{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.9983052215317567,
  "eval_steps": 500,
  "global_step": 4644,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 1.4285714285714287e-07,
      "loss": 1.5165,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.8571428571428575e-07,
      "loss": 1.5417,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.285714285714286e-07,
      "loss": 1.5569,
      "step": 3
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.714285714285715e-07,
      "loss": 1.5622,
      "step": 4
    },
    {
      "epoch": 0.0,
      "learning_rate": 7.142857142857143e-07,
      "loss": 1.547,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.571428571428572e-07,
      "loss": 1.5333,
      "step": 6
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 1.513,
      "step": 7
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.142857142857143e-06,
      "loss": 1.5043,
      "step": 8
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.2857142857142856e-06,
      "loss": 1.4858,
      "step": 9
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.4285714285714286e-06,
      "loss": 1.4728,
      "step": 10
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.5714285714285714e-06,
      "loss": 1.4254,
      "step": 11
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.7142857142857145e-06,
      "loss": 1.3864,
      "step": 12
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.8571428571428573e-06,
      "loss": 1.2765,
      "step": 13
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 1.2683,
      "step": 14
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.1428571428571427e-06,
      "loss": 1.2652,
      "step": 15
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.285714285714286e-06,
      "loss": 1.204,
      "step": 16
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.428571428571429e-06,
      "loss": 1.0444,
      "step": 17
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.571428571428571e-06,
      "loss": 0.9532,
      "step": 18
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.7142857142857144e-06,
      "loss": 0.9477,
      "step": 19
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.8571428571428573e-06,
      "loss": 0.9128,
      "step": 20
    },
    {
      "epoch": 0.01,
      "learning_rate": 3e-06,
      "loss": 0.8812,
      "step": 21
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.142857142857143e-06,
      "loss": 0.8899,
      "step": 22
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.285714285714286e-06,
      "loss": 0.8595,
      "step": 23
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.428571428571429e-06,
      "loss": 0.79,
      "step": 24
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.5714285714285718e-06,
      "loss": 0.7468,
      "step": 25
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.7142857142857146e-06,
      "loss": 0.746,
      "step": 26
    },
    {
      "epoch": 0.02,
      "learning_rate": 3.857142857142858e-06,
      "loss": 0.7081,
      "step": 27
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.7229,
      "step": 28
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.1428571428571435e-06,
      "loss": 0.7201,
      "step": 29
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.2857142857142855e-06,
      "loss": 0.6939,
      "step": 30
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.428571428571429e-06,
      "loss": 0.6923,
      "step": 31
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.571428571428572e-06,
      "loss": 0.6934,
      "step": 32
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.714285714285715e-06,
      "loss": 0.719,
      "step": 33
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.857142857142858e-06,
      "loss": 0.6944,
      "step": 34
    },
    {
      "epoch": 0.02,
      "learning_rate": 5e-06,
      "loss": 0.6601,
      "step": 35
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.142857142857142e-06,
      "loss": 0.6701,
      "step": 36
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.285714285714286e-06,
      "loss": 0.6554,
      "step": 37
    },
    {
      "epoch": 0.02,
      "learning_rate": 5.428571428571429e-06,
      "loss": 0.6637,
      "step": 38
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.571428571428572e-06,
      "loss": 0.6261,
      "step": 39
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.7142857142857145e-06,
      "loss": 0.6468,
      "step": 40
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.857142857142858e-06,
      "loss": 0.6304,
      "step": 41
    },
    {
      "epoch": 0.03,
      "learning_rate": 6e-06,
      "loss": 0.6452,
      "step": 42
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.142857142857144e-06,
      "loss": 0.6445,
      "step": 43
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.285714285714286e-06,
      "loss": 0.6151,
      "step": 44
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.4285714285714295e-06,
      "loss": 0.6259,
      "step": 45
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.571428571428572e-06,
      "loss": 0.6333,
      "step": 46
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.714285714285714e-06,
      "loss": 0.6171,
      "step": 47
    },
    {
      "epoch": 0.03,
      "learning_rate": 6.857142857142858e-06,
      "loss": 0.6167,
      "step": 48
    },
    {
      "epoch": 0.03,
      "learning_rate": 7e-06,
      "loss": 0.6133,
      "step": 49
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.1428571428571436e-06,
      "loss": 0.6136,
      "step": 50
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.285714285714286e-06,
      "loss": 0.6283,
      "step": 51
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.428571428571429e-06,
      "loss": 0.6231,
      "step": 52
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.571428571428572e-06,
      "loss": 0.6077,
      "step": 53
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.714285714285716e-06,
      "loss": 0.6312,
      "step": 54
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.857142857142858e-06,
      "loss": 0.5998,
      "step": 55
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.000000000000001e-06,
      "loss": 0.623,
      "step": 56
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.142857142857143e-06,
      "loss": 0.5723,
      "step": 57
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.285714285714287e-06,
      "loss": 0.5794,
      "step": 58
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.428571428571429e-06,
      "loss": 0.621,
      "step": 59
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.571428571428571e-06,
      "loss": 0.6107,
      "step": 60
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.714285714285715e-06,
      "loss": 0.5981,
      "step": 61
    },
    {
      "epoch": 0.04,
      "learning_rate": 8.857142857142858e-06,
      "loss": 0.607,
      "step": 62
    },
    {
      "epoch": 0.04,
      "learning_rate": 9e-06,
      "loss": 0.5786,
      "step": 63
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.142857142857144e-06,
      "loss": 0.5694,
      "step": 64
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.285714285714288e-06,
      "loss": 0.592,
      "step": 65
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.42857142857143e-06,
      "loss": 0.6035,
      "step": 66
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.571428571428573e-06,
      "loss": 0.5961,
      "step": 67
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.714285714285715e-06,
      "loss": 0.5898,
      "step": 68
    },
    {
      "epoch": 0.04,
      "learning_rate": 9.857142857142859e-06,
      "loss": 0.5731,
      "step": 69
    },
    {
      "epoch": 0.05,
      "learning_rate": 1e-05,
      "loss": 0.6077,
      "step": 70
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.0142857142857143e-05,
      "loss": 0.5916,
      "step": 71
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.0285714285714285e-05,
      "loss": 0.5855,
      "step": 72
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.042857142857143e-05,
      "loss": 0.5863,
      "step": 73
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.0571428571428572e-05,
      "loss": 0.6148,
      "step": 74
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.0714285714285714e-05,
      "loss": 0.5721,
      "step": 75
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.0857142857142858e-05,
      "loss": 0.5798,
      "step": 76
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.1000000000000001e-05,
      "loss": 0.6096,
      "step": 77
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.1142857142857143e-05,
      "loss": 0.566,
      "step": 78
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.1285714285714287e-05,
      "loss": 0.5704,
      "step": 79
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.1428571428571429e-05,
      "loss": 0.5707,
      "step": 80
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.1571428571428573e-05,
      "loss": 0.5707,
      "step": 81
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.1714285714285716e-05,
      "loss": 0.5624,
      "step": 82
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.1857142857142858e-05,
      "loss": 0.5578,
      "step": 83
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.2e-05,
      "loss": 0.5642,
      "step": 84
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.2142857142857142e-05,
      "loss": 0.5774,
      "step": 85
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.2285714285714288e-05,
      "loss": 0.5866,
      "step": 86
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.242857142857143e-05,
      "loss": 0.5866,
      "step": 87
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.2571428571428572e-05,
      "loss": 0.5841,
      "step": 88
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.2714285714285715e-05,
      "loss": 0.5704,
      "step": 89
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.2857142857142859e-05,
      "loss": 0.5707,
      "step": 90
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.3000000000000001e-05,
      "loss": 0.5865,
      "step": 91
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.3142857142857145e-05,
      "loss": 0.5519,
      "step": 92
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.3285714285714287e-05,
      "loss": 0.5732,
      "step": 93
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.3428571428571429e-05,
      "loss": 0.5998,
      "step": 94
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.3571428571428574e-05,
      "loss": 0.5698,
      "step": 95
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.3714285714285716e-05,
      "loss": 0.553,
      "step": 96
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.3857142857142858e-05,
      "loss": 0.5586,
      "step": 97
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.4e-05,
      "loss": 0.5768,
      "step": 98
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.4142857142857145e-05,
      "loss": 0.5627,
      "step": 99
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.4285714285714287e-05,
      "loss": 0.5818,
      "step": 100
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.4428571428571429e-05,
      "loss": 0.5732,
      "step": 101
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.4571428571428573e-05,
      "loss": 0.5758,
      "step": 102
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.4714285714285716e-05,
      "loss": 0.5692,
      "step": 103
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.4857142857142858e-05,
      "loss": 0.5738,
      "step": 104
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.5000000000000002e-05,
      "loss": 0.5573,
      "step": 105
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.5142857142857144e-05,
      "loss": 0.5287,
      "step": 106
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.5285714285714286e-05,
      "loss": 0.5888,
      "step": 107
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.542857142857143e-05,
      "loss": 0.5888,
      "step": 108
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.5571428571428573e-05,
      "loss": 0.5507,
      "step": 109
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.5714285714285715e-05,
      "loss": 0.5513,
      "step": 110
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.5857142857142857e-05,
      "loss": 0.5749,
      "step": 111
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.6000000000000003e-05,
      "loss": 0.598,
      "step": 112
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.6142857142857145e-05,
      "loss": 0.5246,
      "step": 113
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.6285714285714287e-05,
      "loss": 0.5636,
      "step": 114
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.642857142857143e-05,
      "loss": 0.5435,
      "step": 115
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.6571428571428574e-05,
      "loss": 0.5865,
      "step": 116
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.6714285714285716e-05,
      "loss": 0.5478,
      "step": 117
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.6857142857142858e-05,
      "loss": 0.5639,
      "step": 118
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.7e-05,
      "loss": 0.5615,
      "step": 119
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.7142857142857142e-05,
      "loss": 0.558,
      "step": 120
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.7285714285714287e-05,
      "loss": 0.5402,
      "step": 121
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.742857142857143e-05,
      "loss": 0.5589,
      "step": 122
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.757142857142857e-05,
      "loss": 0.5893,
      "step": 123
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.7714285714285717e-05,
      "loss": 0.558,
      "step": 124
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.785714285714286e-05,
      "loss": 0.5624,
      "step": 125
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.8e-05,
      "loss": 0.5643,
      "step": 126
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.8142857142857146e-05,
      "loss": 0.5545,
      "step": 127
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.8285714285714288e-05,
      "loss": 0.5571,
      "step": 128
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.842857142857143e-05,
      "loss": 0.5666,
      "step": 129
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.8571428571428575e-05,
      "loss": 0.5527,
      "step": 130
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.8714285714285717e-05,
      "loss": 0.5663,
      "step": 131
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.885714285714286e-05,
      "loss": 0.5734,
      "step": 132
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9e-05,
      "loss": 0.5537,
      "step": 133
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9142857142857146e-05,
      "loss": 0.5559,
      "step": 134
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.928571428571429e-05,
      "loss": 0.582,
      "step": 135
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.942857142857143e-05,
      "loss": 0.5588,
      "step": 136
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9571428571428572e-05,
      "loss": 0.5578,
      "step": 137
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9714285714285718e-05,
      "loss": 0.5592,
      "step": 138
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.985714285714286e-05,
      "loss": 0.5438,
      "step": 139
    },
    {
      "epoch": 0.09,
      "learning_rate": 2e-05,
      "loss": 0.5777,
      "step": 140
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9999997567387307e-05,
      "loss": 0.5325,
      "step": 141
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9999990269550415e-05,
      "loss": 0.5643,
      "step": 142
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.999997810649287e-05,
      "loss": 0.5806,
      "step": 143
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9999961078220587e-05,
      "loss": 0.5612,
      "step": 144
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.999993918474186e-05,
      "loss": 0.5644,
      "step": 145
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9999912426067335e-05,
      "loss": 0.5737,
      "step": 146
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.999988080221003e-05,
      "loss": 0.5334,
      "step": 147
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9999844313185335e-05,
      "loss": 0.541,
      "step": 148
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9999802959010998e-05,
      "loss": 0.5348,
      "step": 149
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.999975673970714e-05,
      "loss": 0.5642,
      "step": 150
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.999970565529625e-05,
      "loss": 0.5419,
      "step": 151
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9999649705803178e-05,
      "loss": 0.5634,
      "step": 152
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.999958889125515e-05,
      "loss": 0.5292,
      "step": 153
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9999523211681746e-05,
      "loss": 0.5439,
      "step": 154
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.999945266711493e-05,
      "loss": 0.5653,
      "step": 155
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9999377257589012e-05,
      "loss": 0.5286,
      "step": 156
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9999296983140692e-05,
      "loss": 0.536,
      "step": 157
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9999211843809018e-05,
      "loss": 0.5591,
      "step": 158
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9999121839635416e-05,
      "loss": 0.551,
      "step": 159
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.999902697066367e-05,
      "loss": 0.5677,
      "step": 160
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9998927236939943e-05,
      "loss": 0.5327,
      "step": 161
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9998822638512757e-05,
      "loss": 0.5346,
      "step": 162
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9998713175432993e-05,
      "loss": 0.548,
      "step": 163
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9998598847753918e-05,
      "loss": 0.5638,
      "step": 164
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9998479655531145e-05,
      "loss": 0.5493,
      "step": 165
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.999835559882267e-05,
      "loss": 0.5772,
      "step": 166
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9998226677688847e-05,
      "loss": 0.5593,
      "step": 167
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9998092892192403e-05,
      "loss": 0.5432,
      "step": 168
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9997954242398422e-05,
      "loss": 0.573,
      "step": 169
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9997810728374362e-05,
      "loss": 0.5282,
      "step": 170
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.999766235019005e-05,
      "loss": 0.5535,
      "step": 171
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.999750910791767e-05,
      "loss": 0.5368,
      "step": 172
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.999735100163178e-05,
      "loss": 0.5311,
      "step": 173
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9997188031409302e-05,
      "loss": 0.5392,
      "step": 174
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9997020197329525e-05,
      "loss": 0.528,
      "step": 175
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9996847499474102e-05,
      "loss": 0.5416,
      "step": 176
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.999666993792706e-05,
      "loss": 0.5657,
      "step": 177
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.999648751277478e-05,
      "loss": 0.5483,
      "step": 178
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9996300224106023e-05,
      "loss": 0.5429,
      "step": 179
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.99961080720119e-05,
      "loss": 0.5482,
      "step": 180
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9995911056585907e-05,
      "loss": 0.5322,
      "step": 181
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.999570917792389e-05,
      "loss": 0.5418,
      "step": 182
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.999550243612407e-05,
      "loss": 0.5531,
      "step": 183
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9995290831287032e-05,
      "loss": 0.5559,
      "step": 184
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.999507436351572e-05,
      "loss": 0.5436,
      "step": 185
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.999485303291546e-05,
      "loss": 0.5579,
      "step": 186
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.999462683959393e-05,
      "loss": 0.5515,
      "step": 187
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9994395783661177e-05,
      "loss": 0.5471,
      "step": 188
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9994159865229617e-05,
      "loss": 0.5326,
      "step": 189
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.999391908441403e-05,
      "loss": 0.5458,
      "step": 190
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.999367344133156e-05,
      "loss": 0.5343,
      "step": 191
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9993422936101715e-05,
      "loss": 0.5584,
      "step": 192
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9993167568846375e-05,
      "loss": 0.564,
      "step": 193
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9992907339689786e-05,
      "loss": 0.5542,
      "step": 194
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.999264224875855e-05,
      "loss": 0.5451,
      "step": 195
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9992372296181637e-05,
      "loss": 0.5544,
      "step": 196
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.999209748209039e-05,
      "loss": 0.5338,
      "step": 197
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9991817806618512e-05,
      "loss": 0.5715,
      "step": 198
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9991533269902068e-05,
      "loss": 0.5207,
      "step": 199
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9991243872079495e-05,
      "loss": 0.5569,
      "step": 200
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.999094961329159e-05,
      "loss": 0.57,
      "step": 201
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9990650493681517e-05,
      "loss": 0.5621,
      "step": 202
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9990346513394807e-05,
      "loss": 0.5594,
      "step": 203
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9990037672579347e-05,
      "loss": 0.5548,
      "step": 204
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.99897239713854e-05,
      "loss": 0.5468,
      "step": 205
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9989405409965585e-05,
      "loss": 0.5641,
      "step": 206
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9989081988474896e-05,
      "loss": 0.5559,
      "step": 207
