456378689e-07,
+      "loss": 1.0115,
+      "step": 19226
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.953325348098068e-07,
+      "loss": 1.1694,
+      "step": 19227
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.949568594854846e-07,
+      "loss": 1.1362,
+      "step": 19228
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.945814196740182e-07,
+      "loss": 1.1164,
+      "step": 19229
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.942062153845171e-07,
+      "loss": 1.091,
+      "step": 19230
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.938312466260873e-07,
+      "loss": 1.1602,
+      "step": 19231
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.934565134078293e-07,
+      "loss": 1.2021,
+      "step": 19232
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.930820157388348e-07,
+      "loss": 1.0484,
+      "step": 19233
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9270775362819413e-07,
+      "loss": 1.0192,
+      "step": 19234
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9233372708498577e-07,
+      "loss": 0.9489,
+      "step": 19235
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9195993611829123e-07,
+      "loss": 1.0946,
+      "step": 19236
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9158638073717773e-07,
+      "loss": 1.1472,
+      "step": 19237
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.912130609507113e-07,
+      "loss": 1.1015,
+      "step": 19238
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.908399767679504e-07,
+      "loss": 1.1967,
+      "step": 19239
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9046712819795096e-07,
+      "loss": 1.2042,
+      "step": 19240
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.900945152497592e-07,
+      "loss": 1.1289,
+      "step": 19241
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.897221379324189e-07,
+      "loss": 1.0877,
+      "step": 19242
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8934999625496287e-07,
+      "loss": 1.1937,
+      "step": 19243
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.889780902264261e-07,
+      "loss": 1.2115,
+      "step": 19244
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.886064198558325e-07,
+      "loss": 0.9445,
+      "step": 19245
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.882349851522004e-07,
+      "loss": 1.024,
+      "step": 19246
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.878637861245426e-07,
+      "loss": 1.1188,
+      "step": 19247
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8749282278186965e-07,
+      "loss": 1.1079,
+      "step": 19248
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8712209513318325e-07,
+      "loss": 1.059,
+      "step": 19249
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.867516031874773e-07,
+      "loss": 1.1063,
+      "step": 19250
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8638134695374355e-07,
+      "loss": 1.0984,
+      "step": 19251
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8601132644096916e-07,
+      "loss": 1.1665,
+      "step": 19252
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.856415416581326e-07,
+      "loss": 1.0049,
+      "step": 19253
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8527199261420666e-07,
+      "loss": 1.0927,
+      "step": 19254
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8490267931815864e-07,
+      "loss": 1.1163,
+      "step": 19255
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.845336017789524e-07,
+      "loss": 1.246,
+      "step": 19256
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.841647600055442e-07,
+      "loss": 0.9848,
+      "step": 19257
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.837961540068834e-07,
+      "loss": 1.16,
+      "step": 19258
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.834277837919142e-07,
+      "loss": 1.1916,
+      "step": 19259
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.830596493695792e-07,
+      "loss": 1.157,
+      "step": 19260
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8269175074880915e-07,
+      "loss": 1.0445,
+      "step": 19261
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.823240879385325e-07,
+      "loss": 0.9223,
+      "step": 19262
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8195666094767203e-07,
+      "loss": 1.0728,
+      "step": 19263
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.815894697851429e-07,
+      "loss": 1.015,
+      "step": 19264
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8122251445985793e-07,
+      "loss": 1.1563,
+      "step": 19265
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8085579498072003e-07,
+      "loss": 1.2075,
+      "step": 19266
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.804893113566276e-07,
+      "loss": 1.2163,
+      "step": 19267
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.801230635964769e-07,
+      "loss": 1.1022,
+      "step": 19268
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7975705170915526e-07,
+      "loss": 0.9454,
+      "step": 19269
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7939127570354216e-07,
+      "loss": 1.2998,
+      "step": 19270
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.790257355885162e-07,
+      "loss": 0.995,
+      "step": 19271
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7866043137294687e-07,
+      "loss": 1.0777,
+      "step": 19272
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.782953630657004e-07,
+      "loss": 1.1341,
+      "step": 19273
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7793053067563415e-07,
+      "loss": 1.0679,
+      "step": 19274
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7756593421160104e-07,
+      "loss": 1.0793,
+      "step": 19275
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.772015736824518e-07,
+      "loss": 0.9366,
+      "step": 19276
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7683744909702714e-07,
+      "loss": 1.1587,
+      "step": 19277
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.764735604641622e-07,
+      "loss": 1.0725,
+      "step": 19278
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7610990779268655e-07,
+      "loss": 1.1623,
+      "step": 19279
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.757464910914287e-07,
+      "loss": 1.1395,
+      "step": 19280
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7538331036920494e-07,
+      "loss": 1.1542,
+      "step": 19281
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7502036563482826e-07,
+      "loss": 1.2563,
+      "step": 19282
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.746576568971071e-07,
+      "loss": 1.0175,
+      "step": 19283
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.742951841648445e-07,
+      "loss": 1.0625,
+      "step": 19284
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.739329474468355e-07,
+      "loss": 1.0138,
+      "step": 19285
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.735709467518699e-07,
+      "loss": 1.1637,
+      "step": 19286
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7320918208873284e-07,
+      "loss": 1.1708,
+      "step": 19287
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.728476534662039e-07,
+      "loss": 1.0488,
+      "step": 19288
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.724863608930561e-07,
+      "loss": 0.9276,
+      "step": 19289
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7212530437805696e-07,
+      "loss": 1.1378,
+      "step": 19290
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.71764483929966e-07,
+      "loss": 1.2901,
+      "step": 19291
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.714038995575441e-07,
+      "loss": 1.0734,
+      "step": 19292
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.710435512695375e-07,
+      "loss": 1.08,
+      "step": 19293
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7068343907469154e-07,
+      "loss": 1.0714,
+      "step": 19294
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7032356298174576e-07,
+      "loss": 1.202,
+      "step": 19295
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.699639229994344e-07,
+      "loss": 1.1604,
+      "step": 19296
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6960451913648264e-07,
+      "loss": 1.0275,
+      "step": 19297
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6924535140161244e-07,
+      "loss": 1.0989,
+      "step": 19298
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6888641980354125e-07,
+      "loss": 1.1489,
+      "step": 19299
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6852772435097873e-07,
+      "loss": 1.151,
+      "step": 19300
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.681692650526291e-07,
+      "loss": 1.1485,
+      "step": 19301
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6781104191719086e-07,
+      "loss": 1.0345,
+      "step": 19302
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6745305495335717e-07,
+      "loss": 1.1543,
+      "step": 19303
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6709530416981657e-07,
+      "loss": 1.2075,
+      "step": 19304
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6673778957524876e-07,
+      "loss": 1.2182,
+      "step": 19305
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6638051117832906e-07,
+      "loss": 1.038,
+      "step": 19306
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6602346898772944e-07,
+      "loss": 1.1013,
+      "step": 19307
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.656666630121152e-07,
+      "loss": 1.1079,
+      "step": 19308
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6531009326014156e-07,
+      "loss": 1.2394,
+      "step": 19309
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.649537597404639e-07,
+      "loss": 1.2179,
+      "step": 19310
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.645976624617286e-07,
+      "loss": 1.1192,
+      "step": 19311
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.642418014325776e-07,
+      "loss": 1.0325,
+      "step": 19312
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.638861766616463e-07,
+      "loss": 1.1035,
+      "step": 19313
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.635307881575633e-07,
+      "loss": 1.2042,
+      "step": 19314
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.631756359289561e-07,
+      "loss": 1.028,
+      "step": 19315
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6282071998444015e-07,
+      "loss": 1.151,
+      "step": 19316
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6246604033262955e-07,
+      "loss": 1.104,
+      "step": 19317
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6211159698213087e-07,
+      "loss": 1.1749,
+      "step": 19318
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.61757389941546e-07,
+      "loss": 1.0478,
+      "step": 19319
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.614034192194703e-07,
+      "loss": 1.0885,
+      "step": 19320
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6104968482449366e-07,
+      "loss": 1.0771,
+      "step": 19321
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6069618676519806e-07,
+      "loss": 1.1009,
+      "step": 19322
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6034292505016546e-07,
+      "loss": 1.1306,
+      "step": 19323
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.599898996879668e-07,
+      "loss": 1.1517,
+      "step": 19324
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5963711068716867e-07,
+      "loss": 1.0341,
+      "step": 19325
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.592845580563319e-07,
+      "loss": 1.1601,
+      "step": 19326
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5893224180401297e-07,
+      "loss": 1.3346,
+      "step": 19327
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5858016193876066e-07,
+      "loss": 1.2193,
+      "step": 19328
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5822831846912035e-07,
+      "loss": 1.1297,
+      "step": 19329
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5787671140362847e-07,
+      "loss": 1.0289,
+      "step": 19330
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5752534075081825e-07,
+      "loss": 1.0439,
+      "step": 19331
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.571742065192162e-07,
+      "loss": 1.1334,
+      "step": 19332
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.568233087173433e-07,
+      "loss": 1.143,
+      "step": 19333
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.564726473537149e-07,
+      "loss": 1.0894,
+      "step": 19334
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5612222243684206e-07,
+      "loss": 1.0988,
+      "step": 19335
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5577203397522676e-07,
+      "loss": 1.2334,
+      "step": 19336
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.554220819773667e-07,
+      "loss": 1.0213,
+      "step": 19337
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5507236645175515e-07,
+      "loss": 1.1085,
+      "step": 19338
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5472288740687856e-07,
+      "loss": 1.148,
+      "step": 19339
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.543736448512191e-07,
+      "loss": 1.2269,
+      "step": 19340
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.540246387932488e-07,
+      "loss": 0.9814,
+      "step": 19341
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.536758692414387e-07,
+      "loss": 1.0626,
+      "step": 19342
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5332733620425207e-07,
+      "loss": 1.2089,
+      "step": 19343
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5297903969014883e-07,
+      "loss": 0.9532,
+      "step": 19344
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5263097970757767e-07,
+      "loss": 0.9526,
+      "step": 19345
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5228315626498636e-07,
+      "loss": 1.1327,
+      "step": 19346
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.51935569370817e-07,
+      "loss": 1.1811,
+      "step": 19347
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5158821903350284e-07,
+      "loss": 1.0127,
+      "step": 19348
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5124110526147383e-07,
+      "loss": 1.206,
+      "step": 19349
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.508942280631521e-07,
+      "loss": 1.0608,
+      "step": 19350
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.505475874469576e-07,
+      "loss": 1.2612,
+      "step": 19351
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.502011834213014e-07,
+      "loss": 1.0105,
+      "step": 19352
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4985501599458896e-07,
+      "loss": 1.0101,
+      "step": 19353
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4950908517522023e-07,
+      "loss": 1.1671,
+      "step": 19354
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.491633909715918e-07,
+      "loss": 1.1062,
+      "step": 19355
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.488179333920915e-07,
+      "loss": 1.0916,
+      "step": 19356
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.484727124451025e-07,
+      "loss": 1.1485,
+      "step": 19357
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4812772813900155e-07,
+      "loss": 1.1436,
+      "step": 19358
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.47782980482163e-07,
+      "loss": 1.1943,
+      "step": 19359
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4743846948295124e-07,
+      "loss": 1.0406,
+      "step": 19360
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.470941951497263e-07,
+      "loss": 1.0613,
+      "step": 19361
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.467501574908415e-07,
+      "loss": 1.0103,
+      "step": 19362
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.464063565146491e-07,
+      "loss": 1.2342,
+      "step": 19363
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.460627922294889e-07,
+      "loss": 1.0259,
+      "step": 19364
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.457194646437011e-07,
+      "loss": 1.1004,
+      "step": 19365
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.453763737656134e-07,
+      "loss": 1.0762,
+      "step": 19366
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4503351960355473e-07,
+      "loss": 1.0817,
+      "step": 19367
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.44690902165845e-07,
+      "loss": 1.0375,
+      "step": 19368
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.443485214607966e-07,
+      "loss": 1.2255,
+      "step": 19369
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.440063774967194e-07,
+      "loss": 1.1257,
+      "step": 19370
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.436644702819169e-07,
+      "loss": 1.0596,
+      "step": 19371
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4332279982468454e-07,
+      "loss": 1.0347,
+      "step": 19372
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.429813661333147e-07,
+      "loss": 1.1513,
+      "step": 19373
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.426401692160929e-07,
+      "loss": 1.0065,
+      "step": 19374
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.422992090812981e-07,
+      "loss": 1.1455,
+      "step": 19375
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.419584857372059e-07,
+      "loss": 1.0516,
+      "step": 19376
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4161799919208416e-07,
+      "loss": 1.0892,
+      "step": 19377
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4127774945419404e-07,
+      "loss": 1.1404,
+      "step": 19378
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.409377365317944e-07,
+      "loss": 1.0878,
+      "step": 19379
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4059796043313653e-07,
+      "loss": 1.1982,
+      "step": 19380
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4025842116646267e-07,
+      "loss": 1.1526,
+      "step": 19381
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.399191187400163e-07,
+      "loss": 1.1432,
+      "step": 19382
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3958005316202847e-07,
+      "loss": 1.0533,
+      "step": 19383
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.392412244407294e-07,
+      "loss": 1.0122,
+      "step": 19384
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.389026325843391e-07,
+      "loss": 1.1045,
+      "step": 19385
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.385642776010766e-07,
+      "loss": 1.2585,
+      "step": 19386
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3822615949915085e-07,
+      "loss": 1.1912,
+      "step": 19387
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3788827828676864e-07,
+      "loss": 1.0596,
+      "step": 19388
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3755063397212785e-07,
+      "loss": 1.1489,
+      "step": 19389
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3721322656342416e-07,
+      "loss": 1.2104,
+      "step": 19390
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3687605606884324e-07,
+      "loss": 1.0568,
+      "step": 19391
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3653912249656853e-07,
+      "loss": 0.9928,
+      "step": 19392
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.362024258547746e-07,
+      "loss": 1.2059,
+      "step": 19393
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3586596615163492e-07,
+      "loss": 1.122,
+      "step": 19394
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.35529743395313e-07,
+      "loss": 1.146,
+      "step": 19395
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.351937575939678e-07,
+      "loss": 0.9942,
+      "step": 19396
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.348580087557517e-07,
+      "loss": 1.0407,
+      "step": 19397
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.345224968888149e-07,
+      "loss": 1.111,
+      "step": 19398
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.341872220012975e-07,
+      "loss": 1.0247,
+      "step": 19399
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3385218410133526e-07,
+      "loss": 1.0922,
+      "step": 19400
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.335173831970594e-07,
+      "loss": 0.9359,
+      "step": 19401
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.331828192965957e-07,
+      "loss": 0.8461,
+      "step": 19402
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3284849240806096e-07,
+      "loss": 1.2273,
+      "step": 19403
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.325144025395698e-07,
+      "loss": 1.255,
+      "step": 19404
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3218054969922687e-07,
+      "loss": 1.3292,
+      "step": 19405
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.318469338951379e-07,
+      "loss": 1.1763,
+      "step": 19406
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.315135551353964e-07,
+      "loss": 1.0105,
+      "step": 19407
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3118041342809262e-07,
+      "loss": 1.1725,
+      "step": 19408
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3084750878131112e-07,
+      "loss": 1.0919,
+      "step": 19409
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3051484120313105e-07,
+      "loss": 1.1036,
+      "step": 19410
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3018241070162372e-07,
+      "loss": 1.2829,
+      "step": 19411
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2985021728485824e-07,
+      "loss": 1.1139,
+      "step": 19412
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2951826096089368e-07,
+      "loss": 1.2888,
+      "step": 19413
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.291865417377881e-07,
+      "loss": 1.1597,
+      "step": 19414
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2885505962359055e-07,
+      "loss": 1.0896,
+      "step": 19415
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2852381462634354e-07,
+      "loss": 1.0996,
+      "step": 19416
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.281928067540884e-07,
+      "loss": 1.0052,
+      "step": 19417
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2786203601485424e-07,
+      "loss": 1.1177,
+      "step": 19418
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2753150241667022e-07,
+      "loss": 1.156,
+      "step": 19419
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.272012059675577e-07,
+      "loss": 1.1872,
+      "step": 19420
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2687114667552912e-07,
+      "loss": 1.027,
+      "step": 19421
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2654132454859812e-07,
+      "loss": 1.2595,
+      "step": 19422
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2621173959476607e-07,
+      "loss": 1.0523,
+      "step": 19423
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2588239182203097e-07,
+      "loss": 1.1699,
+      "step": 19424
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2555328123838537e-07,
+      "loss": 1.0584,
+      "step": 19425
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.252244078518162e-07,
+      "loss": 1.0022,
+      "step": 19426
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2489577167030486e-07,
+      "loss": 1.1175,
+      "step": 19427
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2456737270182493e-07,
+      "loss": 1.0525,
+      "step": 19428
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2423921095434564e-07,
+      "loss": 1.0991,
+      "step": 19429
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.239112864358317e-07,
+      "loss": 1.0009,
+      "step": 19430
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.235835991542412e-07,
+      "loss": 0.949,
+      "step": 19431
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2325614911752557e-07,
+      "loss": 1.0691,
+      "step": 19432
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2292893633362844e-07,
+      "loss": 1.1231,
+      "step": 19433
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2260196081049458e-07,
+      "loss": 1.1149,
+      "step": 19434
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.222752225560576e-07,
+      "loss": 1.1638,
+      "step": 19435
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2194872157824453e-07,
+      "loss": 1.0703,
+      "step": 19436
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2162245788497906e-07,
+      "loss": 1.16,
+      "step": 19437
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2129643148418034e-07,
+      "loss": 1.0634,
+      "step": 19438
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2097064238375987e-07,
+      "loss": 1.2155,
+      "step": 19439
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2064509059162243e-07,
+      "loss": 1.0813,
+      "step": 19440
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2031977611566723e-07,
+      "loss": 1.0348,
+      "step": 19441
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1999469896379133e-07,
+      "loss": 1.1649,
+      "step": 19442
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1966985914388173e-07,
+      "loss": 1.2367,
+      "step": 19443
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1934525666382211e-07,
+      "loss": 1.1871,
+      "step": 19444
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.190208915314873e-07,
+      "loss": 1.1384,
+      "step": 19445
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1869676375475324e-07,
+      "loss": 1.0876,
+      "step": 19446
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.183728733414814e-07,
+      "loss": 1.1608,
+      "step": 19447
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1804922029953436e-07,
+      "loss": 1.1279,
+      "step": 19448
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1772580463676364e-07,
+      "loss": 1.0968,
+      "step": 19449
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.174026263610196e-07,
+      "loss": 1.2093,
+      "step": 19450
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1707968548014492e-07,
+      "loss": 1.1471,
+      "step": 19451
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1675698200197547e-07,
+      "loss": 1.0362,
+      "step": 19452
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1643451593434172e-07,
+      "loss": 0.9565,
+      "step": 19453
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1611228728507072e-07,
+      "loss": 0.8413,
+      "step": 19454
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1579029606198065e-07,
+      "loss": 1.087,
+      "step": 19455
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.154685422728864e-07,
+      "loss": 1.1264,
+      "step": 19456
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1514702592559388e-07,
+      "loss": 1.0927,
+      "step": 19457
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1482574702790804e-07,
+      "loss": 1.0576,
+      "step": 19458
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.145047055876248e-07,
+      "loss": 1.0917,
+      "step": 19459
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1418390161253356e-07,
+      "loss": 0.8991,
+      "step": 19460
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1386333511041912e-07,
+      "loss": 1.1975,
+      "step": 19461
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1354300608906198e-07,
+      "loss": 1.0417,
+      "step": 19462
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.132229145562359e-07,
+      "loss": 1.0662,
+      "step": 19463
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.12903060519708e-07,
+      "loss": 1.0809,
+      "step": 19464
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1258344398723875e-07,
+      "loss": 1.1277,
+      "step": 19465
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1226406496658635e-07,
+      "loss": 1.0561,
+      "step": 19466
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.119449234655002e-07,
+      "loss": 1.1748,
+      "step": 19467
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1162601949172413e-07,
+      "loss": 1.1907,
+      "step": 19468
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1130735305299966e-07,
+      "loss": 1.0885,
+      "step": 19469
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.109889241570573e-07,
+      "loss": 1.1217,
+      "step": 19470
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1067073281162642e-07,
+      "loss": 0.9673,
+      "step": 19471
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.103527790244253e-07,
+      "loss": 1.1757,
+      "step": 19472
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1003506280317332e-07,
+      "loss": 1.1682,
+      "step": 19473
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0971758415557986e-07,
+      "loss": 0.9107,
+      "step": 19474
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0940034308934764e-07,
+      "loss": 1.0445,
+      "step": 19475
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0908333961217496e-07,
+      "loss": 1.2183,
+      "step": 19476
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0876657373175678e-07,
+      "loss": 1.1551,
+      "step": 19477
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0845004545577919e-07,
+      "loss": 1.1009,
+      "step": 19478
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0813375479192266e-07,
+      "loss": 0.9638,
+      "step": 19479
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0781770174786219e-07,
+      "loss": 1.1281,
+      "step": 19480
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0750188633126943e-07,
+      "loss": 1.1023,
+      "step": 19481
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0718630854980715e-07,
+      "loss": 1.1058,
+      "step": 19482
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0687096841113364e-07,
+      "loss": 1.1635,
+      "step": 19483
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0655586592289944e-07,
+      "loss": 0.9321,
+      "step": 19484
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0624100109275514e-07,
+      "loss": 0.9599,
+      "step": 19485
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.059263739283379e-07,
+      "loss": 1.0041,
+      "step": 19486
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0561198443728502e-07,
+      "loss": 1.0402,
+      "step": 19487
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.052978326272237e-07,
+      "loss": 0.959,
+      "step": 19488
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0498391850578002e-07,
+      "loss": 0.9448,
+      "step": 19489
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0467024208057128e-07,
+      "loss": 1.0303,
+      "step": 19490
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0435680335920805e-07,
+      "loss": 1.1748,
+      "step": 19491
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0404360234929644e-07,
+      "loss": 1.049,
+      "step": 19492
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0373063905843815e-07,
+      "loss": 1.1043,
+      "step": 19493
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0341791349422713e-07,
+      "loss": 1.1846,
+      "step": 19494
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0310542566425396e-07,
+      "loss": 1.0469,
+      "step": 19495
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0279317557609922e-07,
+      "loss": 1.2567,
+      "step": 19496
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0248116323734134e-07,
+      "loss": 1.1279,
+      "step": 19497
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0216938865555312e-07,
+      "loss": 1.2153,
+      "step": 19498
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0185785183829964e-07,
+      "loss": 1.0021,
+      "step": 19499
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0154655279313927e-07,
+      "loss": 1.1954,
+      "step": 19500
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0123549152762823e-07,
+      "loss": 0.9515,
+      "step": 19501
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0092466804931597e-07,
+      "loss": 1.0843,
+      "step": 19502
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.006140823657432e-07,
+      "loss": 1.0364,
+      "step": 19503
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0030373448444719e-07,
+      "loss": 0.9371,
+      "step": 19504
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9999362441296077e-07,
+      "loss": 1.2382,
+      "step": 19505
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.99683752158808e-07,
+      "loss": 0.9988,
+      "step": 19506
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9937411772950831e-07,
+      "loss": 1.1922,
+      "step": 19507
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.990647211325758e-07,
+      "loss": 1.0481,
+      "step": 19508
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9875556237551997e-07,
+      "loss": 1.0479,
+      "step": 19509
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9844664146584258e-07,
+      "loss": 1.1101,
+      "step": 19510
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.981379584110399e-07,
+      "loss": 1.2021,
+      "step": 19511
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9782951321860033e-07,
+      "loss": 1.0637,
+      "step": 19512
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9752130589601348e-07,
+      "loss": 1.0857,
+      "step": 19513
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.972133364507567e-07,
+      "loss": 0.9437,
+      "step": 19514
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9690560489030285e-07,
+      "loss": 1.1584,
+      "step": 19515
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9659811122211936e-07,
+      "loss": 1.1765,
+      "step": 19516
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.962908554536691e-07,
+      "loss": 1.2015,
+      "step": 19517
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9598383759240948e-07,
+      "loss": 1.1404,
+      "step": 19518
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9567705764578894e-07,
+      "loss": 1.0495,
+      "step": 19519
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9537051562125152e-07,
+      "loss": 0.9634,
+      "step": 19520
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.95064211526238e-07,
+      "loss": 1.2268,
+      "step": 19521
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9475814536818127e-07,
+      "loss": 1.0465,
+      "step": 19522
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9445231715450875e-07,
+      "loss": 1.1623,
+      "step": 19523
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.941467268926389e-07,
+      "loss": 1.1594,
+      "step": 19524
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9384137458999252e-07,
+      "loss": 1.003,
+      "step": 19525
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9353626025397588e-07,
+      "loss": 1.207,
+      "step": 19526
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9323138389199526e-07,
+      "loss": 1.1189,
+      "step": 19527
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9292674551144696e-07,
+      "loss": 1.1823,
+      "step": 19528
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9262234511972622e-07,
+      "loss": 1.1969,
+      "step": 19529
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.923181827242182e-07,
+      "loss": 1.062,
+      "step": 19530
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9201425833230481e-07,
+      "loss": 1.0798,
+      "step": 19531
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.917105719513601e-07,
+      "loss": 1.1883,
+      "step": 19532
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9140712358875602e-07,
+      "loss": 1.0291,
+      "step": 19533
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.911039132518544e-07,
+      "loss": 1.0176,
+      "step": 19534
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9080094094801493e-07,
+      "loss": 1.0007,
+      "step": 19535
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.904982066845873e-07,
+      "loss": 1.0526,
+      "step": 19536
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9019571046892006e-07,
+      "loss": 1.0408,
+      "step": 19537
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8989345230835289e-07,
+      "loss": 0.9934,
+      "step": 19538
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8959143221022214e-07,
+      "loss": 1.1129,
+      "step": 19539
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.892896501818553e-07,
+      "loss": 1.1732,
+      "step": 19540
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8898810623057651e-07,
+      "loss": 1.1888,
+      "step": 19541
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8868680036370323e-07,
+      "loss": 1.1231,
+      "step": 19542
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8838573258854743e-07,
+      "loss": 1.0356,
+      "step": 19543
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8808490291241433e-07,
+      "loss": 1.0972,
+      "step": 19544
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8778431134260476e-07,
+      "loss": 1.1572,
+      "step": 19545
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8748395788641404e-07,
+      "loss": 1.1387,
+      "step": 19546
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8718384255113075e-07,
+      "loss": 1.0768,
+      "step": 19547
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8688396534403575e-07,
+      "loss": 1.1237,
+      "step": 19548
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8658432627240764e-07,
+      "loss": 1.0915,
+      "step": 19549
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.862849253435184e-07,
+      "loss": 1.0039,
+      "step": 19550
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8598576256463218e-07,
+      "loss": 1.0175,
+      "step": 19551
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.856868379430099e-07,
+      "loss": 1.1377,
+      "step": 19552
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8538815148590573e-07,
+      "loss": 1.078,
+      "step": 19553
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.850897032005672e-07,
+      "loss": 0.8702,
+      "step": 19554
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8479149309423627e-07,
+      "loss": 1.0558,
+      "step": 19555
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8449352117415052e-07,
+      "loss": 1.2545,
+      "step": 19556
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8419578744754085e-07,
+      "loss": 1.1846,
+      "step": 19557
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8389829192163256e-07,
+      "loss": 1.2484,
+      "step": 19558
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8360103460364432e-07,
+      "loss": 1.0974,
+      "step": 19559
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8330401550078924e-07,
+      "loss": 1.2294,
+      "step": 19560
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8300723462027714e-07,
+      "loss": 1.1924,
+      "step": 19561
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.827106919693089e-07,
+      "loss": 1.0509,
+      "step": 19562
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8241438755507878e-07,
+      "loss": 1.0846,
+      "step": 19563
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8211832138477993e-07,
+      "loss": 1.113,
+      "step": 19564
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8182249346559655e-07,
+      "loss": 1.1553,
+      "step": 19565
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.815269038047063e-07,
+      "loss": 0.9112,
+      "step": 19566
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8123155240928336e-07,
+      "loss": 1.1609,
+      "step": 19567
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.809364392864943e-07,
+      "loss": 1.1879,
+      "step": 19568
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8064156444350223e-07,
+      "loss": 1.1757,
+      "step": 19569
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8034692788746034e-07,
+      "loss": 1.2131,
+      "step": 19570
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8005252962551955e-07,
+      "loss": 1.0961,
+      "step": 19571
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.797583696648264e-07,
+      "loss": 1.0982,
+      "step": 19572
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7946444801251627e-07,
+      "loss": 0.9635,
+      "step": 19573
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7917076467572236e-07,
+      "loss": 1.0914,
+      "step": 19574
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.788773196615723e-07,
+      "loss": 1.1947,
+      "step": 19575
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7858411297718704e-07,
+      "loss": 1.0112,
+      "step": 19576
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.78291144629682e-07,
+      "loss": 1.1073,
+      "step": 19577
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7799841462616595e-07,
+      "loss": 0.8042,
+      "step": 19578
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.777059229737421e-07,
+      "loss": 0.9669,
+      "step": 19579
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7741366967950923e-07,
+      "loss": 1.0816,
+      "step": 19580
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7712165475056053e-07,
+      "loss": 1.0805,
+      "step": 19581
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7682987819398035e-07,
+      "loss": 1.1351,
+      "step": 19582
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7653834001685077e-07,
+      "loss": 1.1434,
+      "step": 19583
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7624704022624395e-07,
+      "loss": 1.0727,
+      "step": 19584
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.759559788292331e-07,
+      "loss": 1.1644,
+      "step": 19585
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7566515583287925e-07,
+      "loss": 1.219,
+      "step": 19586
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7537457124423896e-07,
+      "loss": 1.0311,
+      "step": 19587
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7508422507036438e-07,
+      "loss": 0.974,
+      "step": 19588
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.747941173183021e-07,
+      "loss": 1.2032,
+      "step": 19589
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7450424799509093e-07,
+      "loss": 1.2322,
+      "step": 19590
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7421461710776632e-07,
+      "loss": 1.1019,
+      "step": 19591
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7392522466335714e-07,
+      "loss": 1.1262,
+      "step": 19592
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7363607066888444e-07,
+      "loss": 1.0123,
+      "step": 19593
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.73347155131367e-07,
+      "loss": 1.1602,
+      "step": 19594
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7305847805781373e-07,
+      "loss": 1.3006,
+      "step": 19595
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7277003945523117e-07,
+      "loss": 1.0245,
+      "step": 19596
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7248183933062046e-07,
+      "loss": 1.1648,
+      "step": 19597
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7219387769097262e-07,
+      "loss": 1.0302,
+      "step": 19598
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7190615454327654e-07,
+      "loss": 1.0796,
+      "step": 19599
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7161866989451548e-07,
+      "loss": 1.0701,
+      "step": 19600
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7133142375166502e-07,
+      "loss": 1.0027,
+      "step": 19601
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7104441612169508e-07,
+      "loss": 1.136,
+      "step": 19602
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7075764701157128e-07,
+      "loss": 0.9252,
+      "step": 19603
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7047111642825355e-07,
+      "loss": 0.9941,
+      "step": 19604
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.701848243786941e-07,
+      "loss": 1.1233,
+      "step": 19605
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6989877086984076e-07,
+      "loss": 1.048,
+      "step": 19606
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6961295590863347e-07,
+      "loss": 1.1738,
+      "step": 19607
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6932737950201117e-07,
+      "loss": 1.029,
+      "step": 19608
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6904204165690163e-07,
+      "loss": 1.0111,
+      "step": 19609
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6875694238023042e-07,
+      "loss": 1.1189,
+      "step": 19610
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6847208167891428e-07,
+      "loss": 1.1152,
+      "step": 19611
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.681874595598687e-07,
+      "loss": 0.9735,
+      "step": 19612
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6790307602999823e-07,
+      "loss": 1.13,
+      "step": 19613
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.676189310962062e-07,
+      "loss": 1.0454,
+      "step": 19614
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.673350247653849e-07,
+      "loss": 1.117,
+      "step": 19615
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6705135704442654e-07,
+      "loss": 1.1419,
+      "step": 19616
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6676792794021457e-07,
+      "loss": 1.1886,
+      "step": 19617
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6648473745962678e-07,
+      "loss": 1.0062,