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9988753707070675e-05,
      "loss": 0.5241,
      "step": 208
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9988420565912647e-05,
      "loss": 0.5582,
      "step": 209
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.998808256516289e-05,
      "loss": 0.5595,
      "step": 210
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9987739704985846e-05,
      "loss": 0.5221,
      "step": 211
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9987391985548326e-05,
      "loss": 0.55,
      "step": 212
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9987039407019508e-05,
      "loss": 0.5469,
      "step": 213
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9986681969570924e-05,
      "loss": 0.5509,
      "step": 214
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9986319673376476e-05,
      "loss": 0.5538,
      "step": 215
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.998595251861243e-05,
      "loss": 0.5178,
      "step": 216
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.998558050545741e-05,
      "loss": 0.5523,
      "step": 217
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.998520363409242e-05,
      "loss": 0.5648,
      "step": 218
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.998482190470081e-05,
      "loss": 0.5409,
      "step": 219
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9984435317468298e-05,
      "loss": 0.5454,
      "step": 220
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.998404387258297e-05,
      "loss": 0.508,
      "step": 221
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9983647570235274e-05,
      "loss": 0.5704,
      "step": 222
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9983246410618016e-05,
      "loss": 0.5333,
      "step": 223
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9982840393926374e-05,
      "loss": 0.5423,
      "step": 224
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.998242952035788e-05,
      "loss": 0.5624,
      "step": 225
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9982013790112437e-05,
      "loss": 0.582,
      "step": 226
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9981593203392302e-05,
      "loss": 0.5619,
      "step": 227
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9981167760402104e-05,
      "loss": 0.5571,
      "step": 228
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9980737461348832e-05,
      "loss": 0.5536,
      "step": 229
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9980302306441834e-05,
      "loss": 0.5412,
      "step": 230
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.997986229589282e-05,
      "loss": 0.5462,
      "step": 231
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.997941742991587e-05,
      "loss": 0.523,
      "step": 232
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9978967708727418e-05,
      "loss": 0.5455,
      "step": 233
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.997851313254626e-05,
      "loss": 0.5489,
      "step": 234
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.997805370159357e-05,
      "loss": 0.5374,
      "step": 235
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.997758941609286e-05,
      "loss": 0.5308,
      "step": 236
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9977120276270018e-05,
      "loss": 0.5383,
      "step": 237
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.997664628235329e-05,
      "loss": 0.5378,
      "step": 238
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9976167434573293e-05,
      "loss": 0.533,
      "step": 239
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9975683733162987e-05,
      "loss": 0.564,
      "step": 240
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9975195178357706e-05,
      "loss": 0.5159,
      "step": 241
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9974701770395147e-05,
      "loss": 0.5318,
      "step": 242
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.997420350951536e-05,
      "loss": 0.5453,
      "step": 243
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9973700395960765e-05,
      "loss": 0.5232,
      "step": 244
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9973192429976135e-05,
      "loss": 0.5186,
      "step": 245
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9972679611808603e-05,
      "loss": 0.5772,
      "step": 246
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.997216194170767e-05,
      "loss": 0.5436,
      "step": 247
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9971639419925197e-05,
      "loss": 0.5426,
      "step": 248
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.99711120467154e-05,
      "loss": 0.5514,
      "step": 249
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9970579822334856e-05,
      "loss": 0.5412,
      "step": 250
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.997004274704251e-05,
      "loss": 0.5794,
      "step": 251
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9969500821099654e-05,
      "loss": 0.557,
      "step": 252
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.996895404476995e-05,
      "loss": 0.545,
      "step": 253
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.996840241831942e-05,
      "loss": 0.5505,
      "step": 254
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.996784594201644e-05,
      "loss": 0.5327,
      "step": 255
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.996728461613175e-05,
      "loss": 0.5485,
      "step": 256
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9966718440938443e-05,
      "loss": 0.5157,
      "step": 257
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.996614741671198e-05,
      "loss": 0.5348,
      "step": 258
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.996557154373018e-05,
      "loss": 0.5515,
      "step": 259
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.996499082227321e-05,
      "loss": 0.5141,
      "step": 260
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9964405252623614e-05,
      "loss": 0.5236,
      "step": 261
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9963814835066274e-05,
      "loss": 0.5359,
      "step": 262
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9963219569888454e-05,
      "loss": 0.544,
      "step": 263
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.996261945737975e-05,
      "loss": 0.5512,
      "step": 264
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.996201449783214e-05,
      "loss": 0.5178,
      "step": 265
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9961404691539947e-05,
      "loss": 0.5453,
      "step": 266
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.996079003879986e-05,
      "loss": 0.5194,
      "step": 267
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.996017053991091e-05,
      "loss": 0.5614,
      "step": 268
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.995954619517451e-05,
      "loss": 0.5558,
      "step": 269
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.995891700489441e-05,
      "loss": 0.5551,
      "step": 270
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.995828296937673e-05,
      "loss": 0.5212,
      "step": 271
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.995764408892994e-05,
      "loss": 0.5499,
      "step": 272
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9957000363864872e-05,
      "loss": 0.5495,
      "step": 273
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9956351794494706e-05,
      "loss": 0.5249,
      "step": 274
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9955698381134994e-05,
      "loss": 0.5451,
      "step": 275
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.995504012410363e-05,
      "loss": 0.5735,
      "step": 276
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9954377023720877e-05,
      "loss": 0.5498,
      "step": 277
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.995370908030934e-05,
      "loss": 0.5494,
      "step": 278
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9953036294194003e-05,
      "loss": 0.5351,
      "step": 279
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.995235866570217e-05,
      "loss": 0.5497,
      "step": 280
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9951676195163542e-05,
      "loss": 0.5328,
      "step": 281
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.995098888291015e-05,
      "loss": 0.5339,
      "step": 282
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9950296729276384e-05,
      "loss": 0.5167,
      "step": 283
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9949599734598993e-05,
      "loss": 0.5296,
      "step": 284
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.994889789921708e-05,
      "loss": 0.5455,
      "step": 285
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9948191223472108e-05,
      "loss": 0.5314,
      "step": 286
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.994747970770789e-05,
      "loss": 0.5448,
      "step": 287
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.994676335227059e-05,
      "loss": 0.5407,
      "step": 288
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9946042157508734e-05,
      "loss": 0.5241,
      "step": 289
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.99453161237732e-05,
      "loss": 0.5222,
      "step": 290
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.994458525141722e-05,
      "loss": 0.5473,
      "step": 291
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9943849540796375e-05,
      "loss": 0.5362,
      "step": 292
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9943108992268612e-05,
      "loss": 0.5086,
      "step": 293
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.994236360619422e-05,
      "loss": 0.5345,
      "step": 294
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9941613382935848e-05,
      "loss": 0.5521,
      "step": 295
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9940858322858493e-05,
      "loss": 0.5317,
      "step": 296
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9940098426329518e-05,
      "loss": 0.5357,
      "step": 297
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.993933369371862e-05,
      "loss": 0.5394,
      "step": 298
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9938564125397862e-05,
      "loss": 0.5611,
      "step": 299
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9937789721741654e-05,
      "loss": 0.5735,
      "step": 300
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9937010483126763e-05,
      "loss": 0.5204,
      "step": 301
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.993622640993231e-05,
      "loss": 0.5674,
      "step": 302
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9935437502539762e-05,
      "loss": 0.5214,
      "step": 303
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9934643761332933e-05,
      "loss": 0.5282,
      "step": 304
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9933845186698007e-05,
      "loss": 0.5032,
      "step": 305
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9933041779023502e-05,
      "loss": 0.5617,
      "step": 306
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9932233538700295e-05,
      "loss": 0.5301,
      "step": 307
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9931420466121613e-05,
      "loss": 0.5412,
      "step": 308
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9930602561683035e-05,
      "loss": 0.513,
      "step": 309
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.992977982578249e-05,
      "loss": 0.5272,
      "step": 310
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9928952258820255e-05,
      "loss": 0.5226,
      "step": 311
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9928119861198962e-05,
      "loss": 0.5489,
      "step": 312
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9927282633323596e-05,
      "loss": 0.5454,
      "step": 313
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.992644057560148e-05,
      "loss": 0.518,
      "step": 314
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9925593688442296e-05,
      "loss": 0.5375,
      "step": 315
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9924741972258076e-05,
      "loss": 0.5294,
      "step": 316
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9923885427463193e-05,
      "loss": 0.5467,
      "step": 317
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9923024054474384e-05,
      "loss": 0.5252,
      "step": 318
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.992215785371072e-05,
      "loss": 0.5316,
      "step": 319
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9921286825593632e-05,
      "loss": 0.5402,
      "step": 320
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.992041097054689e-05,
      "loss": 0.5535,
      "step": 321
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9919530288996617e-05,
      "loss": 0.5292,
      "step": 322
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9918644781371292e-05,
      "loss": 0.5497,
      "step": 323
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9917754448101725e-05,
      "loss": 0.5495,
      "step": 324
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.991685928962109e-05,
      "loss": 0.5325,
      "step": 325
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9915959306364897e-05,
      "loss": 0.5479,
      "step": 326
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.991505449877101e-05,
      "loss": 0.5601,
      "step": 327
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9914144867279644e-05,
      "loss": 0.5369,
      "step": 328
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9913230412333348e-05,
      "loss": 0.5402,
      "step": 329
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9912311134377023e-05,
      "loss": 0.54,
      "step": 330
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9911387033857924e-05,
      "loss": 0.5231,
      "step": 331
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9910458111225645e-05,
      "loss": 0.5432,
      "step": 332
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9909524366932127e-05,
      "loss": 0.5065,
      "step": 333
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9908585801431658e-05,
      "loss": 0.5712,
      "step": 334
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9907642415180874e-05,
      "loss": 0.5608,
      "step": 335
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.990669420863875e-05,
      "loss": 0.5315,
      "step": 336
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.990574118226661e-05,
      "loss": 0.53,
      "step": 337
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9904783336528128e-05,
      "loss": 0.5429,
      "step": 338
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9903820671889308e-05,
      "loss": 0.5159,
      "step": 339
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9902853188818518e-05,
      "loss": 0.5351,
      "step": 340
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9901880887786455e-05,
      "loss": 0.5402,
      "step": 341
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9900903769266167e-05,
      "loss": 0.5427,
      "step": 342
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9899921833733042e-05,
      "loss": 0.5347,
      "step": 343
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9898935081664814e-05,
      "loss": 0.5458,
      "step": 344
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9897943513541564e-05,
      "loss": 0.5412,
      "step": 345
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9896947129845707e-05,
      "loss": 0.5266,
      "step": 346
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.989594593106201e-05,
      "loss": 0.5314,
      "step": 347
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9894939917677577e-05,
      "loss": 0.5732,
      "step": 348
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.989392909018186e-05,
      "loss": 0.5612,
      "step": 349
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9892913449066643e-05,
      "loss": 0.5427,
      "step": 350
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.989189299482606e-05,
      "loss": 0.5552,
      "step": 351
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9890867727956587e-05,
      "loss": 0.5122,
      "step": 352
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.988983764895704e-05,
      "loss": 0.5284,
      "step": 353
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9888802758328574e-05,
      "loss": 0.5401,
      "step": 354
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9887763056574688e-05,
      "loss": 0.5387,
      "step": 355
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.988671854420122e-05,
      "loss": 0.5396,
      "step": 356
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9885669221716343e-05,
      "loss": 0.5576,
      "step": 357
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9884615089630584e-05,
      "loss": 0.5345,
      "step": 358
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9883556148456802e-05,
      "loss": 0.5508,
      "step": 359
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9882492398710192e-05,
      "loss": 0.5368,
      "step": 360
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.988142384090829e-05,
      "loss": 0.513,
      "step": 361
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.988035047557098e-05,
      "loss": 0.5137,
      "step": 362
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9879272303220473e-05,
      "loss": 0.5366,
      "step": 363
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.987818932438133e-05,
      "loss": 0.5329,
      "step": 364
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9877101539580437e-05,
      "loss": 0.5417,
      "step": 365
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.987600894934703e-05,
      "loss": 0.527,
      "step": 366
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.987491155421268e-05,
      "loss": 0.5317,
      "step": 367
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.987380935471129e-05,
      "loss": 0.5264,
      "step": 368
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9872702351379108e-05,
      "loss": 0.5193,
      "step": 369
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.987159054475472e-05,
      "loss": 0.5315,
      "step": 370
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9870473935379038e-05,
      "loss": 0.5466,
      "step": 371
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.986935252379532e-05,
      "loss": 0.5237,
      "step": 372
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.986822631054916e-05,
      "loss": 0.5294,
      "step": 373
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9867095296188483e-05,
      "loss": 0.4956,
      "step": 374
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.986595948126356e-05,
      "loss": 0.5384,
      "step": 375
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9864818866326978e-05,
      "loss": 0.5612,
      "step": 376
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9863673451933684e-05,
      "loss": 0.5131,
      "step": 377
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.986252323864094e-05,
      "loss": 0.5381,
      "step": 378
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9861368227008352e-05,
      "loss": 0.5267,
      "step": 379
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9860208417597863e-05,
      "loss": 0.5225,
      "step": 380
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9859043810973747e-05,
      "loss": 0.5399,
      "step": 381
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9857874407702606e-05,
      "loss": 0.5378,
      "step": 382
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9856700208353384e-05,
      "loss": 0.5491,
      "step": 383
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9855521213497355e-05,
      "loss": 0.5325,
      "step": 384
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9854337423708125e-05,
      "loss": 0.5023,
      "step": 385
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9853148839561638e-05,
      "loss": 0.5401,
      "step": 386
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9851955461636164e-05,
      "loss": 0.5376,
      "step": 387
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9850757290512313e-05,
      "loss": 0.5657,
      "step": 388
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9849554326773013e-05,
      "loss": 0.538,
      "step": 389
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.984834657100354e-05,
      "loss": 0.5383,
      "step": 390
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9847134023791497e-05,
      "loss": 0.5422,
      "step": 391
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9845916685726808e-05,
      "loss": 0.5054,
      "step": 392
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.984469455740174e-05,
      "loss": 0.5496,
      "step": 393
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9843467639410885e-05,
      "loss": 0.5269,
      "step": 394
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9842235932351164e-05,
      "loss": 0.5169,
      "step": 395
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9840999436821836e-05,
      "loss": 0.5428,
      "step": 396
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9839758153424478e-05,
      "loss": 0.5226,
      "step": 397
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9838512082763002e-05,
      "loss": 0.5509,
      "step": 398
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9837261225443657e-05,
      "loss": 0.5088,
      "step": 399
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.983600558207501e-05,
      "loss": 0.5259,
      "step": 400
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9834745153267954e-05,
      "loss": 0.5034,
      "step": 401
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9833479939635724e-05,
      "loss": 0.5523,
      "step": 402
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9832209941793865e-05,
      "loss": 0.5178,
      "step": 403
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.983093516036027e-05,
      "loss": 0.5325,
      "step": 404
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9829655595955144e-05,
      "loss": 0.5085,
      "step": 405
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9828371249201025e-05,
      "loss": 0.5554,
      "step": 406
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9827082120722774e-05,
      "loss": 0.5494,
      "step": 407
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9825788211147587e-05,
      "loss": 0.5123,
      "step": 408
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9824489521104975e-05,
      "loss": 0.5323,
      "step": 409
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9823186051226783e-05,
      "loss": 0.5482,
      "step": 410
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.982187780214717e-05,
      "loss": 0.5365,
      "step": 411
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9820564774502644e-05,
      "loss": 0.5223,
      "step": 412
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9819246968932007e-05,
      "loss": 0.5216,
      "step": 413
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9817924386076416e-05,
      "loss": 0.5343,
      "step": 414
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9816597026579326e-05,
      "loss": 0.5283,
      "step": 415
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.981526489108653e-05,
      "loss": 0.5313,
      "step": 416
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.981392798024614e-05,
      "loss": 0.5116,
      "step": 417
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.98125862947086e-05,
      "loss": 0.526,
      "step": 418
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9811239835126667e-05,
      "loss": 0.5288,
      "step": 419
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.980988860215542e-05,
      "loss": 0.5423,
      "step": 420
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.980853259645227e-05,
      "loss": 0.5465,
      "step": 421
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9807171818676944e-05,
      "loss": 0.5568,
      "step": 422
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9805806269491487e-05,
      "loss": 0.5276,
      "step": 423
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.980443594956027e-05,
      "loss": 0.5368,
      "step": 424
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9803060859549988e-05,
      "loss": 0.52,
      "step": 425
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9801681000129652e-05,
      "loss": 0.5274,
      "step": 426
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.980029637197059e-05,
      "loss": 0.5366,
      "step": 427
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9798906975746462e-05,
      "loss": 0.5499,
      "step": 428
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9797512812133236e-05,
      "loss": 0.5214,
      "step": 429
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9796113881809207e-05,
      "loss": 0.5087,
      "step": 430
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9794710185454983e-05,
      "loss": 0.5426,
      "step": 431
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9793301723753494e-05,
      "loss": 0.5256,
      "step": 432
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.979188849738999e-05,
      "loss": 0.5391,
      "step": 433
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9790470507052043e-05,
      "loss": 0.5295,
      "step": 434
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9789047753429527e-05,
      "loss": 0.531,
      "step": 435
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9787620237214648e-05,
      "loss": 0.5447,
      "step": 436
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9786187959101926e-05,
      "loss": 0.5294,
      "step": 437
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9784750919788192e-05,
      "loss": 0.5483,
      "step": 438
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9783309119972607e-05,
      "loss": 0.5397,
      "step": 439
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9781862560356632e-05,
      "loss": 0.5325,
      "step": 440
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9780411241644054e-05,
      "loss": 0.501,
      "step": 441
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9778955164540966e-05,
      "loss": 0.5103,
      "step": 442
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.977749432975579e-05,
      "loss": 0.5448,
      "step": 443
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9776028737999256e-05,
      "loss": 0.541,
      "step": 444
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9774558389984398e-05,
      "loss": 0.5259,
      "step": 445
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.977308328642658e-05,
      "loss": 0.5289,
      "step": 446
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9771603428043476e-05,
      "loss": 0.5693,
      "step": 447
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9770118815555063e-05,
      "loss": 0.5139,
      "step": 448
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9768629449683644e-05,
      "loss": 0.5041,
      "step": 449
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9767135331153827e-05,
      "loss": 0.5147,
      "step": 450
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9765636460692534e-05,
      "loss": 0.5157,
      "step": 451
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9764132839029e-05,
      "loss": 0.5483,
      "step": 452
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9762624466894767e-05,
      "loss": 0.5297,
      "step": 453
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9761111345023702e-05,
      "loss": 0.5424,
      "step": 454
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9759593474151965e-05,
      "loss": 0.5515,
      "step": 455
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9758070855018033e-05,
      "loss": 0.5214,
      "step": 456
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.97565434883627e-05,
      "loss": 0.5137,
      "step": 457
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.975501137492906e-05,
      "loss": 0.5408,
      "step": 458
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9753474515462526e-05,
      "loss": 0.5507,
      "step": 459
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9751932910710808e-05,
      "loss": 0.5226,
      "step": 460
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9750386561423935e-05,
      "loss": 0.551,
      "step": 461
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9748835468354243e-05,
      "loss": 0.5423,
      "step": 462
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.974727963225637e-05,
      "loss": 0.5084,
      "step": 463
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9745719053887265e-05,
      "loss": 0.5031,
      "step": 464
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9744153734006187e-05,
      "loss": 0.5297,
      "step": 465
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.97425836733747e-05,
      "loss": 0.5029,
      "step": 466
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.974100887275667e-05,
      "loss": 0.5294,
      "step": 467
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9739429332918276e-05,
      "loss": 0.528,
      "step": 468
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9737845054627996e-05,
      "loss": 0.5551,
      "step": 469
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9736256038656624e-05,
      "loss": 0.5375,
      "step": 470
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.973466228577725e-05,
      "loss": 0.5556,
      "step": 471
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9733063796765267e-05,
      "loss": 0.5253,
      "step": 472
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9731460572398377e-05,
      "loss": 0.5209,
      "step": 473
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9729852613456586e-05,
      "loss": 0.5211,
      "step": 474
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9728239920722207e-05,
      "loss": 0.5268,
      "step": 475
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.972662249497984e-05,
      "loss": 0.4986,
      "step": 476
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9725000337016407e-05,
      "loss": 0.5404,
      "step": 477
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9723373447621125e-05,
      "loss": 0.5192,
      "step": 478
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9721741827585505e-05,
      "loss": 0.5154,
      "step": 479
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.972010547770338e-05,
      "loss": 0.5226,
      "step": 480
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9718464398770856e-05,
      "loss": 0.5242,
      "step": 481
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9716818591586367e-05,
      "loss": 0.5179,
      "step": 482
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9715168056950626e-05,
      "loss": 0.5344,
      "step": 483
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9713512795666663e-05,
      "loss": 0.5218,
      "step": 484
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9711852808539795e-05,
      "loss": 0.5335,
      "step": 485
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9710188096377645e-05,
      "loss": 0.5395,
      "step": 486
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.970851865999013e-05,
      "loss": 0.5522,
      "step": 487
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.970684450018948e-05,
      "loss": 0.5229,
      "step": 488
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9705165617790195e-05,
      "loss": 0.4993,
      "step": 489
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9703482013609098e-05,
      "loss": 0.5366,
      "step": 490
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9701793688465302e-05,
      "loss": 0.543,
      "step": 491
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9700100643180213e-05,
      "loss": 0.5345,
      "step": 492
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9698402878577528e-05,
      "loss": 0.5308,
      "step": 493
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.969670039548326e-05,
      "loss": 0.5218,
      "step": 494
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9694993194725703e-05,
      "loss": 0.5147,
      "step": 495
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.969328127713544e-05,
      "loss": 0.5315,
      "step": 496
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.969156464354537e-05,
      "loss": 0.5246,
      "step": 497
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.968984329479066e-05,
      "loss": 0.5251,
      "step": 498
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9688117231708795e-05,
      "loss": 0.5446,
      "step": 499
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9686386455139544e-05,
      "loss": 0.552,
      "step": 500
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.968465096592496e-05,
      "loss": 0.5372,
      "step": 501
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9682910764909405e-05,
      "loss": 0.5087,
      "step": 502
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9681165852939526e-05,
      "loss": 0.5324,
      "step": 503
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9679416230864265e-05,
      "loss": 0.5375,
      "step": 504
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9677661899534837e-05,
      "loss": 0.532,
      "step": 505
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9675902859804786e-05,
      "loss": 0.5115,
      "step": 506
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9674139112529908e-05,
      "loss": 0.5388,
      "step": 507
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9672370658568306e-05,
      "loss": 0.5271,
      "step": 508
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9670597498780385e-05,
      "loss": 0.5243,
      "step": 509
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9668819634028816e-05,
      "loss": 0.527,
      "step": 510
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9667037065178577e-05,
      "loss": 0.5067,
      "step": 511
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.966524979309692e-05,
      "loss": 0.5406,
      "step": 512
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.96634578186534e-05,
      "loss": 0.5242,
      "step": 513
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9661661142719856e-05,
      "loss": 0.5377,
      "step": 514
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.96598597661704e-05,
      "loss": 0.5251,
      "step": 515
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9658053689881453e-05,
      "loss": 0.524,
      "step": 516
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9656242914731708e-05,
      "loss": 0.5096,
      "step": 517
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9654427441602145e-05,
      "loss": 0.4962,
      "step": 518
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9652607271376036e-05,
      "loss": 0.5179,
      "step": 519
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9650782404938933e-05,
      "loss": 0.5297,
      "step": 520
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9648952843178676e-05,
      "loss": 0.5068,
      "step": 521
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.964711858698539e-05,
      "loss": 0.5411,
      "step": 522
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9645279637251474e-05,
      "loss": 0.529,
      "step": 523
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9643435994871626e-05,
      "loss": 0.5522,
      "step": 524
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9641587660742822e-05,
      "loss": 0.5147,
      "step": 525
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.963973463576431e-05,
      "loss": 0.5814,
      "step": 526
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9637876920837633e-05,
      "loss": 0.5385,
      "step": 527
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.963601451686661e-05,
      "loss": 0.5507,
      "step": 528
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9634147424757347e-05,
      "loss": 0.5506,
      "step": 529
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9632275645418218e-05,
      "loss": 0.5117,
      "step": 530
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9630399179759888e-05,
      "loss": 0.5417,
      "step": 531
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9628518028695307e-05,
      "loss": 0.5435,
      "step": 532
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.962663219313969e-05,
      "loss": 0.5438,
      "step": 533
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.9624741674010544e-05,
      "loss": 0.5227,
      "step": 534
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.962284647222764e-05,
      "loss": 0.5473,
      "step": 535
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9620946588713048e-05,
      "loss": 0.5534,
      "step": 536
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9619042024391098e-05,
      "loss": 0.5183,
      "step": 537
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.96171327801884e-05,
      "loss": 0.5088,
      "step": 538
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9615218857033852e-05,
      "loss": 0.4927,
      "step": 539
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9613300255858615e-05,
      "loss": 0.5001,
      "step": 540
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9611376977596138e-05,
      "loss": 0.5262,
      "step": 541
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9609449023182133e-05,
      "loss": 0.5534,
      "step": 542
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.960751639355459e-05,
      "loss": 0.5526,
      "step": 543
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.960557908965379e-05,
      "loss": 0.5411,
      "step": 544
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9603637112422263e-05,
      "loss": 0.5257,
      "step": 545
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.960169046280483e-05,
      "loss": 0.5081,
      "step": 546
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.959973914174858e-05,
      "loss": 0.5143,
      "step": 547
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9597783150202873e-05,
      "loss": 0.5431,
      "step": 548
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.959582248911934e-05,
      "loss": 0.524,
      "step": 549
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9593857159451897e-05,
      "loss": 0.5435,
      "step": 550
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.959188716215671e-05,
      "loss": 0.5164,
      "step": 551
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9589912498192233e-05,
      "loss": 0.5172,
      "step": 552
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9587933168519183e-05,
      "loss": 0.5333,
      "step": 553
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.958594917410055e-05,
      "loss": 0.5105,
      "step": 554
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9583960515901586e-05,
      "loss": 0.5283,
      "step": 555
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9581967194889826e-05,
      "loss": 0.5295,
      "step": 556
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.957996921203506e-05,
      "loss": 0.5359,
      "step": 557
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9577966568309358e-05,
      "loss": 0.5194,
      "step": 558
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9575959264687047e-05,
      "loss": 0.522,
      "step": 559
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.957394730214472e-05,
      "loss": 0.5252,
      "step": 560
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9571930681661254e-05,
      "loss": 0.5251,
      "step": 561
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.956990940421777e-05,
      "loss": 0.5317,
      "step": 562
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9567883470797673e-05,
      "loss": 0.5112,
      "step": 563
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.956585288238662e-05,
      "loss": 0.5459,
      "step": 564
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9563817639972535e-05,
      "loss": 0.5297,
      "step": 565
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9561777744545616e-05,
      "loss": 0.551,
      "step": 566
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.955973319709832e-05,
      "loss": 0.5234,
      "step": 567
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.955768399862536e-05,
      "loss": 0.5151,
      "step": 568
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9555630150123717e-05,
      "loss": 0.527,
      "step": 569
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9553571652592637e-05,
      "loss": 0.5268,
      "step": 570
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9551508507033623e-05,
      "loss": 0.5035,
      "step": 571
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9549440714450447e-05,
      "loss": 0.5098,
      "step": 572
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9547368275849128e-05,
      "loss": 0.5506,
      "step": 573
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9545291192237962e-05,
      "loss": 0.4807,
      "step": 574
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9543209464627493e-05,
      "loss": 0.5326,
      "step": 575
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9541123094030528e-05,
      "loss": 0.5183,
      "step": 576
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.953903208146214e-05,
      "loss": 0.5152,
      "step": 577
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9536936427939647e-05,
      "loss": 0.5414,
      "step": 578
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.953483613448263e-05,
      "loss": 0.5182,
      "step": 579
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9532731202112935e-05,
      "loss": 0.532,
      "step": 580
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.953062163185466e-05,
      "loss": 0.5202,
      "step": 581
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9528507424734148e-05,
      "loss": 0.4976,
      "step": 582
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9526388581780016e-05,
      "loss": 0.5149,
      "step": 583
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9524265104023133e-05,
      "loss": 0.5282,
      "step": 584
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.952213699249661e-05,
      "loss": 0.5355,
      "step": 585
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9520004248235826e-05,
      "loss": 0.5371,
      "step": 586
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.951786687227841e-05,
      "loss": 0.5383,
      "step": 587
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9515724865664242e-05,
      "loss": 0.5186,
      "step": 588
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9513578229435454e-05,
      "loss": 0.5302,
      "step": 589
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9511426964636437e-05,
      "loss": 0.4952,
      "step": 590
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9509271072313826e-05,
      "loss": 0.5334,
      "step": 591
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9507110553516518e-05,
      "loss": 0.5343,
      "step": 592
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9504945409295644e-05,
      "loss": 0.4968,
      "step": 593
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9502775640704606e-05,
      "loss": 0.4991,
      "step": 594
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9500601248799035e-05,
      "loss": 0.5088,
      "step": 595
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.949842223463683e-05,
      "loss": 0.5173,
      "step": 596
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.949623859927813e-05,
      "loss": 0.5067,
      "step": 597
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9494050343785317e-05,
      "loss": 0.5392,
      "step": 598
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9491857469223027e-05,
      "loss": 0.5193,
      "step": 599
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9489659976658152e-05,
      "loss": 0.5424,
      "step": 600
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.948745786715981e-05,
      "loss": 0.5283,
      "step": 601
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9485251141799387e-05,
      "loss": 0.5349,
      "step": 602
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9483039801650494e-05,
      "loss": 0.5439,
      "step": 603
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9480823847789007e-05,
      "loss": 0.5412,
      "step": 604
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9478603281293037e-05,
      "loss": 0.5059,
      "step": 605
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9476378103242934e-05,
      "loss": 0.5328,
      "step": 606
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.94741483147213e-05,
      "loss": 0.5072,
      "step": 607
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.947191391681298e-05,
      "loss": 0.5313,
      "step": 608
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9469674910605053e-05,
      "loss": 0.5252,
      "step": 609
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.946743129718685e-05,
      "loss": 0.4906,
      "step": 610
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9465183077649938e-05,
      "loss": 0.5487,
      "step": 611
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.946293025308813e-05,
      "loss": 0.556,
      "step": 612
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9460672824597472e-05,
      "loss": 0.5563,
      "step": 613
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9458410793276256e-05,
      "loss": 0.5186,
      "step": 614
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.945614416022501e-05,
      "loss": 0.5315,
      "step": 615
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9453872926546505e-05,
      "loss": 0.527,
      "step": 616
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.945159709334574e-05,
      "loss": 0.5314,
      "step": 617
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.944931666172997e-05,
      "loss": 0.5245,
      "step": 618
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9447031632808662e-05,