+      "step": 19618
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6620178560953437e-07,
+      "loss": 1.0695,
+      "step": 19619
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6591907239680406e-07,
+      "loss": 1.0852,
+      "step": 19620
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.656365978282981e-07,
+      "loss": 1.0968,
+      "step": 19621
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6535436191086884e-07,
+      "loss": 1.2311,
+      "step": 19622
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6507236465136633e-07,
+      "loss": 1.0345,
+      "step": 19623
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.647906060566351e-07,
+      "loss": 1.0724,
+      "step": 19624
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.645090861335119e-07,
+      "loss": 1.1015,
+      "step": 19625
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6422780488882795e-07,
+      "loss": 1.0243,
+      "step": 19626
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6394676232940777e-07,
+      "loss": 1.0831,
+      "step": 19627
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6366595846207366e-07,
+      "loss": 1.0638,
+      "step": 19628
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.633853932936391e-07,
+      "loss": 1.0911,
+      "step": 19629
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.631050668309131e-07,
+      "loss": 1.067,
+      "step": 19630
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6282497908069684e-07,
+      "loss": 1.0553,
+      "step": 19631
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6254513004978934e-07,
+      "loss": 1.0446,
+      "step": 19632
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6226551974498073e-07,
+      "loss": 1.0987,
+      "step": 19633
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6198614817305557e-07,
+      "loss": 1.0297,
+      "step": 19634
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6170701534079291e-07,
+      "loss": 1.0824,
+      "step": 19635
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6142812125496954e-07,
+      "loss": 1.2171,
+      "step": 19636
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6114946592235115e-07,
+      "loss": 1.1244,
+      "step": 19637
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6087104934969899e-07,
+      "loss": 1.0892,
+      "step": 19638
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6059287154377212e-07,
+      "loss": 1.0613,
+      "step": 19639
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6031493251131846e-07,
+      "loss": 1.1101,
+      "step": 19640
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6003723225908484e-07,
+      "loss": 1.0074,
+      "step": 19641
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.597597707938081e-07,
+      "loss": 0.9765,
+      "step": 19642
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5948254812222397e-07,
+      "loss": 1.1391,
+      "step": 19643
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5920556425105705e-07,
+      "loss": 1.0325,
+      "step": 19644
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5892881918703195e-07,
+      "loss": 1.1348,
+      "step": 19645
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.586523129368611e-07,
+      "loss": 1.1913,
+      "step": 19646
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5837604550725803e-07,
+      "loss": 1.026,
+      "step": 19647
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5810001690492405e-07,
+      "loss": 1.0101,
+      "step": 19648
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5782422713655931e-07,
+      "loss": 1.0295,
+      "step": 19649
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5754867620885516e-07,
+      "loss": 1.003,
+      "step": 19650
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.572733641284996e-07,
+      "loss": 1.0881,
+      "step": 19651
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.569982909021728e-07,
+      "loss": 1.0756,
+      "step": 19652
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.567234565365505e-07,
+      "loss": 1.0516,
+      "step": 19653
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5644886103830082e-07,
+      "loss": 0.9537,
+      "step": 19654
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5617450441408943e-07,
+      "loss": 1.0095,
+      "step": 19655
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5590038667057328e-07,
+      "loss": 1.0974,
+      "step": 19656
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5562650781440368e-07,
+      "loss": 1.1802,
+      "step": 19657
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5535286785222758e-07,
+      "loss": 0.977,
+      "step": 19658
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.550794667906852e-07,
+      "loss": 1.0479,
+      "step": 19659
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5480630463641234e-07,
+      "loss": 1.0808,
+      "step": 19660
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.545333813960359e-07,
+      "loss": 1.0649,
+      "step": 19661
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5426069707617952e-07,
+      "loss": 1.1802,
+      "step": 19662
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.539882516834612e-07,
+      "loss": 1.0814,
+      "step": 19663
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5371604522449235e-07,
+      "loss": 1.1784,
+      "step": 19664
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5344407770587765e-07,
+      "loss": 1.1107,
+      "step": 19665
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.531723491342174e-07,
+      "loss": 1.2928,
+      "step": 19666
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5290085951610521e-07,
+      "loss": 0.9237,
+      "step": 19667
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5262960885813026e-07,
+      "loss": 0.9453,
+      "step": 19668
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5235859716687506e-07,
+      "loss": 1.1924,
+      "step": 19669
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5208782444891656e-07,
+      "loss": 1.1414,
+      "step": 19670
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5181729071082397e-07,
+      "loss": 0.9952,
+      "step": 19671
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5154699595916312e-07,
+      "loss": 1.1069,
+      "step": 19672
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5127694020049432e-07,
+      "loss": 1.0348,
+      "step": 19673
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5100712344136904e-07,
+      "loss": 1.1101,
+      "step": 19674
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5073754568833643e-07,
+      "loss": 0.9836,
+      "step": 19675
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5046820694793794e-07,
+      "loss": 1.1054,
+      "step": 19676
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.501991072267095e-07,
+      "loss": 1.1572,
+      "step": 19677
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.499302465311814e-07,
+      "loss": 1.1205,
+      "step": 19678
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.496616248678784e-07,
+      "loss": 1.1318,
+      "step": 19679
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4939324224331974e-07,
+      "loss": 1.1117,
+      "step": 19680
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.491250986640169e-07,
+      "loss": 0.9868,
+      "step": 19681
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4885719413647692e-07,
+      "loss": 1.174,
+      "step": 19682
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4858952866720234e-07,
+      "loss": 1.074,
+      "step": 19683
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4832210226268794e-07,
+      "loss": 1.07,
+      "step": 19684
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4805491492942415e-07,
+      "loss": 1.0852,
+      "step": 19685
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4778796667389238e-07,
+      "loss": 1.0552,
+      "step": 19686
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4752125750257307e-07,
+      "loss": 1.1677,
+      "step": 19687
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4725478742193766e-07,
+      "loss": 1.0688,
+      "step": 19688
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4698855643845324e-07,
+      "loss": 1.0129,
+      "step": 19689
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4672256455857902e-07,
+      "loss": 0.8933,
+      "step": 19690
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4645681178877103e-07,
+      "loss": 1.1375,
+      "step": 19691
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.461912981354785e-07,
+      "loss": 1.0014,
+      "step": 19692
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4592602360514406e-07,
+      "loss": 1.1553,
+      "step": 19693
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4566098820420483e-07,
+      "loss": 1.1298,
+      "step": 19694
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4539619193909338e-07,
+      "loss": 1.2564,
+      "step": 19695
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.451316348162346e-07,
+      "loss": 1.1152,
+      "step": 19696
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.448673168420489e-07,
+      "loss": 0.9157,
+      "step": 19697
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4460323802295007e-07,
+      "loss": 1.1675,
+      "step": 19698
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4433939836534738e-07,
+      "loss": 1.1863,
+      "step": 19699
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4407579787564353e-07,
+      "loss": 1.0722,
+      "step": 19700
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4381243656023446e-07,
+      "loss": 0.8412,
+      "step": 19701
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4354931442551067e-07,
+      "loss": 1.1296,
+      "step": 19702
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4328643147785925e-07,
+      "loss": 1.1999,
+      "step": 19703
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4302378772365844e-07,
+      "loss": 1.0657,
+      "step": 19704
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.42761383169282e-07,
+      "loss": 1.0226,
+      "step": 19705
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.424992178210971e-07,
+      "loss": 0.9936,
+      "step": 19706
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.422372916854675e-07,
+      "loss": 0.9597,
+      "step": 19707
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4197560476874706e-07,
+      "loss": 1.1333,
+      "step": 19708
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4171415707728843e-07,
+      "loss": 1.1502,
+      "step": 19709
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.414529486174332e-07,
+      "loss": 1.0321,
+      "step": 19710
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.411919793955241e-07,
+      "loss": 1.1125,
+      "step": 19711
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.409312494178905e-07,
+      "loss": 1.112,
+      "step": 19712
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4067075869086178e-07,
+      "loss": 1.0727,
+      "step": 19713
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4041050722075734e-07,
+      "loss": 1.0817,
+      "step": 19714
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4015049501389543e-07,
+      "loss": 1.008,
+      "step": 19715
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3989072207658328e-07,
+      "loss": 1.0716,
+      "step": 19716
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.396311884151269e-07,
+      "loss": 1.0798,
+      "step": 19717
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3937189403582129e-07,
+      "loss": 1.0825,
+      "step": 19718
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3911283894496253e-07,
+      "loss": 1.0126,
+      "step": 19719
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3885402314883445e-07,
+      "loss": 1.0421,
+      "step": 19720
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.385954466537187e-07,
+      "loss": 1.2022,
+      "step": 19721
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.383371094658903e-07,
+      "loss": 0.9644,
+      "step": 19722
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3807901159161862e-07,
+      "loss": 1.0449,
+      "step": 19723
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3782115303716537e-07,
+      "loss": 0.9833,
+      "step": 19724
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3756353380878994e-07,
+      "loss": 1.1878,
+      "step": 19725
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3730615391274293e-07,
+      "loss": 1.047,
+      "step": 19726
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3704901335527044e-07,
+      "loss": 1.2371,
+      "step": 19727
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.367921121426119e-07,
+      "loss": 0.9587,
+      "step": 19728
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3653545028100235e-07,
+      "loss": 0.9816,
+      "step": 19729
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3627902777667123e-07,
+      "loss": 1.1138,
+      "step": 19730
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3602284463583914e-07,
+      "loss": 1.071,
+      "step": 19731
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3576690086472332e-07,
+      "loss": 1.1868,
+      "step": 19732
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3551119646953548e-07,
+      "loss": 1.0536,
+      "step": 19733
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3525573145648062e-07,
+      "loss": 1.1367,
+      "step": 19734
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3500050583175828e-07,
+      "loss": 1.2019,
+      "step": 19735
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3474551960156125e-07,
+      "loss": 1.0368,
+      "step": 19736
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3449077277207678e-07,
+      "loss": 0.9929,
+      "step": 19737
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3423626534948886e-07,
+      "loss": 1.0688,
+      "step": 19738
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3398199733997364e-07,
+      "loss": 1.2169,
+      "step": 19739
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.337279687496995e-07,
+      "loss": 1.0702,
+      "step": 19740
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3347417958483155e-07,
+      "loss": 0.9804,
+      "step": 19741
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.332206298515293e-07,
+      "loss": 1.0471,
+      "step": 19742
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3296731955594443e-07,
+      "loss": 1.0031,
+      "step": 19743
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3271424870422545e-07,
+      "loss": 1.1717,
+      "step": 19744
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3246141730251293e-07,
+      "loss": 1.2056,
+      "step": 19745
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3220882535694312e-07,
+      "loss": 1.0072,
+      "step": 19746
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3195647287364443e-07,
+      "loss": 1.1617,
+      "step": 19747
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3170435985874085e-07,
+      "loss": 1.0363,
+      "step": 19748
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3145248631835084e-07,
+      "loss": 1.0321,
+      "step": 19749
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3120085225858614e-07,
+      "loss": 1.1698,
+      "step": 19750
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3094945768555523e-07,
+      "loss": 1.2318,
+      "step": 19751
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3069830260535653e-07,
+      "loss": 1.0045,
+      "step": 19752
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3044738702408523e-07,
+      "loss": 1.1138,
+      "step": 19753
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.301967109478297e-07,
+      "loss": 1.0619,
+      "step": 19754
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2994627438267514e-07,
+      "loss": 1.2439,
+      "step": 19755
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.296960773346978e-07,
+      "loss": 1.2587,
+      "step": 19756
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2944611980996947e-07,
+      "loss": 1.147,
+      "step": 19757
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2919640181455417e-07,
+      "loss": 1.1587,
+      "step": 19758
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2894692335451376e-07,
+      "loss": 1.0914,
+      "step": 19759
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2869768443590224e-07,
+      "loss": 1.213,
+      "step": 19760
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2844868506476815e-07,
+      "loss": 1.038,
+      "step": 19761
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2819992524715107e-07,
+      "loss": 1.1309,
+      "step": 19762
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2795140498909174e-07,
+      "loss": 1.072,
+      "step": 19763
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.277031242966187e-07,
+      "loss": 1.1279,
+      "step": 19764
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2745508317575704e-07,
+      "loss": 1.1002,
+      "step": 19765
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2720728163252648e-07,
+      "loss": 1.093,
+      "step": 19766
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2695971967294108e-07,
+      "loss": 1.1667,
+      "step": 19767
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.267123973030071e-07,
+      "loss": 1.1259,
+      "step": 19768
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2646531452872645e-07,
+      "loss": 1.1007,
+      "step": 19769
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2621847135609656e-07,
+      "loss": 1.0506,
+      "step": 19770
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2597186779110593e-07,
+      "loss": 1.2881,
+      "step": 19771
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.257255038397398e-07,
+      "loss": 1.1383,
+      "step": 19772
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2547937950797673e-07,
+      "loss": 0.9898,
+      "step": 19773
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.252334948017886e-07,
+      "loss": 1.1503,
+      "step": 19774
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2498784972714395e-07,
+      "loss": 1.0475,
+      "step": 19775
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.247424442900025e-07,
+      "loss": 1.2371,
+      "step": 19776
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2449727849631944e-07,
+      "loss": 1.14,
+      "step": 19777
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2425235235204447e-07,
+      "loss": 0.8905,
+      "step": 19778
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2400766586312174e-07,
+      "loss": 1.1592,
+      "step": 19779
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2376321903548872e-07,
+      "loss": 1.2082,
+      "step": 19780
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.235190118750773e-07,
+      "loss": 1.0588,
+      "step": 19781
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.232750443878139e-07,
+      "loss": 0.9365,
+      "step": 19782
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2303131657961932e-07,
+      "loss": 1.0778,
+      "step": 19783
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.227878284564066e-07,
+      "loss": 1.1512,
+      "step": 19784
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2254458002408544e-07,
+      "loss": 1.1246,
+      "step": 19785
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.223015712885589e-07,
+      "loss": 1.2505,
+      "step": 19786
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2205880225572453e-07,
+      "loss": 1.0915,
+      "step": 19787
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2181627293147312e-07,
+      "loss": 1.1723,
+      "step": 19788
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2157398332168892e-07,
+      "loss": 1.2087,
+      "step": 19789
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.213319334322538e-07,
+      "loss": 1.2094,
+      "step": 19790
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.210901232690398e-07,
+      "loss": 1.0104,
+      "step": 19791
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2084855283791664e-07,
+      "loss": 1.0652,
+      "step": 19792
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2060722214474406e-07,
+      "loss": 1.1222,
+      "step": 19793
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2036613119538075e-07,
+      "loss": 1.0122,
+      "step": 19794
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2012527999567757e-07,
+      "loss": 1.0048,
+      "step": 19795
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.198846685514776e-07,
+      "loss": 1.0138,
+      "step": 19796
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1964429686861844e-07,
+      "loss": 1.2016,
+      "step": 19797
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.194041649529376e-07,
+      "loss": 1.1107,
+      "step": 19798
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1916427281025933e-07,
+      "loss": 0.9982,
+      "step": 19799
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1892462044640452e-07,
+      "loss": 1.1091,
+      "step": 19800
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1868520786719073e-07,
+      "loss": 1.1206,
+      "step": 19801
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1844603507842667e-07,
+      "loss": 1.0232,
+      "step": 19802
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1820710208591768e-07,
+      "loss": 1.1647,
+      "step": 19803
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1796840889546024e-07,
+      "loss": 0.9508,
+      "step": 19804
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.177299555128475e-07,
+      "loss": 1.1317,
+      "step": 19805
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1749174194386592e-07,
+      "loss": 1.018,
+      "step": 19806
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1725376819429757e-07,
+      "loss": 1.055,
+      "step": 19807
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1701603426991448e-07,
+      "loss": 1.0688,
+      "step": 19808
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1677854017648871e-07,
+      "loss": 1.2491,
+      "step": 19809
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.165412859197812e-07,
+      "loss": 1.0457,
+      "step": 19810
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.163042715055518e-07,
+      "loss": 1.1975,
+      "step": 19811
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1606749693954922e-07,
+      "loss": 0.9866,
+      "step": 19812
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1583096222752222e-07,
+      "loss": 1.0085,
+      "step": 19813
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1559466737520842e-07,
+      "loss": 1.0898,
+      "step": 19814
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1535861238834433e-07,
+      "loss": 1.1802,
+      "step": 19815
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.151227972726554e-07,
+      "loss": 1.137,
+      "step": 19816
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1488722203386704e-07,
+      "loss": 1.2405,
+      "step": 19817
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1465188667769467e-07,
+      "loss": 0.9622,
+      "step": 19818
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1441679120984927e-07,
+      "loss": 1.0436,
+      "step": 19819
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1418193563603519e-07,
+      "loss": 1.1083,
+      "step": 19820
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1394731996195229e-07,
+      "loss": 1.0229,
+      "step": 19821
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1371294419329493e-07,
+      "loss": 1.004,
+      "step": 19822
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1347880833574964e-07,
+      "loss": 0.9869,
+      "step": 19823
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1324491239499746e-07,
+      "loss": 1.1142,
+      "step": 19824
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1301125637671717e-07,
+      "loss": 1.1567,
+      "step": 19825
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1277784028657645e-07,
+      "loss": 1.1742,
+      "step": 19826
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1254466413023967e-07,
+      "loss": 1.1324,
+      "step": 19827
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1231172791336564e-07,
+      "loss": 1.1619,
+      "step": 19828
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1207903164160872e-07,
+      "loss": 1.1783,
+      "step": 19829
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1184657532061328e-07,
+      "loss": 0.9899,
+      "step": 19830
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1161435895602147e-07,
+      "loss": 1.1219,
+      "step": 19831
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1138238255346878e-07,
+      "loss": 1.0719,
+      "step": 19832
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1115064611858406e-07,
+      "loss": 1.1074,
+      "step": 19833
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1091914965699168e-07,
+      "loss": 1.0496,
+      "step": 19834
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1068789317430828e-07,
+      "loss": 1.1774,
+      "step": 19835
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1045687667614602e-07,
+      "loss": 1.1575,
+      "step": 19836
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1022610016811153e-07,
+      "loss": 0.9963,
+      "step": 19837
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0999556365580477e-07,
+      "loss": 1.2176,
+      "step": 19838
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0976526714482128e-07,
+      "loss": 1.2777,
+      "step": 19839
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0953521064074768e-07,
+      "loss": 1.1918,
+      "step": 19840
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.093053941491673e-07,
+      "loss": 1.0496,
+      "step": 19841
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.090758176756579e-07,
+      "loss": 1.1075,
+      "step": 19842
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0884648122579056e-07,
+      "loss": 1.059,
+      "step": 19843
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0861738480513084e-07,
+      "loss": 1.0369,
+      "step": 19844
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0838852841923541e-07,
+      "loss": 1.1217,
+      "step": 19845
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0815991207366206e-07,
+      "loss": 1.0669,
+      "step": 19846
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0793153577395743e-07,
+      "loss": 1.0945,
+      "step": 19847
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0770339952566044e-07,
+      "loss": 1.0334,
+      "step": 19848
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0747550333431223e-07,
+      "loss": 1.0574,
+      "step": 19849
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0724784720543946e-07,
+      "loss": 1.1094,
+      "step": 19850
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0702043114456883e-07,
+      "loss": 1.0409,
+      "step": 19851
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0679325515721595e-07,
+      "loss": 1.2375,
+      "step": 19852
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0656631924889749e-07,
+      "loss": 1.0965,
+      "step": 19853
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0633962342511905e-07,
+      "loss": 1.1618,
+      "step": 19854
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.061131676913818e-07,
+      "loss": 1.0661,
+      "step": 19855
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0588695205318022e-07,
+      "loss": 1.1978,
+      "step": 19856
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0566097651600548e-07,
+      "loss": 1.1536,
+      "step": 19857
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0543524108534098e-07,
+      "loss": 1.0577,
+      "step": 19858
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0520974576666343e-07,
+      "loss": 1.0402,
+      "step": 19859
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0498449056544624e-07,
+      "loss": 1.1527,
+      "step": 19860
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0475947548715503e-07,
+      "loss": 1.0834,
+      "step": 19861
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0453470053725101e-07,
+      "loss": 1.1473,
+      "step": 19862
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0431016572118868e-07,
+      "loss": 1.1331,
+      "step": 19863
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0408587104441481e-07,
+      "loss": 1.0846,
+      "step": 19864
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0386181651237504e-07,
+      "loss": 1.1027,
+      "step": 19865
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0363800213050501e-07,
+      "loss": 1.1626,
+      "step": 19866
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0341442790423706e-07,
+      "loss": 1.1242,
+      "step": 19867
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0319109383899572e-07,
+      "loss": 1.1382,
+      "step": 19868
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0296799994020224e-07,
+      "loss": 1.0112,
+      "step": 19869
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.027451462132678e-07,
+      "loss": 1.2294,
+      "step": 19870
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0252253266360257e-07,
+      "loss": 1.1548,
+      "step": 19871
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0230015929660775e-07,
+      "loss": 1.0288,
+      "step": 19872
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0207802611768014e-07,
+      "loss": 0.8579,
+      "step": 19873
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0185613313220988e-07,
+      "loss": 1.0846,
+      "step": 19874
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0163448034558266e-07,
+      "loss": 0.9775,
+      "step": 19875
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0141306776317527e-07,
+      "loss": 1.075,
+      "step": 19876
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0119189539036345e-07,
+      "loss": 1.0139,
+      "step": 19877
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0097096323251177e-07,
+      "loss": 1.0267,
+      "step": 19878
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0075027129498371e-07,
+      "loss": 1.0401,
+      "step": 19879
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.005298195831328e-07,
+      "loss": 1.1883,
+      "step": 19880
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0030960810231139e-07,
+      "loss": 1.0485,
+      "step": 19881
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0008963685786078e-07,
+      "loss": 1.215,
+      "step": 19882
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.986990585512001e-08,
+      "loss": 1.0665,
+      "step": 19883
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.965041509942153e-08,
+      "loss": 1.0495,
+      "step": 19884
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.943116459609214e-08,
+      "loss": 0.9734,
+      "step": 19885
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.921215435045206e-08,
+      "loss": 1.0232,
+      "step": 19886
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.89933843678148e-08,
+      "loss": 0.981,
+      "step": 19887
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.877485465349057e-08,
+      "loss": 0.9946,
+      "step": 19888
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.855656521278289e-08,
+      "loss": 1.1784,
+      "step": 19889
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.833851605098754e-08,
+      "loss": 1.1099,
+      "step": 19890
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.812070717339694e-08,
+      "loss": 1.1395,
+      "step": 19891
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.790313858529577e-08,
+      "loss": 1.2738,
+      "step": 19892
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.768581029196534e-08,
+      "loss": 1.1803,
+      "step": 19893
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.746872229867809e-08,
+      "loss": 0.9644,
+      "step": 19894
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.725187461070207e-08,
+      "loss": 1.2217,
+      "step": 19895
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.703526723330192e-08,
+      "loss": 0.8795,
+      "step": 19896
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.681890017173123e-08,
+      "loss": 1.1695,
+      "step": 19897
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.660277343124135e-08,
+      "loss": 1.0523,
+      "step": 19898
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.638688701707699e-08,
+      "loss": 1.0597,
+      "step": 19899
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.617124093447838e-08,
+      "loss": 1.1192,
+      "step": 19900
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.59558351886769e-08,
+      "loss": 1.1635,
+      "step": 19901
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.57406697849017e-08,
+      "loss": 1.0401,
+      "step": 19902
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.552574472837195e-08,
+      "loss": 1.0224,
+      "step": 19903
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.531106002430457e-08,
+      "loss": 1.1595,
+      "step": 19904
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.509661567790874e-08,
+      "loss": 1.0148,
+      "step": 19905
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.488241169439028e-08,
+      "loss": 0.9744,
+      "step": 19906
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.466844807894393e-08,
+      "loss": 1.2119,
+      "step": 19907
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.445472483676443e-08,
+      "loss": 1.191,
+      "step": 19908
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.424124197303764e-08,
+      "loss": 1.1662,
+      "step": 19909
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.402799949294494e-08,
+      "loss": 1.2124,
+      "step": 19910
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.381499740165889e-08,
+      "loss": 1.0971,
+      "step": 19911
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.36022357043509e-08,
+      "loss": 1.0403,
+      "step": 19912
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.338971440618238e-08,
+      "loss": 0.8919,
+      "step": 19913
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.317743351231256e-08,
+      "loss": 1.1722,
+      "step": 19914
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.296539302788953e-08,
+      "loss": 1.048,
+      "step": 19915
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.275359295806142e-08,
+      "loss": 1.114,
+      "step": 19916
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.25420333079674e-08,
+      "loss": 1.1576,
+      "step": 19917
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.23307140827423e-08,
+      "loss": 1.0691,
+      "step": 19918
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.211963528751311e-08,
+      "loss": 1.2291,
+      "step": 19919
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.190879692740129e-08,
+      "loss": 1.0825,
+      "step": 19920
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.169819900752497e-08,
+      "loss": 1.1744,
+      "step": 19921
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.14878415329945e-08,
+      "loss": 1.1095,
+      "step": 19922
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.12777245089147e-08,
+      "loss": 1.0847,
+      "step": 19923
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.106784794038259e-08,
+      "loss": 1.2001,
+      "step": 19924
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.08582118324941e-08,
+      "loss": 1.1344,
+      "step": 19925
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.064881619033406e-08,
+      "loss": 1.1302,
+      "step": 19926
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.043966101898505e-08,
+      "loss": 1.1854,
+      "step": 19927
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.02307463235219e-08,
+      "loss": 1.2075,
+      "step": 19928
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.002207210901615e-08,
+      "loss": 1.122,
+      "step": 19929
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.981363838053037e-08,
+      "loss": 1.0852,
+      "step": 19930
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.960544514312275e-08,
+      "loss": 0.97,
+      "step": 19931
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.93974924018448e-08,
+      "loss": 1.1941,
+      "step": 19932
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.918978016174473e-08,
+      "loss": 1.1999,
+      "step": 19933
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.898230842786071e-08,
+      "loss": 1.126,
+      "step": 19934
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.877507720522982e-08,
+      "loss": 1.0875,
+      "step": 19935
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.856808649887916e-08,
+      "loss": 1.0688,
+      "step": 19936
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.836133631383359e-08,
+      "loss": 1.0378,
+      "step": 19937
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.81548266551091e-08,
+      "loss": 0.9298,
+      "step": 19938
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.794855752771613e-08,
+      "loss": 1.0682,
+      "step": 19939
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.77425289366629e-08,
+      "loss": 1.1322,
+      "step": 19940
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.753674088694652e-08,
+      "loss": 0.9974,
+      "step": 19941
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.733119338356299e-08,
+      "loss": 1.1317,
+      "step": 19942
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.712588643149834e-08,
+      "loss": 1.0376,
+      "step": 19943
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.692082003573632e-08,
+      "loss": 1.1139,
+      "step": 19944
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.671599420125188e-08,
+      "loss": 1.0651,
+      "step": 19945
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.651140893301768e-08,
+      "loss": 1.0697,
+      "step": 19946
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.630706423599533e-08,
+      "loss": 1.1297,
+      "step": 19947
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.61029601151464e-08,
+      "loss": 1.1311,
+      "step": 19948
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.58990965754225e-08,
+      "loss": 1.1597,
+      "step": 19949
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.569547362177189e-08,
+      "loss": 1.0443,
+      "step": 19950
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.549209125913505e-08,
+      "loss": 1.092,
+      "step": 19951
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.528894949244692e-08,
+      "loss": 1.1553,
+      "step": 19952
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.508604832663913e-08,
+      "loss": 1.1118,
+      "step": 19953
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.48833877666333e-08,
+      "loss": 1.0713,
+      "step": 19954
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.468096781734769e-08,
+      "loss": 1.2285,
+      "step": 19955
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.447878848369618e-08,
+      "loss": 1.0597,
+      "step": 19956
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.42768497705837e-08,
+      "loss": 0.9819,
+      "step": 19957
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.407515168291192e-08,
+      "loss": 1.0734,
+      "step": 19958
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.387369422557246e-08,
+      "loss": 1.1667,
+      "step": 19959
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.367247740345808e-08,
+      "loss": 1.1545,
+      "step": 19960
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.34715012214482e-08,
+      "loss": 1.0429,
+      "step": 19961
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.327076568442339e-08,
+      "loss": 1.0719,
+      "step": 19962
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.307027079725082e-08,
+      "loss": 1.1324,
+      "step": 19963
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.287001656479999e-08,
+      "loss": 1.1474,
+      "step": 19964
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.267000299192806e-08,
+      "loss": 1.0754,
+      "step": 19965
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.247023008348898e-08,
+      "loss": 1.0753,
+      "step": 19966
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.227069784433107e-08,
+      "loss": 1.0909,
+      "step": 19967
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.2071406279296e-08,
+      "loss": 1.1077,
+      "step": 19968
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.187235539322103e-08,
+      "loss": 1.1107,
+      "step": 19969
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.167354519093562e-08,
+      "loss": 0.8954,
+      "step": 19970
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.147497567726482e-08,
+      "loss": 1.1595,
+      "step": 19971
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.127664685702696e-08,
+      "loss": 1.1786,
+      "step": 19972
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.10785587350349e-08,
+      "loss": 1.2408,
+      "step": 19973
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.088071131609587e-08,
+      "loss": 1.2346,
+      "step": 19974
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.068310460500939e-08,
+      "loss": 1.2082,
+      "step": 19975
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.048573860657383e-08,
+      "loss": 1.1105,
+      "step": 19976
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.028861332557759e-08,
+      "loss": 0.9886,
+      "step": 19977
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.009172876680238e-08,
+      "loss": 1.0473,
+      "step": 19978
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.989508493502773e-08,
+      "loss": 1.0774,
+      "step": 19979
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.969868183502649e-08,
+      "loss": 1.1254,
+      "step": 19980
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.950251947156263e-08,
+      "loss": 1.1307,
+      "step": 19981
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.930659784939787e-08,
+      "loss": 1.1527,
+      "step": 19982
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.911091697328621e-08,
+      "loss": 1.0037,
+      "step": 19983
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.891547684797719e-08,
+      "loss": 1.0087,
+      "step": 19984
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.872027747821142e-08,
+      "loss": 1.2624,
+      "step": 19985
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.852531886872738e-08,
+      "loss": 1.1955,
+      "step": 19986
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.833060102425682e-08,
+      "loss": 1.2215,
+      "step": 19987
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.813612394952264e-08,
+      "loss": 1.1361,
+      "step": 19988
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.79418876492466e-08,
+      "loss": 1.166,
+      "step": 19989
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.774789212814049e-08,
+      "loss": 1.0836,
+      "step": 19990
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.755413739091277e-08,
+      "loss": 1.17,
+      "step": 19991
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.736062344226636e-08,
+      "loss": 1.0903,
+      "step": 19992
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.716735028689526e-08,
+      "loss": 1.0673,
+      "step": 19993
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.697431792949017e-08,
+      "loss": 1.1248,
+      "step": 19994
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.678152637473624e-08,
+      "loss": 1.0379,
+      "step": 19995
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.658897562731194e-08,
+      "loss": 1.1968,
+      "step": 19996
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.639666569188908e-08,
+      "loss": 1.1619,
+      "step": 19997
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.620459657313505e-08,
+      "loss": 1.0443,
+      "step": 19998
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.601276827571057e-08,
+      "loss": 1.0762,
+      "step": 19999
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.582118080427081e-08,
+      "loss": 1.0869,
+      "step": 20000
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.562983416346426e-08,
+      "loss": 1.0692,
+      "step": 20001
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.543872835793608e-08,
+      "loss": 1.0836,
+      "step": 20002
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.52478633923226e-08,
+      "loss": 1.0744,
+      "step": 20003
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.505723927125563e-08,
+      "loss": 1.0278,
+      "step": 20004
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.48668559993615e-08,
+      "loss": 1.0176,
+      "step": 20005
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.467671358126093e-08,
+      "loss": 1.0604,
+      "step": 20006
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.448681202156582e-08,
+      "loss": 1.1433,
+      "step": 20007
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.42971513248858e-08,
+      "loss": 1.1045,
+      "step": 20008
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.410773149582496e-08,
+      "loss": 1.0104,
+      "step": 20009
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.391855253897739e-08,
+      "loss": 1.1924,
+      "step": 20010