      "loss": 0.5349,
      "step": 619
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.944474200769355e-05,
      "loss": 0.5257,
      "step": 620
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.944244778749858e-05,
      "loss": 0.5106,
      "step": 621
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9440148973339937e-05,
      "loss": 0.5034,
      "step": 622
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9437845566336056e-05,
      "loss": 0.5187,
      "step": 623
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.943553756760759e-05,
      "loss": 0.5316,
      "step": 624
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9433224978277437e-05,
      "loss": 0.5229,
      "step": 625
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.943090779947072e-05,
      "loss": 0.56,
      "step": 626
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.94285860323148e-05,
      "loss": 0.5181,
      "step": 627
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9426259677939264e-05,
      "loss": 0.5406,
      "step": 628
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9423928737475946e-05,
      "loss": 0.546,
      "step": 629
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9421593212058894e-05,
      "loss": 0.501,
      "step": 630
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9419253102824393e-05,
      "loss": 0.5548,
      "step": 631
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9416908410910965e-05,
      "loss": 0.5203,
      "step": 632
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.941455913745935e-05,
      "loss": 0.5015,
      "step": 633
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9412205283612527e-05,
      "loss": 0.5095,
      "step": 634
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9409846850515694e-05,
      "loss": 0.507,
      "step": 635
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9407483839316284e-05,
      "loss": 0.5094,
      "step": 636
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9405116251163956e-05,
      "loss": 0.5266,
      "step": 637
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9402744087210594e-05,
      "loss": 0.519,
      "step": 638
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.940036734861031e-05,
      "loss": 0.5462,
      "step": 639
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.939798603651944e-05,
      "loss": 0.5288,
      "step": 640
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9395600152096547e-05,
      "loss": 0.5235,
      "step": 641
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9393209696502416e-05,
      "loss": 0.5068,
      "step": 642
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.939081467090006e-05,
      "loss": 0.5055,
      "step": 643
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.938841507645471e-05,
      "loss": 0.4918,
      "step": 644
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9386010914333826e-05,
      "loss": 0.5203,
      "step": 645
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9383602185707082e-05,
      "loss": 0.5184,
      "step": 646
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9381188891746384e-05,
      "loss": 0.5326,
      "step": 647
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9378771033625855e-05,
      "loss": 0.5026,
      "step": 648
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.937634861252183e-05,
      "loss": 0.492,
      "step": 649
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9373921629612876e-05,
      "loss": 0.5208,
      "step": 650
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.937149008607978e-05,
      "loss": 0.5311,
      "step": 651
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9369053983105533e-05,
      "loss": 0.5253,
      "step": 652
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9366613321875357e-05,
      "loss": 0.5196,
      "step": 653
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9364168103576696e-05,
      "loss": 0.506,
      "step": 654
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9361718329399194e-05,
      "loss": 0.4883,
      "step": 655
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9359264000534726e-05,
      "loss": 0.5371,
      "step": 656
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9356805118177377e-05,
      "loss": 0.5559,
      "step": 657
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.935434168352345e-05,
      "loss": 0.5216,
      "step": 658
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.935187369777146e-05,
      "loss": 0.5213,
      "step": 659
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.934940116212214e-05,
      "loss": 0.548,
      "step": 660
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9346924077778433e-05,
      "loss": 0.5042,
      "step": 661
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.93444424459455e-05,
      "loss": 0.5497,
      "step": 662
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9341956267830698e-05,
      "loss": 0.5188,
      "step": 663
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9339465544643623e-05,
      "loss": 0.5258,
      "step": 664
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9336970277596064e-05,
      "loss": 0.5246,
      "step": 665
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9334470467902024e-05,
      "loss": 0.5211,
      "step": 666
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9331966116777713e-05,
      "loss": 0.5291,
      "step": 667
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9329457225441554e-05,
      "loss": 0.5272,
      "step": 668
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.932694379511419e-05,
      "loss": 0.4845,
      "step": 669
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9324425827018452e-05,
      "loss": 0.5061,
      "step": 670
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9321903322379388e-05,
      "loss": 0.5369,
      "step": 671
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9319376282424255e-05,
      "loss": 0.5143,
      "step": 672
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.931684470838252e-05,
      "loss": 0.5188,
      "step": 673
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9314308601485842e-05,
      "loss": 0.5346,
      "step": 674
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9311767962968102e-05,
      "loss": 0.5272,
      "step": 675
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9309222794065373e-05,
      "loss": 0.5136,
      "step": 676
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9306673096015937e-05,
      "loss": 0.5096,
      "step": 677
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9304118870060283e-05,
      "loss": 0.5038,
      "step": 678
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9301560117441097e-05,
      "loss": 0.4892,
      "step": 679
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.929899683940327e-05,
      "loss": 0.5298,
      "step": 680
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9296429037193894e-05,
      "loss": 0.5057,
      "step": 681
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9293856712062267e-05,
      "loss": 0.4937,
      "step": 682
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.929127986525988e-05,
      "loss": 0.5471,
      "step": 683
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9288698498040423e-05,
      "loss": 0.5154,
      "step": 684
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9286112611659795e-05,
      "loss": 0.5058,
      "step": 685
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9283522207376088e-05,
      "loss": 0.5241,
      "step": 686
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9280927286449586e-05,
      "loss": 0.5174,
      "step": 687
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9278327850142783e-05,
      "loss": 0.539,
      "step": 688
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.927572389972036e-05,
      "loss": 0.5153,
      "step": 689
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9273115436449198e-05,
      "loss": 0.5586,
      "step": 690
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9270502461598377e-05,
      "loss": 0.5349,
      "step": 691
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9267884976439163e-05,
      "loss": 0.522,
      "step": 692
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9265262982245023e-05,
      "loss": 0.5196,
      "step": 693
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9262636480291618e-05,
      "loss": 0.5127,
      "step": 694
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.92600054718568e-05,
      "loss": 0.5202,
      "step": 695
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9257369958220612e-05,
      "loss": 0.5104,
      "step": 696
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.925472994066529e-05,
      "loss": 0.52,
      "step": 697
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9252085420475263e-05,
      "loss": 0.5268,
      "step": 698
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9249436398937154e-05,
      "loss": 0.5193,
      "step": 699
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9246782877339767e-05,
      "loss": 0.5164,
      "step": 700
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9244124856974102e-05,
      "loss": 0.5281,
      "step": 701
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9241462339133342e-05,
      "loss": 0.5348,
      "step": 702
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9238795325112867e-05,
      "loss": 0.5222,
      "step": 703
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.923612381621024e-05,
      "loss": 0.5289,
      "step": 704
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9233447813725206e-05,
      "loss": 0.5146,
      "step": 705
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.92307673189597e-05,
      "loss": 0.5453,
      "step": 706
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.922808233321785e-05,
      "loss": 0.5191,
      "step": 707
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9225392857805955e-05,
      "loss": 0.5315,
      "step": 708
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9222698894032508e-05,
      "loss": 0.5296,
      "step": 709
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9220000443208183e-05,
      "loss": 0.4835,
      "step": 710
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9217297506645837e-05,
      "loss": 0.5348,
      "step": 711
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.921459008566051e-05,
      "loss": 0.5128,
      "step": 712
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.921187818156942e-05,
      "loss": 0.5104,
      "step": 713
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9209161795691975e-05,
      "loss": 0.5354,
      "step": 714
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9206440929349756e-05,
      "loss": 0.5066,
      "step": 715
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9203715583866527e-05,
      "loss": 0.4869,
      "step": 716
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9200985760568227e-05,
      "loss": 0.4945,
      "step": 717
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9198251460782974e-05,
      "loss": 0.5308,
      "step": 718
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9195512685841072e-05,
      "loss": 0.5369,
      "step": 719
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9192769437075e-05,
      "loss": 0.5183,
      "step": 720
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.91900217158194e-05,
      "loss": 0.5246,
      "step": 721
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9187269523411108e-05,
      "loss": 0.5276,
      "step": 722
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9184512861189123e-05,
      "loss": 0.5251,
      "step": 723
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.918175173049463e-05,
      "loss": 0.5464,
      "step": 724
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9178986132670973e-05,
      "loss": 0.5311,
      "step": 725
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9176216069063683e-05,
      "loss": 0.5158,
      "step": 726
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9173441541020456e-05,
      "loss": 0.5181,
      "step": 727
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9170662549891162e-05,
      "loss": 0.5258,
      "step": 728
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.916787909702785e-05,
      "loss": 0.5152,
      "step": 729
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9165091183784722e-05,
      "loss": 0.5258,
      "step": 730
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9162298811518166e-05,
      "loss": 0.5436,
      "step": 731
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9159501981586738e-05,
      "loss": 0.521,
      "step": 732
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9156700695351148e-05,
      "loss": 0.514,
      "step": 733
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9153894954174294e-05,
      "loss": 0.5004,
      "step": 734
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.915108475942123e-05,
      "loss": 0.5062,
      "step": 735
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9148270112459178e-05,
      "loss": 0.5086,
      "step": 736
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9145451014657527e-05,
      "loss": 0.5314,
      "step": 737
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9142627467387833e-05,
      "loss": 0.5114,
      "step": 738
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9139799472023814e-05,
      "loss": 0.5255,
      "step": 739
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9136967029941354e-05,
      "loss": 0.5159,
      "step": 740
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.91341301425185e-05,
      "loss": 0.5295,
      "step": 741
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9131288811135465e-05,
      "loss": 0.5263,
      "step": 742
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9128443037174615e-05,
      "loss": 0.5009,
      "step": 743
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9125592822020485e-05,
      "loss": 0.52,
      "step": 744
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9122738167059773e-05,
      "loss": 0.5326,
      "step": 745
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9119879073681328e-05,
      "loss": 0.5362,
      "step": 746
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9117015543276163e-05,
      "loss": 0.5202,
      "step": 747
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9114147577237452e-05,
      "loss": 0.5316,
      "step": 748
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9111275176960527e-05,
      "loss": 0.5556,
      "step": 749
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9108398343842873e-05,
      "loss": 0.4871,
      "step": 750
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9105517079284136e-05,
      "loss": 0.5224,
      "step": 751
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9102631384686116e-05,
      "loss": 0.5302,
      "step": 752
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9099741261452764e-05,
      "loss": 0.5059,
      "step": 753
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9096846710990192e-05,
      "loss": 0.5189,
      "step": 754
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.909394773470667e-05,
      "loss": 0.5161,
      "step": 755
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.909104433401261e-05,
      "loss": 0.4971,
      "step": 756
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.908813651032058e-05,
      "loss": 0.5383,
      "step": 757
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.908522426504531e-05,
      "loss": 0.5312,
      "step": 758
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9082307599603664e-05,
      "loss": 0.5169,
      "step": 759
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9079386515414667e-05,
      "loss": 0.5395,
      "step": 760
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9076461013899495e-05,
      "loss": 0.511,
      "step": 761
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9073531096481475e-05,
      "loss": 0.5164,
      "step": 762
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9070596764586066e-05,
      "loss": 0.5313,
      "step": 763
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9067658019640897e-05,
      "loss": 0.5028,
      "step": 764
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.906471486307573e-05,
      "loss": 0.4968,
      "step": 765
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9061767296322477e-05,
      "loss": 0.5243,
      "step": 766
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9058815320815188e-05,
      "loss": 0.5294,
      "step": 767
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9055858937990083e-05,
      "loss": 0.531,
      "step": 768
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9052898149285494e-05,
      "loss": 0.5058,
      "step": 769
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9049932956141917e-05,
      "loss": 0.5519,
      "step": 770
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9046963360001984e-05,
      "loss": 0.519,
      "step": 771
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9043989362310472e-05,
      "loss": 0.5345,
      "step": 772
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9041010964514296e-05,
      "loss": 0.5451,
      "step": 773
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9038028168062517e-05,
      "loss": 0.5131,
      "step": 774
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.903504097440633e-05,
      "loss": 0.5199,
      "step": 775
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.903204938499907e-05,
      "loss": 0.5356,
      "step": 776
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9029053401296216e-05,
      "loss": 0.5213,
      "step": 777
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9026053024755384e-05,
      "loss": 0.5153,
      "step": 778
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.902304825683632e-05,
      "loss": 0.5268,
      "step": 779
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.902003909900091e-05,
      "loss": 0.549,
      "step": 780
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9017025552713182e-05,
      "loss": 0.5052,
      "step": 781
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.901400761943929e-05,
      "loss": 0.5019,
      "step": 782
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9010985300647528e-05,
      "loss": 0.4966,
      "step": 783
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9007958597808326e-05,
      "loss": 0.5468,
      "step": 784
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.900492751239424e-05,
      "loss": 0.5281,
      "step": 785
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9001892045879963e-05,
      "loss": 0.5306,
      "step": 786
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8998852199742313e-05,
      "loss": 0.5249,
      "step": 787
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8995807975460246e-05,
      "loss": 0.5531,
      "step": 788
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.899275937451485e-05,
      "loss": 0.5122,
      "step": 789
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8989706398389335e-05,
      "loss": 0.5186,
      "step": 790
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8986649048569038e-05,
      "loss": 0.5046,
      "step": 791
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8983587326541437e-05,
      "loss": 0.5123,
      "step": 792
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8980521233796123e-05,
      "loss": 0.4946,
      "step": 793
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.897745077182482e-05,
      "loss": 0.5415,
      "step": 794
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.897437594212138e-05,
      "loss": 0.5111,
      "step": 795
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8971296746181774e-05,
      "loss": 0.5257,
      "step": 796
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8968213185504103e-05,
      "loss": 0.4849,
      "step": 797
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8965125261588586e-05,
      "loss": 0.5204,
      "step": 798
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.896203297593757e-05,
      "loss": 0.5327,
      "step": 799
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8958936330055516e-05,
      "loss": 0.5252,
      "step": 800
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.895583532544902e-05,
      "loss": 0.4923,
      "step": 801
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8952729963626783e-05,
      "loss": 0.5354,
      "step": 802
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.894962024609964e-05,
      "loss": 0.5267,
      "step": 803
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.894650617438054e-05,
      "loss": 0.5327,
      "step": 804
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8943387749984537e-05,
      "loss": 0.5399,
      "step": 805
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8940264974428827e-05,
      "loss": 0.5272,
      "step": 806
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8937137849232705e-05,
      "loss": 0.5367,
      "step": 807
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.893400637591759e-05,
      "loss": 0.5386,
      "step": 808
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8930870556007016e-05,
      "loss": 0.4968,
      "step": 809
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8927730391026625e-05,
      "loss": 0.5157,
      "step": 810
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8924585882504184e-05,
      "loss": 0.5165,
      "step": 811
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8921437031969557e-05,
      "loss": 0.5248,
      "step": 812
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8918283840954745e-05,
      "loss": 0.4947,
      "step": 813
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8915126310993838e-05,
      "loss": 0.5429,
      "step": 814
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8911964443623046e-05,
      "loss": 0.5162,
      "step": 815
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8908798240380692e-05,
      "loss": 0.5193,
      "step": 816
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.89056277028072e-05,
      "loss": 0.5131,
      "step": 817
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.890245283244511e-05,
      "loss": 0.5213,
      "step": 818
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8899273630839073e-05,
      "loss": 0.5522,
      "step": 819
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8896090099535834e-05,
      "loss": 0.5026,
      "step": 820
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8892902240084257e-05,
      "loss": 0.5243,
      "step": 821
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.888971005403531e-05,
      "loss": 0.5508,
      "step": 822
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.888651354294206e-05,
      "loss": 0.5223,
      "step": 823
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.888331270835968e-05,
      "loss": 0.539,
      "step": 824
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8880107551845453e-05,
      "loss": 0.5118,
      "step": 825
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8876898074958757e-05,
      "loss": 0.5131,
      "step": 826
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8873684279261076e-05,
      "loss": 0.5211,
      "step": 827
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8870466166315992e-05,
      "loss": 0.5404,
      "step": 828
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.886724373768919e-05,
      "loss": 0.5187,
      "step": 829
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8864016994948456e-05,
      "loss": 0.5178,
      "step": 830
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8860785939663674e-05,
      "loss": 0.4933,
      "step": 831
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.885755057340682e-05,
      "loss": 0.5214,
      "step": 832
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8854310897751976e-05,
      "loss": 0.4929,
      "step": 833
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.885106691427532e-05,
      "loss": 0.5325,
      "step": 834
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8847818624555117e-05,
      "loss": 0.4969,
      "step": 835
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8844566030171737e-05,
      "loss": 0.5249,
      "step": 836
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8841309132707644e-05,
      "loss": 0.5082,
      "step": 837
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8838047933747386e-05,
      "loss": 0.5319,
      "step": 838
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.883478243487761e-05,
      "loss": 0.5262,
      "step": 839
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8831512637687054e-05,
      "loss": 0.5309,
      "step": 840
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8828238543766554e-05,
      "loss": 0.5136,
      "step": 841
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8824960154709027e-05,
      "loss": 0.5523,
      "step": 842
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8821677472109483e-05,
      "loss": 0.5047,
      "step": 843
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.881839049756502e-05,
      "loss": 0.5187,
      "step": 844
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8815099232674827e-05,
      "loss": 0.5503,
      "step": 845
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8811803679040178e-05,
      "loss": 0.5142,
      "step": 846
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8808503838264435e-05,
      "loss": 0.5394,
      "step": 847
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.880519971195304e-05,
      "loss": 0.5194,
      "step": 848
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8801891301713535e-05,
      "loss": 0.505,
      "step": 849
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8798578609155528e-05,
      "loss": 0.55,
      "step": 850
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8795261635890716e-05,
      "loss": 0.5265,
      "step": 851
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.879194038353289e-05,
      "loss": 0.525,
      "step": 852
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.878861485369791e-05,
      "loss": 0.5396,
      "step": 853
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8785285048003722e-05,
      "loss": 0.4973,
      "step": 854
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8781950968070348e-05,
      "loss": 0.5222,
      "step": 855
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.87786126155199e-05,
      "loss": 0.5353,
      "step": 856
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8775269991976552e-05,
      "loss": 0.5061,
      "step": 857
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8771923099066573e-05,
      "loss": 0.5195,
      "step": 858
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.87685719384183e-05,
      "loss": 0.519,
      "step": 859
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8765216511662153e-05,
      "loss": 0.5245,
      "step": 860
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8761856820430614e-05,
      "loss": 0.4968,
      "step": 861
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.875849286635825e-05,
      "loss": 0.533,
      "step": 862
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.875512465108171e-05,
      "loss": 0.5315,
      "step": 863
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8751752176239693e-05,
      "loss": 0.5213,
      "step": 864
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8748375443472993e-05,
      "loss": 0.5323,
      "step": 865
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8744994454424463e-05,
      "loss": 0.5143,
      "step": 866
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8741609210739034e-05,
      "loss": 0.5116,
      "step": 867
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.87382197140637e-05,
      "loss": 0.5158,
      "step": 868
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8734825966047528e-05,
      "loss": 0.5482,
      "step": 869
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8731427968341654e-05,
      "loss": 0.5203,
      "step": 870
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.872802572259928e-05,
      "loss": 0.5153,
      "step": 871
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8724619230475675e-05,
      "loss": 0.5275,
      "step": 872
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8721208493628174e-05,
      "loss": 0.5034,
      "step": 873
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.871779351371618e-05,
      "loss": 0.5227,
      "step": 874
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.871437429240115e-05,
      "loss": 0.5195,
      "step": 875
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8710950831346623e-05,
      "loss": 0.5445,
      "step": 876
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.870752313221818e-05,
      "loss": 0.5076,
      "step": 877
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.870409119668348e-05,
      "loss": 0.53,
      "step": 878
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8700655026412238e-05,
      "loss": 0.5135,
      "step": 879
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8697214623076222e-05,
      "loss": 0.5071,
      "step": 880
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8693769988349274e-05,
      "loss": 0.5328,
      "step": 881
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8690321123907277e-05,
      "loss": 0.4843,
      "step": 882
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.868686803142819e-05,
      "loss": 0.4909,
      "step": 883
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8683410712592015e-05,
      "loss": 0.5035,
      "step": 884
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8679949169080815e-05,
      "loss": 0.5364,
      "step": 885
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8676483402578714e-05,
      "loss": 0.4987,
      "step": 886
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.867301341477188e-05,
      "loss": 0.5152,
      "step": 887
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8669539207348544e-05,
      "loss": 0.5486,
      "step": 888
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8666060781998982e-05,
      "loss": 0.5312,
      "step": 889
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8662578140415535e-05,
      "loss": 0.5142,
      "step": 890
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.865909128429258e-05,
      "loss": 0.5089,
      "step": 891
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8655600215326547e-05,
      "loss": 0.5301,
      "step": 892
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.865210493521593e-05,
      "loss": 0.5009,
      "step": 893
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8648605445661256e-05,
      "loss": 0.5298,
      "step": 894
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.86451017483651e-05,
      "loss": 0.5371,
      "step": 895
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8641593845032098e-05,
      "loss": 0.5245,
      "step": 896
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8638081737368928e-05,
      "loss": 0.5663,
      "step": 897
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8634565427084295e-05,
      "loss": 0.5337,
      "step": 898
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8631044915888976e-05,
      "loss": 0.5119,
      "step": 899
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8627520205495772e-05,
      "loss": 0.5105,
      "step": 900
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8623991297619536e-05,
      "loss": 0.5125,
      "step": 901
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8620458193977166e-05,
      "loss": 0.529,
      "step": 902
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.861692089628759e-05,
      "loss": 0.526,
      "step": 903
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8613379406271784e-05,
      "loss": 0.5226,
      "step": 904
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8609833725652763e-05,
      "loss": 0.5131,
      "step": 905
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8606283856155585e-05,
      "loss": 0.486,
      "step": 906
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8602729799507335e-05,
      "loss": 0.5129,
      "step": 907
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8599171557437147e-05,
      "loss": 0.5388,
      "step": 908
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.859560913167618e-05,
      "loss": 0.5212,
      "step": 909
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.859204252395764e-05,
      "loss": 0.5015,
      "step": 910
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.858847173601676e-05,
      "loss": 0.4776,
      "step": 911
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.858489676959081e-05,
      "loss": 0.5232,
      "step": 912
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.858131762641909e-05,
      "loss": 0.5153,
      "step": 913
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8577734308242936e-05,
      "loss": 0.5099,
      "step": 914
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.857414681680571e-05,
      "loss": 0.5218,
      "step": 915
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8570555153852806e-05,
      "loss": 0.515,
      "step": 916
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8566959321131656e-05,
      "loss": 0.506,
      "step": 917
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.856335932039171e-05,
      "loss": 0.5418,
      "step": 918
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8559755153384448e-05,
      "loss": 0.495,
      "step": 919
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.855614682186338e-05,
      "loss": 0.4858,
      "step": 920
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8552534327584038e-05,
      "loss": 0.541,
      "step": 921
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8548917672303987e-05,
      "loss": 0.5309,
      "step": 922
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8545296857782808e-05,
      "loss": 0.5,
      "step": 923
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8541671885782106e-05,
      "loss": 0.5141,
      "step": 924
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.853804275806552e-05,
      "loss": 0.5458,
      "step": 925
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8534409476398693e-05,
      "loss": 0.5165,
      "step": 926
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8530772042549303e-05,
      "loss": 0.5256,
      "step": 927
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8527130458287047e-05,
      "loss": 0.5246,
      "step": 928
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.852348472538363e-05,
      "loss": 0.5105,
      "step": 929
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.851983484561279e-05,
      "loss": 0.5209,
      "step": 930
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.851618082075027e-05,
      "loss": 0.5114,
      "step": 931
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.851252265257384e-05,
      "loss": 0.5475,
      "step": 932
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8508860342863283e-05,
      "loss": 0.5389,
      "step": 933
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.850519389340039e-05,
      "loss": 0.5055,
      "step": 934
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.850152330596897e-05,
      "loss": 0.5156,
      "step": 935
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8497848582354852e-05,
      "loss": 0.5216,
      "step": 936
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8494169724345868e-05,
      "loss": 0.5077,
      "step": 937
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.849048673373187e-05,
      "loss": 0.5299,
      "step": 938
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.848679961230471e-05,
      "loss": 0.5328,
      "step": 939
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8483108361858263e-05,
      "loss": 0.5339,
      "step": 940
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.84794129841884e-05,
      "loss": 0.529,
      "step": 941
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8475713481093005e-05,
      "loss": 0.5005,
      "step": 942
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8472009854371973e-05,
      "loss": 0.4962,
      "step": 943
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8468302105827195e-05,
      "loss": 0.5326,
      "step": 944
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8464590237262584e-05,
      "loss": 0.5108,
      "step": 945
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8460874250484045e-05,
      "loss": 0.5204,
      "step": 946
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8457154147299484e-05,
      "loss": 0.5247,
      "step": 947
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.845342992951882e-05,
      "loss": 0.5007,
      "step": 948
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8449701598953965e-05,
      "loss": 0.5375,
      "step": 949
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8445969157418845e-05,
      "loss": 0.4983,
      "step": 950
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8442232606729363e-05,
      "loss": 0.5422,
      "step": 951
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8438491948703445e-05,
      "loss": 0.4993,
      "step": 952
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.8434747185161003e-05,
      "loss": 0.5387,
      "step": 953
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.843099831792395e-05,
      "loss": 0.5283,
      "step": 954
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.842724534881619e-05,
      "loss": 0.5332,
      "step": 955
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.842348827966363e-05,
      "loss": 0.5272,
      "step": 956
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.841972711229417e-05,
      "loss": 0.5037,
      "step": 957
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.8415961848537702e-05,
      "loss": 0.5511,
      "step": 958
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.8412192490226113e-05,
      "loss": 0.5265,
      "step": 959
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.840841903919328e-05,
      "loss": 0.4896,
      "step": 960
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.8404641497275065e-05,
      "loss": 0.5332,
      "step": 961
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.8400859866309337e-05,
      "loss": 0.5106,
      "step": 962
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.839707414813594e-05,
      "loss": 0.5406,
      "step": 963
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.8393284344596715e-05,
      "loss": 0.5167,
      "step": 964
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.8389490457535482e-05,
      "loss": 0.5003,
      "step": 965
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.8385692488798056e-05,
      "loss": 0.5014,
      "step": 966
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.838189044023223e-05,
      "loss": 0.5343,
      "step": 967
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.837808431368779e-05,
      "loss": 0.5372,
      "step": 968
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8374274111016502e-05,
      "loss": 0.5435,
      "step": 969
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8370459834072118e-05,
      "loss": 0.5276,
      "step": 970
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8366641484710362e-05,
      "loss": 0.5148,
      "step": 971
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8362819064788956e-05,
      "loss": 0.534,
      "step": 972
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.835899257616759e-05,
      "loss": 0.5043,
      "step": 973
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8355162020707932e-05,
      "loss": 0.4994,
      "step": 974
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.835132740027364e-05,
      "loss": 0.5301,
      "step": 975
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8347488716730343e-05,
      "loss": 0.5348,
      "step": 976
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8343645971945646e-05,
      "loss": 0.5275,
      "step": 977
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8339799167789127e-05,
      "loss": 0.5139,
      "step": 978
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8335948306132344e-05,
      "loss": 0.5405,
      "step": 979
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8332093388848836e-05,
      "loss": 0.4917,
      "step": 980
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8328234417814097e-05,
      "loss": 0.5296,
      "step": 981
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8324371394905606e-05,
      "loss": 0.5105,
      "step": 982
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8320504322002817e-05,
      "loss": 0.4956,
      "step": 983
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8316633200987143e-05,
      "loss": 0.5166,
      "step": 984
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8312758033741968e-05,
      "loss": 0.525,
      "step": 985
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8308878822152655e-05,
      "loss": 0.5255,
      "step": 986
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8304995568106524e-05,
      "loss": 0.5259,
      "step": 987
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.830110827349287e-05,
      "loss": 0.4996,
      "step": 988
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8297216940202945e-05,
      "loss": 0.5129,
      "step": 989
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.829332157012997e-05,
      "loss": 0.526,
      "step": 990
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.828942216516913e-05,
      "loss": 0.4995,
      "step": 991
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8285518727217578e-05,
      "loss": 0.5211,
      "step": 992
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.828161125817442e-05,
      "loss": 0.5458,
      "step": 993
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8277699759940732e-05,
      "loss": 0.4904,
      "step": 994
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8273784234419544e-05,
      "loss": 0.5224,
      "step": 995
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8269864683515847e-05,
      "loss": 0.527,
      "step": 996
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8265941109136586e-05,
      "loss": 0.4823,
      "step": 997
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8262013513190677e-05,
      "loss": 0.5046,
      "step": 998
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8258081897588978e-05,
      "loss": 0.536,
      "step": 999
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.8254146264244316e-05,
      "loss": 0.5157,
      "step": 1000
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.8250206615071457e-05,
      "loss": 0.5282,
      "step": 1001
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.824626295198713e-05,
      "loss": 0.4845,
      "step": 1002
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.824231527691002e-05,
      "loss": 0.5398,
      "step": 1003
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.8238363591760758e-05,
      "loss": 0.494,
      "step": 1004
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.823440789846193e-05,
      "loss": 0.5336,
      "step": 1005
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.8230448198938067e-05,
      "loss": 0.5056,
      "step": 1006
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.822648449511565e-05,
      "loss": 0.5039,
      "step": 1007
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.822251678892312e-05,
      "loss": 0.5233,
      "step": 1008
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.821854508229084e-05,
      "loss": 0.4799,
      "step": 1009
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.821456937715115e-05,
      "loss": 0.5399,
      "step": 1010
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.8210589675438314e-05,
      "loss": 0.5069,
      "step": 1011
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.8206605979088545e-05,
      "loss": 0.5176,
      "step": 1012
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.820261829004e-05,
      "loss": 0.5164,
      "step": 1013
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.819862661023278e-05,
      "loss": 0.5183,
      "step": 1014
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8194630941608932e-05,
      "loss": 0.4893,
      "step": 1015
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.819063128611244e-05,
      "loss": 0.4829,
      "step": 1016
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8186627645689218e-05,
      "loss": 0.5074,
      "step": 1017
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8182620022287133e-05,
      "loss": 0.5417,
      "step": 1018
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8178608417855984e-05,
      "loss": 0.5348,
      "step": 1019
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8174592834347503e-05,
      "loss": 0.5378,
      "step": 1020
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8170573273715368e-05,
      "loss": 0.5191,
      "step": 1021
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.816654973791518e-05,
      "loss": 0.4881,
      "step": 1022
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8162522228904485e-05,
      "loss": 0.5145,
      "step": 1023
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.815849074864275e-05,
      "loss": 0.5256,
      "step": 1024
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.815445529909139e-05,
      "loss": 0.5284,
      "step": 1025
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8150415882213735e-05,
      "loss": 0.4991,
      "step": 1026
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8146372499975058e-05,
      "loss": 0.513,
      "step": 1027
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.814232515434255e-05,
      "loss": 0.5022,
      "step": 1028
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8138273847285338e-05,
      "loss": 0.5231,
      "step": 1029