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.372961445893501e-08,
+      "loss": 1.1935,
+      "step": 20011
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.354091726028301e-08,
+      "loss": 1.0283,
+      "step": 20012
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.335246094759996e-08,
+      "loss": 1.1885,
+      "step": 20013
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.316424552545998e-08,
+      "loss": 1.1292,
+      "step": 20014
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.297627099842941e-08,
+      "loss": 1.0101,
+      "step": 20015
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.278853737107128e-08,
+      "loss": 1.1324,
+      "step": 20016
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.260104464793971e-08,
+      "loss": 1.068,
+      "step": 20017
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.241379283358552e-08,
+      "loss": 1.0726,
+      "step": 20018
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.222678193255172e-08,
+      "loss": 1.0823,
+      "step": 20019
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.204001194937915e-08,
+      "loss": 1.0605,
+      "step": 20020
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.18534828885975e-08,
+      "loss": 1.0724,
+      "step": 20021
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.166719475473316e-08,
+      "loss": 1.0287,
+      "step": 20022
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.148114755230918e-08,
+      "loss": 1.0766,
+      "step": 20023
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.129534128583971e-08,
+      "loss": 1.1467,
+      "step": 20024
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.110977595983226e-08,
+      "loss": 1.1443,
+      "step": 20025
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.092445157878991e-08,
+      "loss": 1.1137,
+      "step": 20026
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.073936814721128e-08,
+      "loss": 1.1374,
+      "step": 20027
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.05545256695872e-08,
+      "loss": 1.0637,
+      "step": 20028
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.036992415040411e-08,
+      "loss": 1.1425,
+      "step": 20029
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.018556359413953e-08,
+      "loss": 1.0214,
+      "step": 20030
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.000144400526987e-08,
+      "loss": 1.1169,
+      "step": 20031
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.981756538826268e-08,
+      "loss": 1.155,
+      "step": 20032
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.963392774757883e-08,
+      "loss": 1.0082,
+      "step": 20033
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.945053108767475e-08,
+      "loss": 0.8932,
+      "step": 20034
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.926737541300243e-08,
+      "loss": 0.9697,
+      "step": 20035
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.9084460728005e-08,
+      "loss": 0.9974,
+      "step": 20036
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.890178703712225e-08,
+      "loss": 0.9999,
+      "step": 20037
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.871935434478727e-08,
+      "loss": 1.2748,
+      "step": 20038
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.853716265542654e-08,
+      "loss": 1.0009,
+      "step": 20039
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.835521197346096e-08,
+      "loss": 0.9822,
+      "step": 20040
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.817350230330699e-08,
+      "loss": 1.2471,
+      "step": 20041
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.799203364937334e-08,
+      "loss": 0.9726,
+      "step": 20042
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.781080601606427e-08,
+      "loss": 1.0893,
+      "step": 20043
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.762981940777735e-08,
+      "loss": 1.254,
+      "step": 20044
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.744907382890464e-08,
+      "loss": 1.1934,
+      "step": 20045
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.726856928383152e-08,
+      "loss": 1.2065,
+      "step": 20046
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.708830577694004e-08,
+      "loss": 1.2188,
+      "step": 20047
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.690828331260446e-08,
+      "loss": 1.119,
+      "step": 20048
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.672850189519131e-08,
+      "loss": 1.1554,
+      "step": 20049
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.654896152906487e-08,
+      "loss": 1.2053,
+      "step": 20050
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.636966221858276e-08,
+      "loss": 0.8789,
+      "step": 20051
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.619060396809484e-08,
+      "loss": 1.1094,
+      "step": 20052
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.601178678194653e-08,
+      "loss": 1.0502,
+      "step": 20053
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.583321066447657e-08,
+      "loss": 1.0511,
+      "step": 20054
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.565487562002037e-08,
+      "loss": 1.0216,
+      "step": 20055
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.54767816529045e-08,
+      "loss": 1.1272,
+      "step": 20056
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.529892876745104e-08,
+      "loss": 1.0237,
+      "step": 20057
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.51213169679743e-08,
+      "loss": 1.14,
+      "step": 20058
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.494394625878641e-08,
+      "loss": 1.0369,
+      "step": 20059
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.476681664419171e-08,
+      "loss": 1.0765,
+      "step": 20060
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.458992812848675e-08,
+      "loss": 1.1314,
+      "step": 20061
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.441328071596587e-08,
+      "loss": 1.0932,
+      "step": 20062
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.423687441091453e-08,
+      "loss": 1.1794,
+      "step": 20063
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.406070921761487e-08,
+      "loss": 1.164,
+      "step": 20064
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.388478514034124e-08,
+      "loss": 1.0753,
+      "step": 20065
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.370910218336245e-08,
+      "loss": 1.149,
+      "step": 20066
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.353366035094288e-08,
+      "loss": 1.0461,
+      "step": 20067
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.335845964733911e-08,
+      "loss": 1.1781,
+      "step": 20068
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.318350007680219e-08,
+      "loss": 1.2174,
+      "step": 20069
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.300878164357871e-08,
+      "loss": 0.9476,
+      "step": 20070
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.283430435190973e-08,
+      "loss": 0.9212,
+      "step": 20071
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.266006820602633e-08,
+      "loss": 1.1533,
+      "step": 20072
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.248607321015954e-08,
+      "loss": 1.077,
+      "step": 20073
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.231231936853044e-08,
+      "loss": 1.1292,
+      "step": 20074
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.213880668535455e-08,
+      "loss": 1.0341,
+      "step": 20075
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.196553516484515e-08,
+      "loss": 0.9627,
+      "step": 20076
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.179250481120446e-08,
+      "loss": 0.9354,
+      "step": 20077
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.161971562863245e-08,
+      "loss": 1.0939,
+      "step": 20078
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.144716762132353e-08,
+      "loss": 1.0571,
+      "step": 20079
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.127486079346212e-08,
+      "loss": 1.1064,
+      "step": 20080
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.110279514923156e-08,
+      "loss": 1.0468,
+      "step": 20081
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.09309706928074e-08,
+      "loss": 1.161,
+      "step": 20082
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.075938742835741e-08,
+      "loss": 1.2277,
+      "step": 20083
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.058804536004826e-08,
+      "loss": 1.1633,
+      "step": 20084
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.041694449203661e-08,
+      "loss": 0.9649,
+      "step": 20085
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.02460848284736e-08,
+      "loss": 0.9946,
+      "step": 20086
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.007546637350704e-08,
+      "loss": 1.0243,
+      "step": 20087
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.99050891312758e-08,
+      "loss": 1.0568,
+      "step": 20088
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.97349531059166e-08,
+      "loss": 1.0009,
+      "step": 20089
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.9565058301555015e-08,
+      "loss": 1.1464,
+      "step": 20090
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.939540472231664e-08,
+      "loss": 0.975,
+      "step": 20091
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.9225992372318184e-08,
+      "loss": 1.0273,
+      "step": 20092
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.905682125566969e-08,
+      "loss": 1.0624,
+      "step": 20093
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.888789137647566e-08,
+      "loss": 1.1749,
+      "step": 20094
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8719202738838355e-08,
+      "loss": 1.0145,
+      "step": 20095
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.855075534684895e-08,
+      "loss": 1.2245,
+      "step": 20096
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.838254920459641e-08,
+      "loss": 1.1343,
+      "step": 20097
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.821458431616189e-08,
+      "loss": 1.0089,
+      "step": 20098
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8046860685622156e-08,
+      "loss": 1.1447,
+      "step": 20099
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.787937831704615e-08,
+      "loss": 0.9468,
+      "step": 20100
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.771213721449953e-08,
+      "loss": 0.9509,
+      "step": 20101
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.754513738203904e-08,
+      "loss": 1.1273,
+      "step": 20102
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.737837882371921e-08,
+      "loss": 1.0837,
+      "step": 20103
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.721186154358571e-08,
+      "loss": 1.1645,
+      "step": 20104
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.704558554567974e-08,
+      "loss": 1.2075,
+      "step": 20105
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.687955083403695e-08,
+      "loss": 1.1454,
+      "step": 20106
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.671375741268526e-08,
+      "loss": 1.1172,
+      "step": 20107
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6548205285648084e-08,
+      "loss": 1.16,
+      "step": 20108
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.638289445694334e-08,
+      "loss": 1.0412,
+      "step": 20109
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6217824930582257e-08,
+      "loss": 1.1385,
+      "step": 20110
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.605299671057163e-08,
+      "loss": 1.1151,
+      "step": 20111
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.588840980091159e-08,
+      "loss": 1.0274,
+      "step": 20112
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5724064205593396e-08,
+      "loss": 1.133,
+      "step": 20113
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.55599599286083e-08,
+      "loss": 1.0421,
+      "step": 20114
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5396096973937554e-08,
+      "loss": 0.9675,
+      "step": 20115
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.523247534555687e-08,
+      "loss": 1.0822,
+      "step": 20116
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.506909504743751e-08,
+      "loss": 0.9633,
+      "step": 20117
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.49059560835441e-08,
+      "loss": 0.8844,
+      "step": 20118
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.474305845783678e-08,
+      "loss": 1.1685,
+      "step": 20119
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.458040217426686e-08,
+      "loss": 1.0743,
+      "step": 20120
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4417987236781156e-08,
+      "loss": 1.0992,
+      "step": 20121
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.42558136493232e-08,
+      "loss": 1.087,
+      "step": 20122
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.40938814158265e-08,
+      "loss": 1.048,
+      "step": 20123
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.3932190540222364e-08,
+      "loss": 1.1061,
+      "step": 20124
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.377074102643209e-08,
+      "loss": 1.0333,
+      "step": 20125
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.3609532878377e-08,
+      "loss": 1.2197,
+      "step": 20126
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.344856609996618e-08,
+      "loss": 1.1085,
+      "step": 20127
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.328784069510762e-08,
+      "loss": 0.9256,
+      "step": 20128
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.31273566676993e-08,
+      "loss": 1.189,
+      "step": 20129
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2967114021639234e-08,
+      "loss": 0.9977,
+      "step": 20130
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2807112760813186e-08,
+      "loss": 1.0029,
+      "step": 20131
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.264735288910472e-08,
+      "loss": 1.0698,
+      "step": 20132
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2487834410391844e-08,
+      "loss": 1.1118,
+      "step": 20133
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.232855732854369e-08,
+      "loss": 1.0936,
+      "step": 20134
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.216952164742717e-08,
+      "loss": 0.9894,
+      "step": 20135
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.201072737090029e-08,
+      "loss": 1.1443,
+      "step": 20136
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1852174502816654e-08,
+      "loss": 0.9728,
+      "step": 20137
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1693863047025395e-08,
+      "loss": 1.1435,
+      "step": 20138
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1535793007366774e-08,
+      "loss": 1.1086,
+      "step": 20139
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.137796438767662e-08,
+      "loss": 1.2026,
+      "step": 20140
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.122037719178519e-08,
+      "loss": 1.0045,
+      "step": 20141
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.106303142351721e-08,
+      "loss": 0.9758,
+      "step": 20142
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.090592708669184e-08,
+      "loss": 1.1185,
+      "step": 20143
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.074906418511938e-08,
+      "loss": 1.1637,
+      "step": 20144
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0592442722606775e-08,
+      "loss": 1.047,
+      "step": 20145
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0436062702956536e-08,
+      "loss": 1.0826,
+      "step": 20146
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0279924129962296e-08,
+      "loss": 1.2308,
+      "step": 20147
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0124027007412145e-08,
+      "loss": 1.0357,
+      "step": 20148
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9968371339090824e-08,
+      "loss": 1.2136,
+      "step": 20149
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.981295712877643e-08,
+      "loss": 1.1297,
+      "step": 20150
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.965778438023816e-08,
+      "loss": 0.9515,
+      "step": 20151
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9502853097241896e-08,
+      "loss": 1.1255,
+      "step": 20152
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.934816328354908e-08,
+      "loss": 1.1768,
+      "step": 20153
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.919371494291225e-08,
+      "loss": 1.2649,
+      "step": 20154
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9039508079079535e-08,
+      "loss": 1.0772,
+      "step": 20155
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8885542695792374e-08,
+      "loss": 1.0021,
+      "step": 20156
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8731818796788896e-08,
+      "loss": 1.0451,
+      "step": 20157
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8578336385799454e-08,
+      "loss": 0.9004,
+      "step": 20158
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.842509546654661e-08,
+      "loss": 1.0768,
+      "step": 20159
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.827209604274963e-08,
+      "loss": 1.0118,
+      "step": 20160
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8119338118122196e-08,
+      "loss": 1.1711,
+      "step": 20161
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.7966821696372455e-08,
+      "loss": 1.1153,
+      "step": 20162
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.7814546781198565e-08,
+      "loss": 1.147,
+      "step": 20163
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.766251337629757e-08,
+      "loss": 1.1665,
+      "step": 20164
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.7510721485358734e-08,
+      "loss": 0.984,
+      "step": 20165
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.7359171112064675e-08,
+      "loss": 1.1882,
+      "step": 20166
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.720786226009466e-08,
+      "loss": 1.2239,
+      "step": 20167
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.705679493312021e-08,
+      "loss": 1.1606,
+      "step": 20168
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6905969134806164e-08,
+      "loss": 1.0664,
+      "step": 20169
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6755384868812925e-08,
+      "loss": 1.07,
+      "step": 20170
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6605042138796466e-08,
+      "loss": 1.1192,
+      "step": 20171
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.645494094840275e-08,
+      "loss": 1.0951,
+      "step": 20172
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6305081301274425e-08,
+      "loss": 1.0495,
+      "step": 20173
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.61554632010508e-08,
+      "loss": 1.0201,
+      "step": 20174
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.600608665136008e-08,
+      "loss": 1.1392,
+      "step": 20175
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.585695165582715e-08,
+      "loss": 1.206,
+      "step": 20176
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.570805821807245e-08,
+      "loss": 1.1065,
+      "step": 20177
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.555940634170974e-08,
+      "loss": 1.1778,
+      "step": 20178
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.541099603034394e-08,
+      "loss": 1.2233,
+      "step": 20179
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5262827287578804e-08,
+      "loss": 1.1593,
+      "step": 20180
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5114900117008144e-08,
+      "loss": 1.1445,
+      "step": 20181
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.496721452222352e-08,
+      "loss": 1.2334,
+      "step": 20182
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.481977050680764e-08,
+      "loss": 1.0182,
+      "step": 20183
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.467256807433873e-08,
+      "loss": 1.1179,
+      "step": 20184
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.452560722838839e-08,
+      "loss": 1.222,
+      "step": 20185
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.437888797252377e-08,
+      "loss": 1.2799,
+      "step": 20186
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.423241031030534e-08,
+      "loss": 1.0824,
+      "step": 20187
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.408617424528805e-08,
+      "loss": 1.1335,
+      "step": 20188
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.394017978101905e-08,
+      "loss": 0.9479,
+      "step": 20189
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3794426921042185e-08,
+      "loss": 1.0965,
+      "step": 20190
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.364891566889573e-08,
+      "loss": 1.1317,
+      "step": 20191
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.350364602810797e-08,
+      "loss": 1.1484,
+      "step": 20192
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3358618002206086e-08,
+      "loss": 1.2297,
+      "step": 20193
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.321383159470949e-08,
+      "loss": 1.1174,
+      "step": 20194
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3069286809132026e-08,
+      "loss": 1.0554,
+      "step": 20195
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.292498364897979e-08,
+      "loss": 1.069,
+      "step": 20196
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.278092211775553e-08,
+      "loss": 1.077,
+      "step": 20197
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.263710221895534e-08,
+      "loss": 1.1068,
+      "step": 20198
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2493523956068646e-08,
+      "loss": 0.8995,
+      "step": 20199
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.235018733258045e-08,
+      "loss": 1.0705,
+      "step": 20200
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.220709235196907e-08,
+      "loss": 1.0439,
+      "step": 20201
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.206423901770618e-08,
+      "loss": 1.0969,
+      "step": 20202
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1921627333260104e-08,
+      "loss": 1.1844,
+      "step": 20203
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1779257302089204e-08,
+      "loss": 1.0665,
+      "step": 20204
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.16371289276507e-08,
+      "loss": 1.1907,
+      "step": 20205
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.149524221339185e-08,
+      "loss": 1.0388,
+      "step": 20206
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1353597162756555e-08,