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8134218580774475e-05,
      "loss": 0.5012,
      "step": 1030
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.813015935678294e-05,
      "loss": 0.5237,
      "step": 1031
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8126096177285637e-05,
      "loss": 0.5188,
      "step": 1032
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8122029044259392e-05,
      "loss": 0.5052,
      "step": 1033
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.811795795968296e-05,
      "loss": 0.5018,
      "step": 1034
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8113882925537008e-05,
      "loss": 0.5126,
      "step": 1035
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8109803943804146e-05,
      "loss": 0.5115,
      "step": 1036
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8105721016468878e-05,
      "loss": 0.5347,
      "step": 1037
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8101634145517644e-05,
      "loss": 0.5126,
      "step": 1038
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.80975433329388e-05,
      "loss": 0.5369,
      "step": 1039
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8093448580722617e-05,
      "loss": 0.5032,
      "step": 1040
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.808934989086129e-05,
      "loss": 0.4957,
      "step": 1041
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8085247265348913e-05,
      "loss": 0.5294,
      "step": 1042
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8081140706181514e-05,
      "loss": 0.4911,
      "step": 1043
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8077030215357024e-05,
      "loss": 0.5171,
      "step": 1044
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.807291579487529e-05,
      "loss": 0.5337,
      "step": 1045
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8068797446738072e-05,
      "loss": 0.5144,
      "step": 1046
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8064675172949035e-05,
      "loss": 0.5479,
      "step": 1047
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.806054897551376e-05,
      "loss": 0.5253,
      "step": 1048
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8056418856439737e-05,
      "loss": 0.5048,
      "step": 1049
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.805228481773636e-05,
      "loss": 0.4952,
      "step": 1050
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8048146861414933e-05,
      "loss": 0.4907,
      "step": 1051
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8044004989488662e-05,
      "loss": 0.5216,
      "step": 1052
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.803985920397267e-05,
      "loss": 0.5049,
      "step": 1053
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8035709506883962e-05,
      "loss": 0.5135,
      "step": 1054
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.803155590024147e-05,
      "loss": 0.5291,
      "step": 1055
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.802739838606601e-05,
      "loss": 0.5104,
      "step": 1056
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8023236966380312e-05,
      "loss": 0.5054,
      "step": 1057
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8019071643208996e-05,
      "loss": 0.5083,
      "step": 1058
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8014902418578588e-05,
      "loss": 0.534,
      "step": 1059
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.801072929451751e-05,
      "loss": 0.5085,
      "step": 1060
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.800655227305608e-05,
      "loss": 0.523,
      "step": 1061
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.8002371356226512e-05,
      "loss": 0.519,
      "step": 1062
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.799818654606292e-05,
      "loss": 0.5452,
      "step": 1063
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.7993997844601305e-05,
      "loss": 0.504,
      "step": 1064
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.798980525387956e-05,
      "loss": 0.4886,
      "step": 1065
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.7985608775937492e-05,
      "loss": 0.5289,
      "step": 1066
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.7981408412816762e-05,
      "loss": 0.5087,
      "step": 1067
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.7977204166560954e-05,
      "loss": 0.513,
      "step": 1068
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.7972996039215523e-05,
      "loss": 0.4867,
      "step": 1069
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.796878403282782e-05,
      "loss": 0.5551,
      "step": 1070
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.7964568149447078e-05,
      "loss": 0.5326,
      "step": 1071
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.7960348391124422e-05,
      "loss": 0.5351,
      "step": 1072
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.7956124759912862e-05,
      "loss": 0.5282,
      "step": 1073
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.7951897257867284e-05,
      "loss": 0.5109,
      "step": 1074
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.7947665887044468e-05,
      "loss": 0.4869,
      "step": 1075
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.7943430649503065e-05,
      "loss": 0.5301,
      "step": 1076
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.793919154730362e-05,
      "loss": 0.5078,
      "step": 1077
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7934948582508554e-05,
      "loss": 0.5214,
      "step": 1078
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.793070175718216e-05,
      "loss": 0.5383,
      "step": 1079
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7926451073390612e-05,
      "loss": 0.5209,
      "step": 1080
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7922196533201967e-05,
      "loss": 0.5077,
      "step": 1081
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7917938138686152e-05,
      "loss": 0.5375,
      "step": 1082
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7913675891914976e-05,
      "loss": 0.5061,
      "step": 1083
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7909409794962115e-05,
      "loss": 0.5184,
      "step": 1084
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.790513984990312e-05,
      "loss": 0.4921,
      "step": 1085
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7900866058815424e-05,
      "loss": 0.5379,
      "step": 1086
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.789658842377831e-05,
      "loss": 0.5047,
      "step": 1087
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7892306946872952e-05,
      "loss": 0.471,
      "step": 1088
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7888021630182385e-05,
      "loss": 0.5089,
      "step": 1089
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.7883732475791512e-05,
      "loss": 0.4987,
      "step": 1090
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.78794394857871e-05,
      "loss": 0.4782,
      "step": 1091
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7875142662257788e-05,
      "loss": 0.5214,
      "step": 1092
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7870842007294075e-05,
      "loss": 0.5142,
      "step": 1093
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.786653752298833e-05,
      "loss": 0.5426,
      "step": 1094
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7862229211434778e-05,
      "loss": 0.5114,
      "step": 1095
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7857917074729513e-05,
      "loss": 0.5259,
      "step": 1096
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7853601114970483e-05,
      "loss": 0.4828,
      "step": 1097
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7849281334257504e-05,
      "loss": 0.5044,
      "step": 1098
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7844957734692247e-05,
      "loss": 0.5169,
      "step": 1099
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7840630318378233e-05,
      "loss": 0.5216,
      "step": 1100
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7836299087420856e-05,
      "loss": 0.5524,
      "step": 1101
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7831964043927355e-05,
      "loss": 0.5026,
      "step": 1102
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7827625190006824e-05,
      "loss": 0.5217,
      "step": 1103
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7823282527770214e-05,
      "loss": 0.5167,
      "step": 1104
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7818936059330333e-05,
      "loss": 0.4926,
      "step": 1105
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.7814585786801826e-05,
      "loss": 0.5173,
      "step": 1106
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.781023171230121e-05,
      "loss": 0.5072,
      "step": 1107
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7805873837946833e-05,
      "loss": 0.5005,
      "step": 1108
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.78015121658589e-05,
      "loss": 0.5084,
      "step": 1109
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.779714669815946e-05,
      "loss": 0.4994,
      "step": 1110
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7792777436972423e-05,
      "loss": 0.5261,
      "step": 1111
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.778840438442352e-05,
      "loss": 0.5068,
      "step": 1112
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7784027542640346e-05,
      "loss": 0.5095,
      "step": 1113
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7779646913752334e-05,
      "loss": 0.5227,
      "step": 1114
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7775262499890753e-05,
      "loss": 0.5273,
      "step": 1115
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7770874303188727e-05,
      "loss": 0.5188,
      "step": 1116
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7766482325781205e-05,
      "loss": 0.504,
      "step": 1117
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.776208656980499e-05,
      "loss": 0.5027,
      "step": 1118
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.775768703739871e-05,
      "loss": 0.5304,
      "step": 1119
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7753283730702837e-05,
      "loss": 0.5231,
      "step": 1120
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7748876651859686e-05,
      "loss": 0.5142,
      "step": 1121
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.7744465803013394e-05,
      "loss": 0.5475,
      "step": 1122
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7740051186309937e-05,
      "loss": 0.538,
      "step": 1123
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7735632803897135e-05,
      "loss": 0.5015,
      "step": 1124
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.773121065792462e-05,
      "loss": 0.5445,
      "step": 1125
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7726784750543867e-05,
      "loss": 0.5315,
      "step": 1126
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7722355083908187e-05,
      "loss": 0.5043,
      "step": 1127
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7717921660172708e-05,
      "loss": 0.5245,
      "step": 1128
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.771348448149439e-05,
      "loss": 0.5187,
      "step": 1129
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.770904355003202e-05,
      "loss": 0.54,
      "step": 1130
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7704598867946212e-05,
      "loss": 0.516,
      "step": 1131
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7700150437399405e-05,
      "loss": 0.5358,
      "step": 1132
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.769569826055586e-05,
      "loss": 0.5294,
      "step": 1133
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7691242339581664e-05,
      "loss": 0.4928,
      "step": 1134
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.768678267664472e-05,
      "loss": 0.49,
      "step": 1135
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7682319273914755e-05,
      "loss": 0.4948,
      "step": 1136
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.7677852133563314e-05,
      "loss": 0.5264,
      "step": 1137
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.767338125776376e-05,
      "loss": 0.5141,
      "step": 1138
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.766890664869128e-05,
      "loss": 0.5175,
      "step": 1139
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.766442830852287e-05,
      "loss": 0.5031,
      "step": 1140
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.765994623943734e-05,
      "loss": 0.516,
      "step": 1141
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.7655460443615327e-05,
      "loss": 0.4813,
      "step": 1142
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.765097092323926e-05,
      "loss": 0.5086,
      "step": 1143
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.76464776804934e-05,
      "loss": 0.5578,
      "step": 1144
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.7641980717563806e-05,
      "loss": 0.5103,
      "step": 1145
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.7637480036638356e-05,
      "loss": 0.5266,
      "step": 1146
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.7632975639906727e-05,
      "loss": 0.5198,
      "step": 1147
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.7628467529560417e-05,
      "loss": 0.5132,
      "step": 1148
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.7623955707792717e-05,
      "loss": 0.5124,
      "step": 1149
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.7619440176798733e-05,
      "loss": 0.5161,
      "step": 1150
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.761492093877537e-05,
      "loss": 0.511,
      "step": 1151
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.7610397995921348e-05,
      "loss": 0.4985,
      "step": 1152
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.7605871350437165e-05,
      "loss": 0.527,
      "step": 1153
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.760134100452515e-05,
      "loss": 0.5057,
      "step": 1154
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.7596806960389415e-05,
      "loss": 0.5129,
      "step": 1155
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.759226922023587e-05,
      "loss": 0.5186,
      "step": 1156
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.758772778627223e-05,
      "loss": 0.543,
      "step": 1157
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.758318266070801e-05,
      "loss": 0.5008,
      "step": 1158
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.757863384575451e-05,
      "loss": 0.4937,
      "step": 1159
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.7574081343624827e-05,
      "loss": 0.5187,
      "step": 1160
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.7569525156533864e-05,
      "loss": 0.5193,
      "step": 1161
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.756496528669831e-05,
      "loss": 0.4996,
      "step": 1162
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.7560401736336642e-05,
      "loss": 0.503,
      "step": 1163
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.7555834507669124e-05,
      "loss": 0.5226,
      "step": 1164
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.7551263602917823e-05,
      "loss": 0.4975,
      "step": 1165
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.7546689024306585e-05,
      "loss": 0.5039,
      "step": 1166
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.7542110774061045e-05,
      "loss": 0.5548,
      "step": 1167
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.7537528854408625e-05,
      "loss": 0.5324,
      "step": 1168
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.7532943267578532e-05,
      "loss": 0.5187,
      "step": 1169
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.752835401580176e-05,
      "loss": 0.5137,
      "step": 1170
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.752376110131108e-05,
      "loss": 0.5121,
      "step": 1171
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.751916452634105e-05,
      "loss": 0.5006,
      "step": 1172
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.7514564293128008e-05,
      "loss": 0.5244,
      "step": 1173
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.750996040391007e-05,
      "loss": 0.5006,
      "step": 1174
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.750535286092713e-05,
      "loss": 0.5276,
      "step": 1175
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.7500741666420863e-05,
      "loss": 0.4907,
      "step": 1176
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.7496126822634723e-05,
      "loss": 0.5101,
      "step": 1177
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.7491508331813928e-05,
      "loss": 0.5108,
      "step": 1178
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.7486886196205487e-05,
      "loss": 0.5123,
      "step": 1179
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.7482260418058167e-05,
      "loss": 0.5297,
      "step": 1180
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.747763099962251e-05,
      "loss": 0.5126,
      "step": 1181
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.747299794315084e-05,
      "loss": 0.5374,
      "step": 1182
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.746836125089724e-05,
      "loss": 0.4925,
      "step": 1183
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.7463720925117565e-05,
      "loss": 0.5104,
      "step": 1184
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.745907696806944e-05,
      "loss": 0.5239,
      "step": 1185
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.7454429382012255e-05,
      "loss": 0.5216,
      "step": 1186
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.744977816920716e-05,
      "loss": 0.5455,
      "step": 1187
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.744512333191708e-05,
      "loss": 0.517,
      "step": 1188
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.74404648724067e-05,
      "loss": 0.5106,
      "step": 1189
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.743580279294246e-05,
      "loss": 0.4949,
      "step": 1190
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.7431137095792566e-05,
      "loss": 0.4996,
      "step": 1191
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.7426467783226992e-05,
      "loss": 0.5081,
      "step": 1192
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.7421794857517457e-05,
      "loss": 0.5195,
      "step": 1193
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.7417118320937452e-05,
      "loss": 0.5183,
      "step": 1194
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.7412438175762206e-05,
      "loss": 0.5022,
      "step": 1195
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.7407754424268727e-05,
      "loss": 0.5088,
      "step": 1196
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.740306706873576e-05,
      "loss": 0.5048,
      "step": 1197
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.739837611144381e-05,
      "loss": 0.5093,
      "step": 1198
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.7393681554675135e-05,
      "loss": 0.5248,
      "step": 1199
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.7388983400713736e-05,
      "loss": 0.4883,
      "step": 1200
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.738428165184538e-05,
      "loss": 0.5036,
      "step": 1201
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7379576310357568e-05,
      "loss": 0.5281,
      "step": 1202
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7374867378539558e-05,
      "loss": 0.5454,
      "step": 1203
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7370154858682347e-05,
      "loss": 0.5094,
      "step": 1204
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7365438753078686e-05,
      "loss": 0.5013,
      "step": 1205
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7360719064023067e-05,
      "loss": 0.5153,
      "step": 1206
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.735599579381172e-05,
      "loss": 0.5335,
      "step": 1207
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7351268944742626e-05,
      "loss": 0.509,
      "step": 1208
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7346538519115507e-05,
      "loss": 0.4957,
      "step": 1209
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7341804519231815e-05,
      "loss": 0.5072,
      "step": 1210
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7337066947394755e-05,
      "loss": 0.5127,
      "step": 1211
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7332325805909256e-05,
      "loss": 0.4991,
      "step": 1212
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.732758109708199e-05,
      "loss": 0.5347,
      "step": 1213
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.732283282322137e-05,
      "loss": 0.5175,
      "step": 1214
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.7318080986637538e-05,
      "loss": 0.4978,
      "step": 1215
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7313325589642363e-05,
      "loss": 0.4965,
      "step": 1216
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.730856663454946e-05,
      "loss": 0.5195,
      "step": 1217
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7303804123674165e-05,
      "loss": 0.5201,
      "step": 1218
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7299038059333545e-05,
      "loss": 0.5059,
      "step": 1219
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7294268443846403e-05,
      "loss": 0.5275,
      "step": 1220
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7289495279533257e-05,
      "loss": 0.4882,
      "step": 1221
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7284718568716362e-05,
      "loss": 0.5261,
      "step": 1222
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7279938313719703e-05,
      "loss": 0.518,
      "step": 1223
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.727515451686897e-05,
      "loss": 0.5007,
      "step": 1224
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7270367180491594e-05,
      "loss": 0.5089,
      "step": 1225
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.726557630691672e-05,
      "loss": 0.5096,
      "step": 1226
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.726078189847522e-05,
      "loss": 0.4808,
      "step": 1227
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7255983957499676e-05,
      "loss": 0.5111,
      "step": 1228
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7251182486324403e-05,
      "loss": 0.5067,
      "step": 1229
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.7246377487285415e-05,
      "loss": 0.5149,
      "step": 1230
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.724156896272046e-05,
      "loss": 0.5086,
      "step": 1231
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.7236756914968985e-05,
      "loss": 0.518,
      "step": 1232
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.723194134637217e-05,
      "loss": 0.4971,
      "step": 1233
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.722712225927289e-05,
      "loss": 0.5038,
      "step": 1234
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.7222299656015742e-05,
      "loss": 0.5244,
      "step": 1235
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.7217473538947032e-05,
      "loss": 0.5114,
      "step": 1236
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.7212643910414776e-05,
      "loss": 0.5074,
      "step": 1237
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.7207810772768692e-05,
      "loss": 0.4893,
      "step": 1238
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.7202974128360213e-05,
      "loss": 0.5326,
      "step": 1239
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.719813397954248e-05,
      "loss": 0.5206,
      "step": 1240
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.7193290328670328e-05,
      "loss": 0.5236,
      "step": 1241
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.7188443178100306e-05,
      "loss": 0.5103,
      "step": 1242
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.7183592530190662e-05,
      "loss": 0.5098,
      "step": 1243
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.7178738387301342e-05,
      "loss": 0.5331,
      "step": 1244
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.7173880751794e-05,
      "loss": 0.5452,
      "step": 1245
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.7169019626031985e-05,
      "loss": 0.5106,
      "step": 1246
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.716415501238034e-05,
      "loss": 0.5172,
      "step": 1247
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7159286913205813e-05,
      "loss": 0.4953,
      "step": 1248
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.715441533087684e-05,
      "loss": 0.5438,
      "step": 1249
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7149540267763566e-05,
      "loss": 0.5049,
      "step": 1250
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7144661726237804e-05,
      "loss": 0.5036,
      "step": 1251
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7139779708673084e-05,
      "loss": 0.5386,
      "step": 1252
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7134894217444617e-05,
      "loss": 0.5092,
      "step": 1253
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.71300052549293e-05,
      "loss": 0.5154,
      "step": 1254
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7125112823505736e-05,
      "loss": 0.5381,
      "step": 1255
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7120216925554185e-05,
      "loss": 0.5169,
      "step": 1256
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.711531756345662e-05,
      "loss": 0.5025,
      "step": 1257
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7110414739596697e-05,
      "loss": 0.5171,
      "step": 1258
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7105508456359742e-05,
      "loss": 0.5183,
      "step": 1259
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7100598716132775e-05,
      "loss": 0.5098,
      "step": 1260
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7095685521304498e-05,
      "loss": 0.5048,
      "step": 1261
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.7090768874265285e-05,
      "loss": 0.5032,
      "step": 1262
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.70858487774072e-05,
      "loss": 0.5145,
      "step": 1263
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.708092523312398e-05,
      "loss": 0.5152,
      "step": 1264
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7075998243811043e-05,
      "loss": 0.5217,
      "step": 1265
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7071067811865477e-05,
      "loss": 0.5208,
      "step": 1266
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.706613393968605e-05,
      "loss": 0.5141,
      "step": 1267
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7061196629673198e-05,
      "loss": 0.5043,
      "step": 1268
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.705625588422904e-05,
      "loss": 0.5179,
      "step": 1269
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7051311705757353e-05,
      "loss": 0.5049,
      "step": 1270
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7046364096663598e-05,
      "loss": 0.5058,
      "step": 1271
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7041413059354893e-05,
      "loss": 0.4851,
      "step": 1272
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.703645859624003e-05,
      "loss": 0.5148,
      "step": 1273
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.703150070972947e-05,
      "loss": 0.5357,
      "step": 1274
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.702653940223534e-05,
      "loss": 0.518,
      "step": 1275
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.7021574676171418e-05,
      "loss": 0.4938,
      "step": 1276
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.701660653395316e-05,
      "loss": 0.5286,
      "step": 1277
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.7011634977997683e-05,
      "loss": 0.5147,
      "step": 1278
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.7006660010723754e-05,
      "loss": 0.5299,
      "step": 1279
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.7001681634551813e-05,
      "loss": 0.5135,
      "step": 1280
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.6996699851903942e-05,
      "loss": 0.5311,
      "step": 1281
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.6991714665203905e-05,
      "loss": 0.5238,
      "step": 1282
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.69867260768771e-05,
      "loss": 0.4854,
      "step": 1283
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.6981734089350585e-05,
      "loss": 0.4983,
      "step": 1284
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.6976738705053078e-05,
      "loss": 0.5203,
      "step": 1285
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.6971739926414946e-05,
      "loss": 0.4819,
      "step": 1286
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.6966737755868207e-05,
      "loss": 0.5458,
      "step": 1287
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.696173219584653e-05,
      "loss": 0.5173,
      "step": 1288
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.6956723248785235e-05,
      "loss": 0.4975,
      "step": 1289
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.695171091712128e-05,
      "loss": 0.4983,
      "step": 1290
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.6946695203293286e-05,
      "loss": 0.5362,
      "step": 1291
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.6941676109741506e-05,
      "loss": 0.504,
      "step": 1292
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.6936653638907846e-05,
      "loss": 0.5189,
      "step": 1293
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6931627793235845e-05,
      "loss": 0.5047,
      "step": 1294
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6926598575170695e-05,
      "loss": 0.5084,
      "step": 1295
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6921565987159226e-05,
      "loss": 0.5248,
      "step": 1296
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.69165300316499e-05,
      "loss": 0.5173,
      "step": 1297
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6911490711092824e-05,
      "loss": 0.5361,
      "step": 1298
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6906448027939744e-05,
      "loss": 0.4966,
      "step": 1299
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6901401984644034e-05,
      "loss": 0.5204,
      "step": 1300
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6896352583660712e-05,
      "loss": 0.5283,
      "step": 1301
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6891299827446428e-05,
      "loss": 0.5093,
      "step": 1302
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6886243718459457e-05,
      "loss": 0.5409,
      "step": 1303
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6881184259159708e-05,
      "loss": 0.5153,
      "step": 1304
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.687612145200873e-05,
      "loss": 0.5153,
      "step": 1305
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6871055299469686e-05,
      "loss": 0.5058,
      "step": 1306
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.6865985804007373e-05,
      "loss": 0.5116,
      "step": 1307
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.686091296808822e-05,
      "loss": 0.5132,
      "step": 1308
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.685583679418027e-05,
      "loss": 0.5255,
      "step": 1309
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.6850757284753202e-05,
      "loss": 0.5406,
      "step": 1310
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.6845674442278307e-05,
      "loss": 0.4971,
      "step": 1311
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.6840588269228507e-05,
      "loss": 0.5071,
      "step": 1312
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.6835498768078333e-05,
      "loss": 0.5115,
      "step": 1313
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.6830405941303948e-05,
      "loss": 0.5121,
      "step": 1314
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.6825309791383128e-05,
      "loss": 0.5044,
      "step": 1315
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.682021032079526e-05,
      "loss": 0.5117,
      "step": 1316
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.6815107532021354e-05,
      "loss": 0.4931,
      "step": 1317
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.681000142754403e-05,
      "loss": 0.4823,
      "step": 1318
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.6804892009847528e-05,
      "loss": 0.5225,
      "step": 1319
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.6799779281417685e-05,
      "loss": 0.5079,
      "step": 1320
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.679466324474197e-05,
      "loss": 0.4893,
      "step": 1321
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.6789543902309443e-05,
      "loss": 0.5187,
      "step": 1322
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.6784421256610778e-05,
      "loss": 0.4944,
      "step": 1323
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.6779295310138264e-05,
      "loss": 0.5076,
      "step": 1324
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6774166065385787e-05,
      "loss": 0.5204,
      "step": 1325
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6769033524848833e-05,
      "loss": 0.52,
      "step": 1326
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6763897691024514e-05,
      "loss": 0.5268,
      "step": 1327
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6758758566411516e-05,
      "loss": 0.5077,
      "step": 1328
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6753616153510145e-05,
      "loss": 0.5178,
      "step": 1329
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6748470454822295e-05,
      "loss": 0.5056,
      "step": 1330
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.674332147285147e-05,
      "loss": 0.493,
      "step": 1331
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6738169210102765e-05,
      "loss": 0.5219,
      "step": 1332
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6733013669082868e-05,
      "loss": 0.5168,
      "step": 1333
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6727854852300073e-05,
      "loss": 0.5074,
      "step": 1334
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.672269276226425e-05,
      "loss": 0.5129,
      "step": 1335
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6717527401486882e-05,
      "loss": 0.481,
      "step": 1336
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6712358772481032e-05,
      "loss": 0.5219,
      "step": 1337
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.670718687776135e-05,
      "loss": 0.5052,
      "step": 1338
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.6702011719844083e-05,
      "loss": 0.5042,
      "step": 1339
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.669683330124706e-05,
      "loss": 0.4991,
      "step": 1340
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.66916516244897e-05,
      "loss": 0.5027,
      "step": 1341
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.6686466692093007e-05,
      "loss": 0.5023,
      "step": 1342
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.6681278506579562e-05,
      "loss": 0.5302,
      "step": 1343
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.667608707047354e-05,
      "loss": 0.5045,
      "step": 1344
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.6670892386300687e-05,
      "loss": 0.543,
      "step": 1345
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.6665694456588335e-05,
      "loss": 0.5269,
      "step": 1346
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.6660493283865394e-05,
      "loss": 0.5089,
      "step": 1347
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.6655288870662354e-05,
      "loss": 0.508,
      "step": 1348
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.665008121951128e-05,
      "loss": 0.5125,
      "step": 1349
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.6644870332945807e-05,
      "loss": 0.4898,
      "step": 1350
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.6639656213501155e-05,
      "loss": 0.5182,
      "step": 1351
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.6634438863714108e-05,
      "loss": 0.523,
      "step": 1352
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.662921828612302e-05,
      "loss": 0.5151,
      "step": 1353
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.6623994483267823e-05,
      "loss": 0.5171,
      "step": 1354
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.6618767457690012e-05,
      "loss": 0.5299,
      "step": 1355
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.661353721193266e-05,
      "loss": 0.5199,
      "step": 1356
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.6608303748540393e-05,
      "loss": 0.5336,
      "step": 1357
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.6603067070059413e-05,
      "loss": 0.5143,
      "step": 1358
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.6597827179037475e-05,
      "loss": 0.5243,
      "step": 1359
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.6592584078023915e-05,
      "loss": 0.4958,
      "step": 1360
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.658733776956961e-05,
      "loss": 0.5049,
      "step": 1361
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.658208825622701e-05,
      "loss": 0.5326,
      "step": 1362
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.6576835540550128e-05,
      "loss": 0.5201,
      "step": 1363
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.657157962509452e-05,
      "loss": 0.4997,
      "step": 1364
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.656632051241731e-05,
      "loss": 0.5272,
      "step": 1365
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.656105820507718e-05,
      "loss": 0.5172,
      "step": 1366
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.6555792705634352e-05,
      "loss": 0.4955,
      "step": 1367
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.6550524016650617e-05,
      "loss": 0.5285,
      "step": 1368
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.6545252140689307e-05,
      "loss": 0.5164,
      "step": 1369
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.6539977080315313e-05,
      "loss": 0.5184,
      "step": 1370
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.6534698838095066e-05,
      "loss": 0.5073,
      "step": 1371
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.652941741659655e-05,
      "loss": 0.5208,
      "step": 1372
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.65241328183893e-05,
      "loss": 0.4928,
      "step": 1373
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.651884504604439e-05,
      "loss": 0.5012,
      "step": 1374
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.6513554102134438e-05,
      "loss": 0.5292,
      "step": 1375
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.650825998923361e-05,
      "loss": 0.4998,
      "step": 1376
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.6502962709917612e-05,
      "loss": 0.5095,
      "step": 1377
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.6497662266763685e-05,
      "loss": 0.5255,
      "step": 1378
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.649235866235062e-05,
      "loss": 0.488,
      "step": 1379
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.6487051899258738e-05,
      "loss": 0.5265,
      "step": 1380
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.6481741980069898e-05,
      "loss": 0.5199,
      "step": 1381
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.6476428907367497e-05,
      "loss": 0.5006,
      "step": 1382
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.6471112683736462e-05,
      "loss": 0.527,
      "step": 1383
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.6465793311763255e-05,
      "loss": 0.4793,
      "step": 1384
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.6460470794035875e-05,
      "loss": 0.4945,
      "step": 1385
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.6455145133143843e-05,
      "loss": 0.4958,
      "step": 1386
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.6449816331678214e-05,
      "loss": 0.5085,
      "step": 1387
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.6444484392231574e-05,
      "loss": 0.535,
      "step": 1388
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.6439149317398025e-05,
      "loss": 0.5011,
      "step": 1389
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.6433811109773202e-05,
      "loss": 0.5181,
      "step": 1390
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.6428469771954267e-05,
      "loss": 0.5112,
      "step": 1391
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.6423125306539903e-05,
      "loss": 0.4876,
      "step": 1392
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.6417777716130305e-05,
      "loss": 0.4965,
      "step": 1393
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.64124270033272e-05,
      "loss": 0.5294,
      "step": 1394
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.6407073170733836e-05,
      "loss": 0.5222,
      "step": 1395
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.6401716220954968e-05,
      "loss": 0.5012,
      "step": 1396
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.639635615659687e-05,
      "loss": 0.5064,
      "step": 1397
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.639099298026734e-05,
      "loss": 0.5192,
      "step": 1398
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.638562669457568e-05,
      "loss": 0.4917,
      "step": 1399
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.638025730213271e-05,
      "loss": 0.5064,
      "step": 1400
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.637488480555076e-05,
      "loss": 0.5274,
      "step": 1401
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6369509207443676e-05,
      "loss": 0.5113,
      "step": 1402
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6364130510426798e-05,
      "loss": 0.5173,
      "step": 1403
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6358748717116993e-05,
      "loss": 0.5251,
      "step": 1404
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6353363830132618e-05,
      "loss": 0.5186,
      "step": 1405
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6347975852093544e-05,
      "loss": 0.5136,
      "step": 1406
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6342584785621146e-05,
      "loss": 0.5028,
      "step": 1407
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6337190633338294e-05,
      "loss": 0.5086,
      "step": 1408
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6331793397869372e-05,
      "loss": 0.5086,
      "step": 1409
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.632639308184025e-05,
      "loss": 0.507,
      "step": 1410
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6320989687878304e-05,
      "loss": 0.4973,
      "step": 1411
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.631558321861241e-05,
      "loss": 0.5337,
      "step": 1412
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.631017367667294e-05,
      "loss": 0.5183,
      "step": 1413
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6304761064691752e-05,
      "loss": 0.5192,
      "step": 1414
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6299345385302204e-05,
      "loss": 0.4876,
      "step": 1415
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.6293926641139154e-05,
      "loss": 0.5069,
      "step": 1416
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.628850483483893e-05,
      "loss": 0.5058,
      "step": 1417
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6283079969039376e-05,
      "loss": 0.5173,
      "step": 1418
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6277652046379804e-05,
      "loss": 0.5151,
      "step": 1419
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.627222106950102e-05,
      "loss": 0.4989,
      "step": 1420
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6266787041045322e-05,
      "loss": 0.523,
      "step": 1421
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6261349963656482e-05,
      "loss": 0.5282,
      "step": 1422
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6255909839979764e-05,
      "loss": 0.5199,
      "step": 1423
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.625046667266191e-05,
      "loss": 0.5036,
      "step": 1424
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6245020464351142e-05,
      "loss": 0.4837,
      "step": 1425
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6239571217697164e-05,
      "loss": 0.5184,
      "step": 1426
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6234118935351162e-05,
      "loss": 0.5194,
      "step": 1427
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6228663619965787e-05,
      "loss": 0.508,
      "step": 1428
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6223205274195178e-05,
      "loss": 0.4745,
      "step": 1429
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.621774390069494e-05,
      "loss": 0.5002,
      "step": 1430
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6212279502122158e-05,
      "loss": 0.5043,
      "step": 1431
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.620681208113538e-05,
      "loss": 0.5216,
      "step": 1432
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.620134164039463e-05,
      "loss": 0.5129,
      "step": 1433
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.619586818256141e-05,
      "loss": 0.5296,
      "step": 1434
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.6190391710298664e-05,
      "loss": 0.5182,
      "step": 1435
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.6184912226270833e-05,