+      "loss": 1.2219,
+      "step": 20207
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1212193779182065e-08,
+      "loss": 1.0081,
+      "step": 20208
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.107103206610119e-08,
+      "loss": 1.0604,
+      "step": 20209
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.093011202693675e-08,
+      "loss": 1.1574,
+      "step": 20210
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.078943366511046e-08,
+      "loss": 1.1201,
+      "step": 20211
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.064899698403513e-08,
+      "loss": 1.0347,
+      "step": 20212
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.050880198712026e-08,
+      "loss": 1.097,
+      "step": 20213
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.036884867776647e-08,
+      "loss": 1.1624,
+      "step": 20214
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.0229137059369925e-08,
+      "loss": 1.1548,
+      "step": 20215
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.008966713532236e-08,
+      "loss": 1.0078,
+      "step": 20216
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.995043890900663e-08,
+      "loss": 1.246,
+      "step": 20217
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.981145238380335e-08,
+      "loss": 0.9952,
+      "step": 20218
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.967270756308428e-08,
+      "loss": 0.9939,
+      "step": 20219
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.953420445021561e-08,
+      "loss": 1.1185,
+      "step": 20220
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.939594304855909e-08,
+      "loss": 1.073,
+      "step": 20221
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9257923361469826e-08,
+      "loss": 1.1218,
+      "step": 20222
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.912014539229847e-08,
+      "loss": 1.145,
+      "step": 20223
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8982609144385676e-08,
+      "loss": 1.163,
+      "step": 20224
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.884531462107211e-08,
+      "loss": 0.9334,
+      "step": 20225
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.870826182568732e-08,
+      "loss": 1.1306,
+      "step": 20226
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.857145076155755e-08,
+      "loss": 1.0178,
+      "step": 20227
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.843488143200347e-08,
+      "loss": 1.0885,
+      "step": 20228
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8298553840339095e-08,
+      "loss": 1.0403,
+      "step": 20229
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8162467989872885e-08,
+      "loss": 1.1375,
+      "step": 20230
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8026623883906656e-08,
+      "loss": 1.0063,
+      "step": 20231
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.789102152573665e-08,
+      "loss": 1.1668,
+      "step": 20232
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.775566091865468e-08,
+      "loss": 0.9321,
+      "step": 20233
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.762054206594479e-08,
+      "loss": 1.0866,
+      "step": 20234
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.748566497088546e-08,
+      "loss": 0.9924,
+      "step": 20235
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7351029636750745e-08,
+      "loss": 1.1228,
+      "step": 20236
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.721663606680692e-08,
+      "loss": 1.1191,
+      "step": 20237
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.708248426431693e-08,
+      "loss": 1.2053,
+      "step": 20238
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.694857423253373e-08,
+      "loss": 1.1948,
+      "step": 20239
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.681490597471027e-08,
+      "loss": 1.0948,
+      "step": 20240
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.668147949408729e-08,
+      "loss": 0.9936,
+      "step": 20241
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6548294793903314e-08,
+      "loss": 1.0989,
+      "step": 20242
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.641535187739131e-08,
+      "loss": 1.0259,
+      "step": 20243
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6282650747776484e-08,
+      "loss": 1.1151,
+      "step": 20244
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6150191408279576e-08,
+      "loss": 0.9849,
+      "step": 20245
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6017973862115805e-08,
+      "loss": 1.0477,
+      "step": 20246
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5885998112491495e-08,
+      "loss": 1.1826,
+      "step": 20247
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.575426416261074e-08,
+      "loss": 1.0889,
+      "step": 20248
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.562277201567099e-08,
+      "loss": 0.8866,
+      "step": 20249
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5491521674861914e-08,
+      "loss": 1.0355,
+      "step": 20250
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5360513143368745e-08,
+      "loss": 1.109,
+      "step": 20251
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.522974642437116e-08,
+      "loss": 1.1182,
+      "step": 20252
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.509922152104328e-08,
+      "loss": 1.0829,
+      "step": 20253
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.496893843655036e-08,
+      "loss": 1.0866,
+      "step": 20254
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.483889717405542e-08,
+      "loss": 1.026,
+      "step": 20255
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4709097736714825e-08,
+      "loss": 1.0764,
+      "step": 20256
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.457954012767606e-08,
+      "loss": 1.044,
+      "step": 20257
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.445022435008549e-08,
+      "loss": 1.0689,
+      "step": 20258
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4321150407080615e-08,
+      "loss": 1.0862,
+      "step": 20259
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.419231830179337e-08,
+      "loss": 1.1366,
+      "step": 20260
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.406372803735014e-08,
+      "loss": 1.0088,
+      "step": 20261
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3935379616870654e-08,
+      "loss": 1.2033,
+      "step": 20262
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.380727304347131e-08,
+      "loss": 1.1655,
+      "step": 20263
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.367940832025962e-08,
+      "loss": 1.1461,
+      "step": 20264
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.355178545033866e-08,
+      "loss": 1.1424,
+      "step": 20265
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.342440443680595e-08,
+      "loss": 0.9845,
+      "step": 20266
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3297265282751237e-08,
+      "loss": 1.1363,
+      "step": 20267
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.317036799126205e-08,
+      "loss": 1.1094,
+      "step": 20268
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.304371256541594e-08,
+      "loss": 1.1292,
+      "step": 20269
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.291729900828822e-08,
+      "loss": 1.1218,
+      "step": 20270
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.2791127322944205e-08,
+      "loss": 1.0478,
+      "step": 20271
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.266519751244812e-08,
+      "loss": 1.2584,
+      "step": 20272
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.253950957985419e-08,
+      "loss": 1.1615,
+      "step": 20273
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2414063528213305e-08,
+      "loss": 0.8975,
+      "step": 20274
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.228885936056858e-08,
+      "loss": 1.1526,
+      "step": 20275
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.216389707995982e-08,
+      "loss": 1.1621,
+      "step": 20276
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2039176689420137e-08,
+      "loss": 1.0841,
+      "step": 20277
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.191469819197379e-08,
+      "loss": 1.0662,
+      "step": 20278
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1790461590642804e-08,
+      "loss": 1.0953,
+      "step": 20279
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.166646688844144e-08,
+      "loss": 1.1225,
+      "step": 20280
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.15427140883795e-08,
+      "loss": 1.1811,
+      "step": 20281
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.141920319346015e-08,
+      "loss": 0.9689,
+      "step": 20282
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.129593420667987e-08,
+      "loss": 1.1346,
+      "step": 20283
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.117290713102961e-08,
+      "loss": 1.1025,
+      "step": 20284
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.105012196949697e-08,
+      "loss": 1.0674,
+      "step": 20285
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.092757872505958e-08,
+      "loss": 1.1215,
+      "step": 20286
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0805277400691724e-08,
+      "loss": 0.997,
+      "step": 20287
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.068321799936103e-08,
+      "loss": 1.0114,
+      "step": 20288
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.056140052403067e-08,
+      "loss": 1.2325,
+      "step": 20289
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.043982497765607e-08,
+      "loss": 1.1956,
+      "step": 20290
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0318491363187096e-08,
+      "loss": 1.101,
+      "step": 20291
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.019739968356805e-08,
+      "loss": 1.0152,
+      "step": 20292
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.007654994173881e-08,
+      "loss": 1.0656,
+      "step": 20293
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9955942140631466e-08,
+      "loss": 1.1408,
+      "step": 20294
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.983557628317257e-08,
+      "loss": 1.0703,
+      "step": 20295
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9715452372282015e-08,
+      "loss": 1.1188,
+      "step": 20296
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.959557041087635e-08,
+      "loss": 1.0345,
+      "step": 20297
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9475930401864362e-08,
+      "loss": 1.1116,
+      "step": 20298
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.935653234814928e-08,
+      "loss": 1.095,
+      "step": 20299
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.923737625262879e-08,
+      "loss": 1.2118,
+      "step": 20300
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9118462118193913e-08,
+      "loss": 0.9069,
+      "step": 20301
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.899978994773123e-08,
+      "loss": 1.0127,
+      "step": 20302
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.888135974411954e-08,
+      "loss": 1.0129,
+      "step": 20303
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8763171510233224e-08,
+      "loss": 1.0714,
+      "step": 20304
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8645225248939977e-08,
+      "loss": 1.0771,
+      "step": 20305
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.852752096310307e-08,
+      "loss": 1.0129,
+      "step": 20306
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8410058655578e-08,
+      "loss": 1.2174,
+      "step": 20307
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.829283832921581e-08,
+      "loss": 1.1152,
+      "step": 20308
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8175859986859788e-08,
+      "loss": 1.1512,
+      "step": 20309
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8059123631349884e-08,
+      "loss": 0.9972,
+      "step": 20310
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7942629265519383e-08,
+      "loss": 1.1137,
+      "step": 20311
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7826376892193807e-08,
+      "loss": 0.9328,
+      "step": 20312
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7710366514195343e-08,
+      "loss": 0.941,
+      "step": 20313
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.75945981343384e-08,
+      "loss": 1.0564,
+      "step": 20314
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7479071755431853e-08,
+      "loss": 1.2178,
+      "step": 20315
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7363787380281226e-08,
+      "loss": 1.0074,
+      "step": 20316
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7248745011683175e-08,
+      "loss": 1.1364,
+      "step": 20317
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7133944652429912e-08,
+      "loss": 1.182,
+      "step": 20318
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7019386305304763e-08,
+      "loss": 1.1223,
+      "step": 20319
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6905069973091058e-08,
+      "loss": 1.0495,
+      "step": 20320
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6790995658561026e-08,
+      "loss": 0.9579,
+      "step": 20321
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.667716336448356e-08,
+      "loss": 0.9386,
+      "step": 20322
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.65635730936209e-08,
+      "loss": 1.0574,
+      "step": 20323
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6450224848728612e-08,
+      "loss": 1.1292,
+      "step": 20324
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6337118632558945e-08,
+      "loss": 1.2929,
+      "step": 20325
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6224254447856368e-08,
+      "loss": 1.0262,
+      "step": 20326
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6111632297358693e-08,
+      "loss": 1.1043,
+      "step": 20327
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5999252183799285e-08,
+      "loss": 1.1427,
+      "step": 20328
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5887114109907073e-08,
+      "loss": 0.9689,
+      "step": 20329
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5775218078399888e-08,
+      "loss": 1.0936,
+      "step": 20330
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.566356409199555e-08,
+      "loss": 1.2116,
+      "step": 20331
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.555215215340412e-08,
+      "loss": 1.1383,
+      "step": 20332
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5440982265326764e-08,
+      "loss": 1.0044,
+      "step": 20333
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5330054430463547e-08,
+      "loss": 1.1188,
+      "step": 20334
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5219368651504537e-08,
+      "loss": 1.1194,
+      "step": 20335
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.510892493113759e-08,
+      "loss": 0.9346,
+      "step": 20336
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.499872327204167e-08,
+      "loss": 1.0614,
+      "step": 20337
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4888763676891305e-08,
+      "loss": 1.161,
+      "step": 20338
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4779046148354358e-08,
+      "loss": 1.1758,
+      "step": 20339
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4669570689094258e-08,
+      "loss": 1.1119,
+      "step": 20340
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4560337301767767e-08,
+      "loss": 1.1634,
+      "step": 20341
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4451345989024987e-08,
+      "loss": 1.1482,
+      "step": 20342
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4342596753510473e-08,
+      "loss": 1.0363,
+      "step": 20343
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.423408959786433e-08,
+      "loss": 1.0387,
+      "step": 20344
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.41258245247189e-08,
+      "loss": 1.0718,
+      "step": 20345
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4017801536702078e-08,
+      "loss": 1.1773,
+      "step": 20346
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3910020636433994e-08,
+      "loss": 1.097,
+      "step": 20347
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3802481826532554e-08,
+      "loss": 1.1256,
+      "step": 20348
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.369518510960456e-08,
+      "loss": 1.0368,
+      "step": 20349
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.35881304882557e-08,
+      "loss": 1.0716,
+      "step": 20350
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3481317965083905e-08,
+      "loss": 1.1545,
+      "step": 20351
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.337474754267932e-08,
+      "loss": 1.1192,
+      "step": 20352
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.326841922363099e-08,
+      "loss": 1.0322,
+      "step": 20353
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3162333010516847e-08,
+      "loss": 1.2731,
+      "step": 20354
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.30564889059115e-08,
+      "loss": 1.0913,
+      "step": 20355
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2950886912385117e-08,
+      "loss": 1.0242,
+      "step": 20356
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.284552703249787e-08,
+      "loss": 1.1319,
+      "step": 20357
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2740409268808826e-08,
+      "loss": 1.0776,
+      "step": 20358
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2635533623868166e-08,
+      "loss": 1.0889,
+      "step": 20359
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.253090010022052e-08,
+      "loss": 1.0555,
+      "step": 20360
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.242650870040497e-08,
+      "loss": 1.1583,
+      "step": 20361
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2322359426955042e-08,
+      "loss": 1.0857,
+      "step": 20362
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2218452282398716e-08,
+      "loss": 1.0737,
+      "step": 20363
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2114787269256198e-08,
+      "loss": 1.0581,
+      "step": 20364
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2011364390044367e-08,
+      "loss": 1.0461,
+      "step": 20365
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.190818364727232e-08,
+      "loss": 1.2346,
+      "step": 20366
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.180524504344472e-08,
+      "loss": 1.1787,
+      "step": 20367
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1702548581058468e-08,
+      "loss": 1.1613,
+      "step": 20368
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.160009426260712e-08,
+      "loss": 1.1473,
+      "step": 20369
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.149788209057424e-08,
+      "loss": 0.9499,
+      "step": 20370
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1395912067443404e-08,
+      "loss": 1.1577,
+      "step": 20371
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1294184195688183e-08,
+      "loss": 1.0067,
+      "step": 20372
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1192698477775498e-08,
+      "loss": 1.0509,
+      "step": 20373
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.109145491617004e-08,
+      "loss": 1.1591,
+      "step": 20374
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0990453513327625e-08,
+      "loss": 1.0456,
+      "step": 20375
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0889694271700732e-08,
+      "loss": 1.0948,
+      "step": 20376
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0789177193732967e-08,
+      "loss": 1.221,
+      "step": 20377
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0688902281863487e-08,
+      "loss": 1.1422,
+      "step": 20378
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0588869538527012e-08,
+      "loss": 1.0781,
+      "step": 20379
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0489078966149378e-08,
+      "loss": 1.0685,
+      "step": 20380
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.03895305671542e-08,
+      "loss": 1.0845,
+      "step": 20381
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0290224343955113e-08,
+      "loss": 1.1506,
+      "step": 20382
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0191160298963508e-08,
+      "loss": 1.2179,
+      "step": 20383
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0092338434583026e-08,
+      "loss": 1.1903,
+      "step": 20384
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9993758753211746e-08,
+      "loss": 1.2338,
+      "step": 20385
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9895421257241086e-08,
+      "loss": 1.1175,
+      "step": 20386
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9797325949058032e-08,
+      "loss": 0.912,
+      "step": 20387
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9699472831044007e-08,
+      "loss": 1.087,
+      "step": 20388
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9601861905572673e-08,
+      "loss": 1.0731,
+      "step": 20389
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.950449317501213e-08,
+      "loss": 1.0513,
+      "step": 20390
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9407366641727153e-08,
+      "loss": 1.0767,
+      "step": 20391
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.931048230807253e-08,
+      "loss": 1.2466,
+      "step": 20392
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9213840176400823e-08,
+      "loss": 1.1488,
+      "step": 20393
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9117440249055708e-08,
+      "loss": 1.1162,
+      "step": 20394
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.902128252837865e-08,
+      "loss": 0.9703,
+      "step": 20395
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8925367016702223e-08,
+      "loss": 1.0898,
+      "step": 20396
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8829693716353457e-08,
+      "loss": 1.2024,
+      "step": 20397
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.873426262965383e-08,
+      "loss": 1.1067,
+      "step": 20398
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.863907375892149e-08,
+      "loss": 1.1496,
+      "step": 20399
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8544127106463473e-08,
+      "loss": 1.1418,
+      "step": 20400
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8449422674585715e-08,
+      "loss": 0.7578,
+      "step": 20401
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.835496046558527e-08,
+      "loss": 1.1063,
+      "step": 20402
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8260740481754747e-08,
+      "loss": 1.0936,
+      "step": 20403
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8166762725381205e-08,
+      "loss": 1.044,
+      "step": 20404
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8073027198745042e-08,
+      "loss": 0.9837,
+      "step": 20405
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7979533904119993e-08,
+      "loss": 0.8327,
+      "step": 20406
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.788628284377647e-08,
+      "loss": 1.178,
+      "step": 20407
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7793274019975992e-08,
+      "loss": 1.1346,
+      "step": 20408
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7700507434975644e-08,
+      "loss": 1.2708,
+      "step": 20409
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.760798309102807e-08,
+      "loss": 1.1522,
+      "step": 20410
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7515700990377028e-08,
+      "loss": 1.1287,
+      "step": 20411
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7423661135262948e-08,
+      "loss": 0.9269,
+      "step": 20412
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.733186352791849e-08,
+      "loss": 0.9706,
+      "step": 20413
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7240308170571875e-08,
+      "loss": 1.1779,
+      "step": 20414
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7148995065444653e-08,
+      "loss": 1.0958,
+      "step": 20415
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.705792421475283e-08,
+      "loss": 1.0315,
+      "step": 20416
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6967095620705753e-08,
+      "loss": 1.0865,
+      "step": 20417
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6876509285509434e-08,
+      "loss": 1.2312,
+      "step": 20418
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6786165211359895e-08,
+      "loss": 1.1984,
+      "step": 20419
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6696063400450936e-08,
+      "loss": 1.1234,
+      "step": 20420
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6606203854969694e-08,
+      "loss": 1.057,
+      "step": 20421
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6516586577094428e-08,
+      "loss": 0.9927,
+      "step": 20422
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6427211569001178e-08,
+      "loss": 0.9421,
+      "step": 20423
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6338078832860427e-08,
+      "loss": 1.0436,
+      "step": 20424
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6249188370832668e-08,
+      "loss": 0.9926,
+      "step": 20425
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6160540185075068e-08,
+      "loss": 1.0503,
+      "step": 20426
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6072134277740346e-08,
+      "loss": 1.1545,
+      "step": 20427
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5983970650973457e-08,
+      "loss": 1.1752,
+      "step": 20428
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5896049306913797e-08,
+      "loss": 1.1695,
+      "step": 20429
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5808370247695215e-08,
+      "loss": 1.1484,
+      "step": 20430
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.572093347544379e-08,
+      "loss": 1.0705,
+      "step": 20431
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5633738992283377e-08,
+      "loss": 0.9505,
+      "step": 20432
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.554678680033006e-08,
+      "loss": 1.0606,
+      "step": 20433
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5460076901691046e-08,
+      "loss": 1.1161,
+      "step": 20434
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5373609298473536e-08,
+      "loss": 1.1797,
+      "step": 20435
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.528738399277474e-08,
+      "loss": 1.0581,
+      "step": 20436
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5201400986687432e-08,
+      "loss": 0.9342,
+      "step": 20437
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5115660282297717e-08,
+      "loss": 0.9954,
+      "step": 20438
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5030161881686155e-08,
+      "loss": 1.1619,
+      "step": 20439
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4944905786927756e-08,
+      "loss": 1.0048,
+      "step": 20440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4859892000091968e-08,
+      "loss": 0.8995,
+      "step": 20441
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4775120523241593e-08,
+      "loss": 1.0938,
+      "step": 20442
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.469059135843276e-08,
+      "loss": 1.0325,
+      "step": 20443
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4606304507718271e-08,
+      "loss": 1.1768,
+      "step": 20444
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.452225997314205e-08,
+      "loss": 1.1162,
+      "step": 20445
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4438457756744683e-08,
+      "loss": 1.0143,
+      "step": 20446
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4354897860558992e-08,
+      "loss": 0.9842,
+      "step": 20447
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4271580286614461e-08,
+      "loss": 1.1079,
+      "step": 20448
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4188505036930588e-08,
+      "loss": 1.0881,
+      "step": 20449
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4105672113524648e-08,
+      "loss": 1.1803,
+      "step": 20450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4023081518406145e-08,
+      "loss": 1.1329,
+      "step": 20451
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3940733253581251e-08,
+      "loss": 1.0806,
+      "step": 20452
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3858627321045037e-08,
+      "loss": 1.2046,
+      "step": 20453
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3776763722792574e-08,
+      "loss": 1.1764,
+      "step": 20454
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.369514246081005e-08,
+      "loss": 1.1653,
+      "step": 20455
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3613763537078106e-08,
+      "loss": 0.9221,
+      "step": 20456
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3532626953569606e-08,
+      "loss": 1.2131,
+      "step": 20457
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3451732712256305e-08,
+      "loss": 1.1518,
+      "step": 20458
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3371080815099969e-08,
+      "loss": 1.0738,
+      "step": 20459
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.329067126405792e-08,
+      "loss": 1.1122,
+      "step": 20460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.321050406108082e-08,
+      "loss": 1.13,
+      "step": 20461
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.313057920811489e-08,
+      "loss": 1.0839,
+      "step": 20462
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3050896707099692e-08,
+      "loss": 1.228,
+      "step": 20463
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.297145655996812e-08,
+      "loss": 1.1281,
+      "step": 20464
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2892258768649746e-08,
+      "loss": 1.1817,
+      "step": 20465
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2813303335064143e-08,
+      "loss": 1.0249,
+      "step": 20466
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2734590261128665e-08,
+      "loss": 1.1002,
+      "step": 20467
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2656119548752898e-08,
+      "loss": 1.1858,
+      "step": 20468
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2577891199841986e-08,
+      "loss": 0.8523,
+      "step": 20469
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2499905216293295e-08,
+      "loss": 1.1979,
+      "step": 20470
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.242216159999976e-08,
+      "loss": 1.1588,
+      "step": 20471
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2344660352848758e-08,
+      "loss": 1.2537,
+      "step": 20472
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2267401476718787e-08,
+      "loss": 1.1146,
+      "step": 20473
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2190384973487234e-08,
+      "loss": 1.0789,
+      "step": 20474
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2113610845021495e-08,
+      "loss": 0.9435,
+      "step": 20475
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2037079093185632e-08,
+      "loss": 1.0976,
+      "step": 20476
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1960789719835941e-08,
+      "loss": 0.9522,
+      "step": 20477
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1884742726824272e-08,
+      "loss": 0.9812,
+      "step": 20478
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1808938115995815e-08,
+      "loss": 1.0844,
+      "step": 20479
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1733375889189102e-08,
+      "loss": 1.0009,
+      "step": 20480
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1658056048240441e-08,
+      "loss": 1.1554,
+      "step": 20481
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1582978594975037e-08,
+      "loss": 0.9985,
+      "step": 20482
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1508143531216986e-08,
+      "loss": 1.1729,
+      "step": 20483
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1433550858780396e-08,
+      "loss": 1.0593,
+      "step": 20484
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1359200579476038e-08,
+      "loss": 1.1604,
+      "step": 20485
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1285092695109134e-08,
+      "loss": 1.0458,
+      "step": 20486
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1211227207477139e-08,
+      "loss": 1.1218,
+      "step": 20487
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1137604118371947e-08,
+      "loss": 1.0341,
+      "step": 20488
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1064223429581022e-08,
+      "loss": 1.1246,
+      "step": 20489
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.099108514288627e-08,
+      "loss": 1.1405,
+      "step": 20490
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0918189260059609e-08,
+      "loss": 1.1069,
+      "step": 20491
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0845535782872952e-08,
+      "loss": 1.1739,
+      "step": 20492
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0773124713087112e-08,
+      "loss": 1.0816,
+      "step": 20493
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0700956052460686e-08,
+      "loss": 1.0454,
+      "step": 20494
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0629029802744494e-08,
+      "loss": 0.9505,
+      "step": 20495
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0557345965683808e-08,
+      "loss": 1.0874,
+      "step": 20496
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0485904543019453e-08,
+      "loss": 1.0384,
+      "step": 20497
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0414705536483383e-08,
+      "loss": 1.0038,
+      "step": 20498
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0343748947804211e-08,
+      "loss": 1.1815,
+      "step": 20499
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0273034778703894e-08,
+      "loss": 0.9261,
+      "step": 20500
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0202563030897727e-08,
+      "loss": 1.1658,
+      "step": 20501
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0132333706096564e-08,
+      "loss": 0.8873,
+      "step": 20502
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0062346806004597e-08,
+      "loss": 0.9988,
+      "step": 20503
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.992602332320467e-09,
+      "loss": 1.0138,
+      "step": 20504
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.923100286736153e-09,
+      "loss": 1.1197,
+      "step": 20505
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.853840670938087e-09,
+      "loss": 1.0757,
+      "step": 20506
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.784823486608253e-09,
+      "loss": 1.0027,
+      "step": 20507
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.716048735419758e-09,
+      "loss": 1.2458,
+      "step": 20508
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.64751641904238e-09,
+      "loss": 1.1698,
+      "step": 20509
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.57922653913923e-09,
+      "loss": 1.2043,
+      "step": 20510
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.511179097367872e-09,
+      "loss": 1.0725,
+      "step": 20511
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.44337409537921e-09,
+      "loss": 1.1649,
+      "step": 20512
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.375811534818591e-09,
+      "loss": 1.0609,
+      "step": 20513
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.308491417324705e-09,
+      "loss": 1.1919,
+      "step": 20514
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.241413744532912e-09,
+      "loss": 1.0542,
+      "step": 20515
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.174578518070799e-09,
+      "loss": 1.1699,
+      "step": 20516
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.107985739559289e-09,
+      "loss": 1.058,
+      "step": 20517
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.041635410614868e-09,
+      "loss": 1.0182,
+      "step": 20518
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.97552753284736e-09,
+      "loss": 1.1939,
+      "step": 20519
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.90966210786215e-09,
+      "loss": 1.1159,
+      "step": 20520
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.844039137255734e-09,
+      "loss": 1.1477,
+      "step": 20521
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.778658622621283e-09,
+      "loss": 1.1556,
+      "step": 20522
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.713520565546418e-09,
+      "loss": 1.0235,
+      "step": 20523
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.648624967610985e-09,
+      "loss": 1.1003,
+      "step": 20524
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.583971830389281e-09,
+      "loss": 1.2478,
+      "step": 20525
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.519561155451161e-09,
+      "loss": 1.0079,
+      "step": 20526
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.45539294435982e-09,
+      "loss": 1.078,
+      "step": 20527
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.391467198671788e-09,
+      "loss": 1.0636,
+      "step": 20528
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.327783919939158e-09,
+      "loss": 1.1219,
+      "step": 20529
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.264343109706252e-09,
+      "loss": 1.1839,
+      "step": 20530
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.201144769512948e-09,
+      "loss": 0.973,
+      "step": 20531
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.138188900892464e-09,
+      "loss": 0.9768,
+      "step": 20532
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.075475505373576e-09,
+      "loss": 1.0308,
+      "step": 20533
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.013004584478401e-09,
+      "loss": 1.087,
+      "step": 20534
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.950776139722394e-09,
+      "loss": 1.0516,
+      "step": 20535
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.888790172614347e-09,
+      "loss": 1.0664,
+      "step": 20536
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.827046684659723e-09,
+      "loss": 1.1349,
+      "step": 20537
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.765545677357323e-09,
+      "loss": 1.0377,
+      "step": 20538
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.704287152199285e-09,
+      "loss": 0.962,
+      "step": 20539
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.643271110671091e-09,
+      "loss": 1.0577,
+      "step": 20540
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.582497554254887e-09,
+      "loss": 1.0861,
+      "step": 20541
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.521966484423936e-09,
+      "loss": 1.1402,
+      "step": 20542
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.46167790264929e-09,
+      "loss": 1.082,
+      "step": 20543
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.401631810391996e-09,
+      "loss": 1.0806,
+      "step": 20544
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.34182820910978e-09,
+      "loss": 1.0917,
+      "step": 20545
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.282267100254814e-09,
+      "loss": 1.09,
+      "step": 20546
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.222948485271497e-09,
+      "loss": 1.0701,
+      "step": 20547
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.163872365598679e-09,
+      "loss": 1.1764,
+      "step": 20548
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.105038742670767e-09,
+      "loss": 1.0766,
+      "step": 20549
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.046447617916618e-09,
+      "loss": 1.0423,
+      "step": 20550
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.988098992756209e-09,
+      "loss": 1.1817,
+      "step": 20551
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.929992868606184e-09,
+      "loss": 1.2364,
+      "step": 20552
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.872129246877635e-09,
+      "loss": 1.3125,
+      "step": 20553
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.814508128972774e-09,
+      "loss": 1.1167,
+      "step": 20554
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.757129516291594e-09,
+      "loss": 0.9345,
+      "step": 20555
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.699993410225203e-09,
+      "loss": 1.2332,
+      "step": 20556
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.64309981216138e-09,
+      "loss": 1.1149,
+      "step": 20557
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.586448723479022e-09,
+      "loss": 1.062,
+      "step": 20558
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.530040145554806e-09,
+      "loss": 1.0298,
+      "step": 20559
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.473874079756526e-09,
+      "loss": 1.062,
+      "step": 20560
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.417950527447536e-09,
+      "loss": 1.1737,
+      "step": 20561
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.3622694899845294e-09,
+      "loss": 0.9851,
+      "step": 20562
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.306830968719757e-09,
+      "loss": 1.1315,
+      "step": 20563
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.251634964997699e-09,
+      "loss": 1.1417,
+      "step": 20564
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.196681480157285e-09,
+      "loss": 1.1346,
+      "step": 20565
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.141970515533002e-09,
+      "loss": 1.1789,
+      "step": 20566
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.0875020724515675e-09,
+      "loss": 1.1292,
+      "step": 20567
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.033276152236367e-09,
+      "loss": 1.112,
+      "step": 20568
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.979292756201904e-09,
+      "loss": 1.11,
+      "step": 20569
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.925551885659353e-09,
+      "loss": 1.0509,
+      "step": 20570
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.872053541911005e-09,
+      "loss": 1.2618,
+      "step": 20571
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.818797726256931e-09,
+      "loss": 0.9608,
+      "step": 20572
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.76578443998832e-09,
+      "loss": 0.9507,
+      "step": 20573
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.7130136843930315e-09,
+      "loss": 1.0883,
+      "step": 20574
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.6604854607500425e-09,
+      "loss": 1.173,
+      "step": 20575
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.608199770334999e-09,
+      "loss": 1.0634,
+      "step": 20576
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.556156614415775e-09,
+      "loss": 1.0425,
+      "step": 20577
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.504355994256916e-09,
+      "loss": 0.978,
+      "step": 20578
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.4527979111140824e-09,
+      "loss": 1.0946,
+      "step": 20579
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.4014823662396074e-09,
+      "loss": 1.132,
+      "step": 20580
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.35040936087694e-09,
+      "loss": 0.9024,
+      "step": 20581
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.2995788962684204e-09,
+      "loss": 1.0183,
+      "step": 20582
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.2489909736441744e-09,
+      "loss": 1.0362,
+      "step": 20583
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.1986455942332205e-09,
+      "loss": 1.1076,
+      "step": 20584
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.148542759259023e-09,
+      "loss": 1.0085,
+      "step": 20585
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0986824699350565e-09,
+      "loss": 1.1334,
+      "step": 20586
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.049064727471464e-09,
+      "loss": 1.0667,
+      "step": 20587
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.999689533073948e-09,
+      "loss": 1.1648,
+      "step": 20588
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.950556887939329e-09,
+      "loss": 1.1707,
+      "step": 20589
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.901666793261095e-09,
+      "loss": 1.1244,
+      "step": 20590
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.853019250223856e-09,
+      "loss": 1.1149,
+      "step": 20591
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.804614260008889e-09,
+      "loss": 1.2155,
+      "step": 20592
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.756451823791919e-09,
+      "loss": 1.16,
+      "step": 20593
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.708531942740902e-09,
+      "loss": 0.9993,
+      "step": 20594
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.660854618019351e-09,
+      "loss": 1.0607,
+      "step": 20595
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.613419850781897e-09,
+      "loss": 0.964,
+      "step": 20596
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.566227642182064e-09,
+      "loss": 1.175,
+      "step": 20597
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.5192779933656e-09,
+      "loss": 1.1937,
+      "step": 20598
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.4725709054693756e-09,
+      "loss": 0.956,
+      "step": 20599
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.4261063796280365e-09,
+      "loss": 1.2197,
+      "step": 20600
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.379884416969571e-09,
+      "loss": 0.9578,
+      "step": 20601
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.333905018615303e-09,
+      "loss": 1.028,
+      "step": 20602
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.288168185681007e-09,
+      "loss": 1.1092,
+      "step": 20603
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.242673919275797e-09,
+      "loss": 1.2324,
+      "step": 20604
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.197422220505453e-09,
+      "loss": 1.2116,
+      "step": 20605
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.152413090465768e-09,
+      "loss": 1.0299,
+      "step": 20606
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.107646530251419e-09,
+      "loss": 1.2493,
+      "step": 20607
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.063122540947095e-09,
+      "loss": 1.0187,
+      "step": 20608
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.018841123634154e-09,
+      "loss": 1.0527,
+      "step": 20609
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.974802279386181e-09,
+      "loss": 0.9841,
+      "step": 20610
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.93100600927343e-09,
+      "loss": 1.1897,
+      "step": 20611
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.887452314357276e-09,
+      "loss": 0.7584,
+      "step": 20612
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8441411956946505e-09,
+      "loss": 1.109,
+      "step": 20613
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.801072654338045e-09,
+      "loss": 1.1509,
+      "step": 20614
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.758246691331069e-09,
+      "loss": 1.0922,
+      "step": 20615
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.7156633077140012e-09,
+      "loss": 1.0776,
+      "step": 20616
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.67332250451935e-09,
+      "loss": 1.1298,
+      "step": 20617
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.631224282775181e-09,
+      "loss": 1.1512,
+      "step": 20618
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5893686435028995e-09,
+      "loss": 1.0842,
+      "step": 20619
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.54775558771836e-09,
+      "loss": 1.1126,
+      "step": 20620
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.506385116430755e-09,
+      "loss": 1.0567,
+      "step": 20621
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.465257230644836e-09,
+      "loss": 1.0815,
+      "step": 20622
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.424371931358694e-09,
+      "loss": 1.0209,
+      "step": 20623
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3837292195637583e-09,
+      "loss": 1.1767,
+      "step": 20624
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3433290962470166e-09,
+      "loss": 1.0685,
+      "step": 20625
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.303171562388796e-09,
+      "loss": 1.0565,
+      "step": 20626
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2632566189627622e-09,
+      "loss": 1.1003,
+      "step": 20627
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.22358426693814e-09,
+      "loss": 1.0865,
+      "step": 20628
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1841545072786027e-09,
+      "loss": 1.0872,
+      "step": 20629
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1449673409389425e-09,
+      "loss": 0.9764,
+      "step": 20630
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1060227688717303e-09,
+      "loss": 1.0631,
+      "step": 20631
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0673207920217664e-09,
+      "loss": 1.2196,
+      "step": 20632
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0288614113282987e-09,
+      "loss": 0.9948,
+      "step": 20633
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9906446277239154e-09,
+      "loss": 1.1122,
+      "step": 20634
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9526704421356523e-09,
+      "loss": 0.9914,
+      "step": 20635
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9149388554861045e-09,
+      "loss": 1.0728,
+      "step": 20636
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8774498686912066e-09,
+      "loss": 1.058,
+      "step": 20637
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.840203482660231e-09,
+      "loss": 1.1471,
+      "step": 20638
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.803199698296899e-09,
+      "loss": 1.1516,
+      "step": 20639
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.766438516498271e-09,
+      "loss": 0.9822,
+      "step": 20640
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7299199381580766e-09,
+      "loss": 1.098,
+      "step": 20641
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6936439641622737e-09,
+      "loss": 1.1544,
+      "step": 20642
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.657610595390159e-09,
+      "loss": 1.0888,
+      "step": 20643
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.621819832716588e-09,
+      "loss": 1.2309,
+      "step": 20644
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5862716770108654e-09,
+      "loss": 1.1368,
+      "step": 20645
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5509661291345245e-09,
+      "loss": 1.0411,
+      "step": 20646
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5159031899446574e-09,
+      "loss": 1.0815,
+      "step": 20647
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4810828602916947e-09,
+      "loss": 1.1095,
+      "step": 20648
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4465051410227368e-09,
+      "loss": 0.9664,
+      "step": 20649
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.412170032973782e-09,
+      "loss": 1.1987,
+      "step": 20650
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3780775369808275e-09,
+      "loss": 1.0862,
+      "step": 20651
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.344227653868769e-09,
+      "loss": 1.1277,
+      "step": 20652
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.310620384461393e-09,
+      "loss": 1.1201,
+      "step": 20653
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.277255729572492e-09,
+      "loss": 1.0254,
+      "step": 20654
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2441336900125288e-09,
+      "loss": 0.9297,
+      "step": 20655
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2112542665853054e-09,
+      "loss": 1.1226,
+      "step": 20656
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.178617460087962e-09,
+      "loss": 1.1155,
+      "step": 20657
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1462232713143072e-09,
+      "loss": 1.1377,
+      "step": 20658
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1140717010481595e-09,
+      "loss": 1.0761,
+      "step": 20659
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.082162750071115e-09,
+      "loss": 1.0656,
+      "step": 20660
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0504964191569996e-09,
+      "loss": 1.1147,
+      "step": 20661
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.019072709074088e-09,
+      "loss": 1.2363,
+      "step": 20662
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.987891620585103e-09,
+      "loss": 1.1173,
+      "step": 20663
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9569531544472163e-09,
+      "loss": 1.1402,
+      "step": 20664
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9262573114109395e-09,
+      "loss": 1.1031,
+      "step": 20665
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8958040922212316e-09,
+      "loss": 1.0924,
+      "step": 20666
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8655934976175017e-09,
+      "loss": 1.0414,
+      "step": 20667
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8356255283313862e-09,
+      "loss": 1.126,
+      "step": 20668
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8059001850911917e-09,
+      "loss": 1.0392,
+      "step": 20669
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7764174686185631e-09,
+      "loss": 1.0914,
+      "step": 20670
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.747177379628484e-09,
+      "loss": 1.1344,
+      "step": 20671
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7181799188314974e-09,
+      "loss": 1.2585,
+      "step": 20672
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.689425086929264e-09,
+      "loss": 1.0547,
+      "step": 20673
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6609128846212242e-09,
+      "loss": 1.2404,
+      "step": 20674
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6326433125979368e-09,
+      "loss": 0.7651,
+      "step": 20675
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.60461637154663e-09,
+      "loss": 1.1785,
+      "step": 20676
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5768320621478705e-09,
+      "loss": 1.0319,
+      "step": 20677
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5492903850744534e-09,
+      "loss": 1.0237,
+      "step": 20678
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5219913409947329e-09,
+      "loss": 1.0923,
+      "step": 20679
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4949349305726224e-09,
+      "loss": 1.1635,
+      "step": 20680
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4681211544642638e-09,
+      "loss": 1.1166,
+      "step": 20681
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4415500133191374e-09,
+      "loss": 1.2263,
+      "step": 20682
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.415221507783393e-09,
+      "loss": 1.1083,
+      "step": 20683
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3891356384954092e-09,
+      "loss": 0.9972,
+      "step": 20684
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3632924060880126e-09,
+      "loss": 1.0704,
+      "step": 20685
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3376918111884796e-09,
+      "loss": 1.1604,
+      "step": 20686
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.312333854418535e-09,
+      "loss": 1.1818,
+      "step": 20687
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.287218536393242e-09,
+      "loss": 1.0079,
+      "step": 20688
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2623458577210035e-09,
+      "loss": 1.1559,
+      "step": 20689
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2377158190068906e-09,
+      "loss": 0.9778,
+      "step": 20690
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2133284208493135e-09,
+      "loss": 1.0464,
+      "step": 20691
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1891836638378008e-09,
+      "loss": 1.1959,
+      "step": 20692
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1652815485585501e-09,
+      "loss": 1.0015,
+      "step": 20693
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1416220755933183e-09,
+      "loss": 1.0557,
+      "step": 20694
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1182052455160908e-09,
+      "loss": 1.1213,
+      "step": 20695
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0950310588930812e-09,
+      "loss": 1.0918,
+      "step": 20696
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.072099516288283e-09,
+      "loss": 1.2027,
+      "step": 20697
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.049410618259028e-09,
+      "loss": 1.0351,
+      "step": 20698
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0269643653526563e-09,
+      "loss": 1.1151,
+      "step": 20699
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0047607581176178e-09,
+      "loss": 1.0211,
+      "step": 20700
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.827997970901503e-10,
+      "loss": 0.9794,
+      "step": 20701
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.610814828042715e-10,
+      "loss": 1.169,
+      "step": 20702
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.396058157873367e-10,
+      "loss": 1.0964,
+      "step": 20703
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.183727965600409e-10,
+      "loss": 1.0212,
+      "step": 20704
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.973824256364172e-10,
+      "loss": 1.0755,
+      "step": 20705
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.766347035282785e-10,
+      "loss": 0.983,
+      "step": 20706
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.561296307374456e-10,
+      "loss": 1.1523,
+      "step": 20707
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.358672077612983e-10,
+      "loss": 1.0035,
+      "step": 20708
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.158474350927758e-10,
+      "loss": 1.0876,
+      "step": 20709
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.960703132170456e-10,
+      "loss": 1.1547,
+      "step": 20710
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.765358426137237e-10,
+      "loss": 1.0488,
+      "step": 20711
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.572440237579859e-10,
+      "loss": 1.1219,
+      "step": 20712
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.381948571161257e-10,
+      "loss": 0.9866,
+      "step": 20713
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.193883431522163e-10,
+      "loss": 1.0754,
+      "step": 20714
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.008244823214494e-10,
+      "loss": 1.1613,
+      "step": 20715
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.825032750756855e-10,
+      "loss": 0.9932,
+      "step": 20716
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.644247218579036e-10,
+      "loss": 1.0592,
+      "step": 20717
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.465888231077522e-10,
+      "loss": 0.9378,
+      "step": 20718
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.289955792582181e-10,
+      "loss": 0.9449,
+      "step": 20719
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.11644990735627e-10,
+      "loss": 1.1034,
+      "step": 20720
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.945370579607534e-10,
+      "loss": 1.0557,
+      "step": 20721
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.77671781349931e-10,
+      "loss": 0.9511,
+      "step": 20722
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.610491613117219e-10,
+      "loss": 1.1244,
+      "step": 20723
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.44669198249137e-10,
+      "loss": 1.0793,
+      "step": 20724
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.285318925596361e-10,
+      "loss": 1.1355,
+      "step": 20725
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.126372446362383e-10,
+      "loss": 0.9225,
+      "step": 20726
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.969852548630804e-10,
+      "loss": 1.1745,
+      "step": 20727
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.815759236209694e-10,
+      "loss": 1.0388,
+      "step": 20728
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.664092512840502e-10,
+      "loss": 1.0742,
+      "step": 20729
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.514852382186963e-10,
+      "loss": 1.115,
+      "step": 20730
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.36803884789061e-10,
+      "loss": 0.9947,
+      "step": 20731
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.2236519135041566e-10,
+      "loss": 1.1861,
+      "step": 20732
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.081691582524805e-10,
+      "loss": 1.2058,
+      "step": 20733
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.942157858416451e-10,
+      "loss": 1.173,
+      "step": 20734
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.805050744543071e-10,
+      "loss": 1.0476,
+      "step": 20735
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.670370244246435e-10,
+      "loss": 1.1556,
+      "step": 20736
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.5381163607906e-10,
+      "loss": 1.0926,
+      "step": 20737
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.408289097395212e-10,
+      "loss": 1.0387,
+      "step": 20738
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.2808884571911006e-10,
+      "loss": 1.1478,
+      "step": 20739
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.1559144432868893e-10,
+      "loss": 1.0296,
+      "step": 20740
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.033367058702386e-10,
+      "loss": 1.0508,
+      "step": 20741
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.913246306412987e-10,
+      "loss": 0.9731,
+      "step": 20742
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7955521893496817e-10,
+      "loss": 1.1164,
+      "step": 20743
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.6802847103546413e-10,
+      "loss": 1.1641,
+      "step": 20744
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.567443872225628e-10,
+      "loss": 1.1353,
+      "step": 20745
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.4570296777048917e-10,
+      "loss": 1.181,
+      "step": 20746
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.349042129468071e-10,
+      "loss": 1.1725,
+      "step": 20747
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2434812301352915e-10,
+      "loss": 1.1772,
+      "step": 20748
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1403469822711686e-10,
+      "loss": 1.1891,
+      "step": 20749
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0396393883737042e-10,
+      "loss": 1.1676,
+      "step": 20750
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.941358450896491e-10,
+      "loss": 1.0245,
+      "step": 20751
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8455041722154066e-10,
+      "loss": 1.0045,
+      "step": 20752
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.752076554661919e-10,
+      "loss": 0.9953,
+      "step": 20753
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6610756005008832e-10,
+      "loss": 1.0992,
+      "step": 20754
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5725013119416432e-10,
+      "loss": 1.0653,
+      "step": 20755
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4863536911380317e-10,
+      "loss": 1.1116,
+      "step": 20756
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4026327401661654e-10,
+      "loss": 1.0407,
+      "step": 20757
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3213384610799574e-10,
+      "loss": 1.1662,
+      "step": 20758
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2424708558222976e-10,
+      "loss": 1.0839,
+      "step": 20759
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1660299263360764e-10,
+      "loss": 1.0318,
+      "step": 20760
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.092015674464264e-10,
+      "loss": 1.1072,
+      "step": 20761
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0204281019943196e-10,
+      "loss": 1.0892,
+      "step": 20762
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.512672106803955e-11,
+      "loss": 1.2234,
+      "step": 20763
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.845330021878262e-11,
+      "loss": 1.1391,
+      "step": 20764
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.202254781486397e-11,
+      "loss": 1.1085,
+      "step": 20765
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.583446401060457e-11,
+      "loss": 1.0487,
+      "step": 20766
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.9889048958105e-11,
+      "loss": 1.1421,
+      "step": 20767
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.41863027994738e-11,
+      "loss": 1.2039,
+      "step": 20768
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.872622567459907e-11,
+      "loss": 1.1348,
+      "step": 20769
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.350881771670757e-11,
+      "loss": 0.9425,
+      "step": 20770
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.853407905014429e-11,
+      "loss": 1.0807,
+      "step": 20771
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.380200979703375e-11,
+      "loss": 0.9731,
+      "step": 20772
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.931261007283915e-11,
+      "loss": 1.0669,
+      "step": 20773
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.506587998525213e-11,
+      "loss": 1.1528,
+      "step": 20774
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.106181963752342e-11,
+      "loss": 0.9673,
+      "step": 20775
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7300429127352647e-11,
+      "loss": 1.0426,
+      "step": 20776
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3781708545778102e-11,
+      "loss": 1.1549,
+      "step": 20777
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0505657978286964e-11,
+      "loss": 1.108,
+      "step": 20778
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.747227750370506e-11,
+      "loss": 1.0953,
+      "step": 20779
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4681567196417333e-11,
+      "loss": 0.8802,
+      "step": 20780
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2133527124147392e-11,
+      "loss": 1.0867,
+      "step": 20781
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.828157347957501e-12,
+      "loss": 1.2158,
+      "step": 20782
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.765457924469033e-12,
+      "loss": 0.9562,
+      "step": 20783
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.945428903642025e-12,
+      "loss": 1.0428,
+      "step": 20784
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.368070329885399e-12,
+      "loss": 0.9639,
+      "step": 20785
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.0333822409467362e-12,
+      "loss": 1.0351,
+      "step": 20786
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9413646690225053e-12,
+      "loss": 1.1027,
+      "step": 20787
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0920176418682815e-12,
+      "loss": 0.9896,
+      "step": 20788
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.853411794680796e-13,
+      "loss": 1.1189,
+      "step": 20789
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2133529514457565e-13,
+      "loss": 1.1514,
+      "step": 20790
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0,
+      "loss": 1.2172,
+      "step": 20791
+    },
+    {
+      "epoch": 1.0,
+      "step": 20791,
+      "total_flos": 1.1036824432508142e+19,
+      "train_loss": 1.1433841733583656,
+      "train_runtime": 403972.4923,
+      "train_samples_per_second": 1.647,
+      "train_steps_per_second": 0.051
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 20791,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 5000,
+  "total_flos": 1.1036824432508142e+19,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}