      "loss": 0.5254,
      "step": 1436
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.6179429733143807e-05,
      "loss": 0.5026,
      "step": 1437
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.6173944233584936e-05,
      "loss": 0.4971,
      "step": 1438
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.6168455730263048e-05,
      "loss": 0.5064,
      "step": 1439
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.6162964225848416e-05,
      "loss": 0.49,
      "step": 1440
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.615746972301279e-05,
      "loss": 0.5023,
      "step": 1441
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.6151972224429356e-05,
      "loss": 0.5344,
      "step": 1442
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.6146471732772783e-05,
      "loss": 0.5106,
      "step": 1443
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.6140968250719177e-05,
      "loss": 0.5374,
      "step": 1444
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.613546178094611e-05,
      "loss": 0.513,
      "step": 1445
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.6129952326132603e-05,
      "loss": 0.5195,
      "step": 1446
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.6124439888959128e-05,
      "loss": 0.5115,
      "step": 1447
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.611892447210761e-05,
      "loss": 0.4948,
      "step": 1448
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6113406078261427e-05,
      "loss": 0.5326,
      "step": 1449
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.61078847101054e-05,
      "loss": 0.5116,
      "step": 1450
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6102360370325797e-05,
      "loss": 0.4932,
      "step": 1451
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.609683306161034e-05,
      "loss": 0.5219,
      "step": 1452
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.609130278664818e-05,
      "loss": 0.5207,
      "step": 1453
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6085769548129928e-05,
      "loss": 0.4943,
      "step": 1454
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.608023334874762e-05,
      "loss": 0.5191,
      "step": 1455
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6074694191194758e-05,
      "loss": 0.5183,
      "step": 1456
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.606915207816625e-05,
      "loss": 0.5353,
      "step": 1457
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6063607012358474e-05,
      "loss": 0.5192,
      "step": 1458
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6058058996469213e-05,
      "loss": 0.4887,
      "step": 1459
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6052508033197713e-05,
      "loss": 0.5243,
      "step": 1460
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.604695412524464e-05,
      "loss": 0.4927,
      "step": 1461
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6041397275312102e-05,
      "loss": 0.5017,
      "step": 1462
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.6035837486103617e-05,
      "loss": 0.5185,
      "step": 1463
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.6030274760324163e-05,
      "loss": 0.5145,
      "step": 1464
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.602470910068012e-05,
      "loss": 0.4996,
      "step": 1465
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.6019140509879312e-05,
      "loss": 0.5129,
      "step": 1466
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.6013568990630986e-05,
      "loss": 0.493,
      "step": 1467
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.6007994545645807e-05,
      "loss": 0.494,
      "step": 1468
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.6002417177635868e-05,
      "loss": 0.5111,
      "step": 1469
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.599683688931469e-05,
      "loss": 0.5222,
      "step": 1470
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5991253683397205e-05,
      "loss": 0.5072,
      "step": 1471
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.598566756259977e-05,
      "loss": 0.4955,
      "step": 1472
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.598007852964015e-05,
      "loss": 0.5214,
      "step": 1473
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5974486587237554e-05,
      "loss": 0.5053,
      "step": 1474
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.596889173811257e-05,
      "loss": 0.501,
      "step": 1475
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.596329398498723e-05,
      "loss": 0.5176,
      "step": 1476
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5957693330584957e-05,
      "loss": 0.5073,
      "step": 1477
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5952089777630604e-05,
      "loss": 0.5094,
      "step": 1478
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.5946483328850426e-05,
      "loss": 0.5098,
      "step": 1479
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.5940873986972078e-05,
      "loss": 0.5364,
      "step": 1480
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.593526175472464e-05,
      "loss": 0.4954,
      "step": 1481
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.5929646634838583e-05,
      "loss": 0.4943,
      "step": 1482
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.59240286300458e-05,
      "loss": 0.5158,
      "step": 1483
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.5918407743079564e-05,
      "loss": 0.5005,
      "step": 1484
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.591278397667457e-05,
      "loss": 0.5145,
      "step": 1485
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.59071573335669e-05,
      "loss": 0.5202,
      "step": 1486
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.5901527816494056e-05,
      "loss": 0.4848,
      "step": 1487
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.5895895428194915e-05,
      "loss": 0.4877,
      "step": 1488
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.5890260171409765e-05,
      "loss": 0.5099,
      "step": 1489
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.5884622048880283e-05,
      "loss": 0.5028,
      "step": 1490
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.5878981063349543e-05,
      "loss": 0.5397,
      "step": 1491
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.5873337217562012e-05,
      "loss": 0.5019,
      "step": 1492
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.5867690514263548e-05,
      "loss": 0.5302,
      "step": 1493
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.58620409562014e-05,
      "loss": 0.4937,
      "step": 1494
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.5856388546124208e-05,
      "loss": 0.5006,
      "step": 1495
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.585073328678199e-05,
      "loss": 0.5192,
      "step": 1496
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.5845075180926163e-05,
      "loss": 0.4996,
      "step": 1497
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.583941423130952e-05,
      "loss": 0.5166,
      "step": 1498
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.5833750440686242e-05,
      "loss": 0.487,
      "step": 1499
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.582808381181189e-05,
      "loss": 0.4884,
      "step": 1500
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.582241434744341e-05,
      "loss": 0.5056,
      "step": 1501
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.581674205033912e-05,
      "loss": 0.5273,
      "step": 1502
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.5811066923258718e-05,
      "loss": 0.5266,
      "step": 1503
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.5805388968963286e-05,
      "loss": 0.5133,
      "step": 1504
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.579970819021528e-05,
      "loss": 0.5338,
      "step": 1505
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.5794024589778518e-05,
      "loss": 0.5346,
      "step": 1506
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.5788338170418208e-05,
      "loss": 0.5348,
      "step": 1507
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.5782648934900915e-05,
      "loss": 0.489,
      "step": 1508
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.5776956885994584e-05,
      "loss": 0.5334,
      "step": 1509
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.577126202646852e-05,
      "loss": 0.499,
      "step": 1510
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.576556435909341e-05,
      "loss": 0.4841,
      "step": 1511
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.575986388664129e-05,
      "loss": 0.5462,
      "step": 1512
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5754160611885566e-05,
      "loss": 0.479,
      "step": 1513
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.574845453760102e-05,
      "loss": 0.5198,
      "step": 1514
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.574274566656378e-05,
      "loss": 0.5182,
      "step": 1515
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5737034001551336e-05,
      "loss": 0.5035,
      "step": 1516
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.573131954534255e-05,
      "loss": 0.5155,
      "step": 1517
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5725602300717628e-05,
      "loss": 0.5047,
      "step": 1518
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5719882270458138e-05,
      "loss": 0.5021,
      "step": 1519
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5714159457347007e-05,
      "loss": 0.5158,
      "step": 1520
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5708433864168512e-05,
      "loss": 0.5096,
      "step": 1521
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5702705493708283e-05,
      "loss": 0.5298,
      "step": 1522
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5696974348753298e-05,
      "loss": 0.5062,
      "step": 1523
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5691240432091892e-05,
      "loss": 0.4926,
      "step": 1524
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.5685503746513744e-05,
      "loss": 0.5314,
      "step": 1525
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5679764294809882e-05,
      "loss": 0.5064,
      "step": 1526
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5674022079772675e-05,
      "loss": 0.5277,
      "step": 1527
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.566827710419584e-05,
      "loss": 0.507,
      "step": 1528
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5662529370874444e-05,
      "loss": 0.48,
      "step": 1529
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.565677888260488e-05,
      "loss": 0.5373,
      "step": 1530
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5651025642184898e-05,
      "loss": 0.5091,
      "step": 1531
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5645269652413574e-05,
      "loss": 0.5269,
      "step": 1532
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5639510916091326e-05,
      "loss": 0.5,
      "step": 1533
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5633749436019913e-05,
      "loss": 0.5446,
      "step": 1534
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5627985215002423e-05,
      "loss": 0.4965,
      "step": 1535
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5622218255843276e-05,
      "loss": 0.5218,
      "step": 1536
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5616448561348233e-05,
      "loss": 0.5086,
      "step": 1537
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.561067613432438e-05,
      "loss": 0.507,
      "step": 1538
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5604900977580132e-05,
      "loss": 0.4896,
      "step": 1539
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.559912309392523e-05,
      "loss": 0.5189,
      "step": 1540
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.559334248617075e-05,
      "loss": 0.5107,
      "step": 1541
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5587559157129078e-05,
      "loss": 0.5429,
      "step": 1542
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5581773109613947e-05,
      "loss": 0.5176,
      "step": 1543
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5575984346440393e-05,
      "loss": 0.505,
      "step": 1544
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5570192870424775e-05,
      "loss": 0.5305,
      "step": 1545
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5564398684384787e-05,
      "loss": 0.5323,
      "step": 1546
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.555860179113942e-05,
      "loss": 0.5274,
      "step": 1547
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5552802193509003e-05,
      "loss": 0.504,
      "step": 1548
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5546999894315166e-05,
      "loss": 0.5059,
      "step": 1549
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5541194896380863e-05,
      "loss": 0.4478,
      "step": 1550
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5535387202530346e-05,
      "loss": 0.4446,
      "step": 1551
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.55295768155892e-05,
      "loss": 0.45,
      "step": 1552
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5523763738384303e-05,
      "loss": 0.4117,
      "step": 1553
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.551794797374385e-05,
      "loss": 0.4297,
      "step": 1554
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.551212952449734e-05,
      "loss": 0.4501,
      "step": 1555
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.5506308393475582e-05,
      "loss": 0.41,
      "step": 1556
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5500484583510685e-05,
      "loss": 0.4113,
      "step": 1557
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.549465809743607e-05,
      "loss": 0.4185,
      "step": 1558
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5488828938086444e-05,
      "loss": 0.4531,
      "step": 1559
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5482997108297834e-05,
      "loss": 0.4381,
      "step": 1560
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.547716261090755e-05,
      "loss": 0.4539,
      "step": 1561
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5471325448754207e-05,
      "loss": 0.4189,
      "step": 1562
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.546548562467772e-05,
      "loss": 0.4226,
      "step": 1563
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.545964314151929e-05,
      "loss": 0.4294,
      "step": 1564
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.545379800212142e-05,
      "loss": 0.4324,
      "step": 1565
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5447950209327905e-05,
      "loss": 0.4274,
      "step": 1566
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.544209976598382e-05,
      "loss": 0.4211,
      "step": 1567
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5436246674935543e-05,
      "loss": 0.4201,
      "step": 1568
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5430390939030734e-05,
      "loss": 0.4164,
      "step": 1569
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.542453256111834e-05,
      "loss": 0.3999,
      "step": 1570
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5418671544048597e-05,
      "loss": 0.4034,
      "step": 1571
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.5412807890673015e-05,
      "loss": 0.4327,
      "step": 1572
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.54069416038444e-05,
      "loss": 0.4145,
      "step": 1573
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.5401072686416826e-05,
      "loss": 0.4092,
      "step": 1574
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.5395201141245662e-05,
      "loss": 0.4275,
      "step": 1575
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.5389326971187543e-05,
      "loss": 0.4166,
      "step": 1576
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.538345017910038e-05,
      "loss": 0.415,
      "step": 1577
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.5377570767843377e-05,
      "loss": 0.4251,
      "step": 1578
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.5371688740276988e-05,
      "loss": 0.4235,
      "step": 1579
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.536580409926296e-05,
      "loss": 0.4399,
      "step": 1580
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.53599168476643e-05,
      "loss": 0.4129,
      "step": 1581
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.5354026988345284e-05,
      "loss": 0.4214,
      "step": 1582
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.534813452417147e-05,
      "loss": 0.4435,
      "step": 1583
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.5342239458009675e-05,
      "loss": 0.4283,
      "step": 1584
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.533634179272797e-05,
      "loss": 0.4392,
      "step": 1585
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.5330441531195714e-05,
      "loss": 0.4045,
      "step": 1586
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.5324538676283512e-05,
      "loss": 0.4215,
      "step": 1587
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5318633230863237e-05,
      "loss": 0.4385,
      "step": 1588
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.531272519780802e-05,
      "loss": 0.4292,
      "step": 1589
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5306814579992254e-05,
      "loss": 0.4234,
      "step": 1590
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5300901380291583e-05,
      "loss": 0.4342,
      "step": 1591
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5294985601582922e-05,
      "loss": 0.4395,
      "step": 1592
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5289067246744423e-05,
      "loss": 0.4022,
      "step": 1593
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.52831463186555e-05,
      "loss": 0.4138,
      "step": 1594
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5277222820196816e-05,
      "loss": 0.4228,
      "step": 1595
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5271296754250296e-05,
      "loss": 0.4325,
      "step": 1596
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5265368123699093e-05,
      "loss": 0.4096,
      "step": 1597
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5259436931427624e-05,
      "loss": 0.4184,
      "step": 1598
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.525350318032155e-05,
      "loss": 0.4267,
      "step": 1599
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.524756687326777e-05,
      "loss": 0.4248,
      "step": 1600
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5241628013154435e-05,
      "loss": 0.4172,
      "step": 1601
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5235686602870932e-05,
      "loss": 0.4086,
      "step": 1602
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.5229742645307894e-05,
      "loss": 0.4395,
      "step": 1603
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5223796143357188e-05,
      "loss": 0.4212,
      "step": 1604
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5217847099911919e-05,
      "loss": 0.4489,
      "step": 1605
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5211895517866437e-05,
      "loss": 0.4199,
      "step": 1606
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5205941400116316e-05,
      "loss": 0.4257,
      "step": 1607
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5199984749558367e-05,
      "loss": 0.4331,
      "step": 1608
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5194025569090638e-05,
      "loss": 0.4525,
      "step": 1609
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5188063861612405e-05,
      "loss": 0.4509,
      "step": 1610
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5182099630024172e-05,
      "loss": 0.4287,
      "step": 1611
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5176132877227674e-05,
      "loss": 0.4517,
      "step": 1612
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5170163606125867e-05,
      "loss": 0.4353,
      "step": 1613
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5164191819622937e-05,
      "loss": 0.4318,
      "step": 1614
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5158217520624296e-05,
      "loss": 0.422,
      "step": 1615
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5152240712036573e-05,
      "loss": 0.4314,
      "step": 1616
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.5146261396767615e-05,
      "loss": 0.4422,
      "step": 1617
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.51402795777265e-05,
      "loss": 0.3957,
      "step": 1618
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5134295257823519e-05,
      "loss": 0.4189,
      "step": 1619
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5128308439970174e-05,
      "loss": 0.4223,
      "step": 1620
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5122319127079185e-05,
      "loss": 0.4255,
      "step": 1621
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5116327322064497e-05,
      "loss": 0.4187,
      "step": 1622
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5110333027841248e-05,
      "loss": 0.4265,
      "step": 1623
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5104336247325803e-05,
      "loss": 0.4195,
      "step": 1624
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5098336983435729e-05,
      "loss": 0.4096,
      "step": 1625
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5092335239089803e-05,
      "loss": 0.4335,
      "step": 1626
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5086331017208011e-05,
      "loss": 0.4308,
      "step": 1627
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5080324320711542e-05,
      "loss": 0.4553,
      "step": 1628
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5074315152522787e-05,
      "loss": 0.4162,
      "step": 1629
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.506830351556534e-05,
      "loss": 0.43,
      "step": 1630
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5062289412764003e-05,
      "loss": 0.4262,
      "step": 1631
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.505627284704477e-05,
      "loss": 0.4305,
      "step": 1632
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.5050253821334837e-05,
      "loss": 0.4365,
      "step": 1633
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.504423233856259e-05,
      "loss": 0.4106,
      "step": 1634
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.5038208401657626e-05,
      "loss": 0.4323,
      "step": 1635
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.5032182013550719e-05,
      "loss": 0.4318,
      "step": 1636
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.5026153177173841e-05,
      "loss": 0.4331,
      "step": 1637
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.5020121895460165e-05,
      "loss": 0.4387,
      "step": 1638
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.501408817134404e-05,
      "loss": 0.4137,
      "step": 1639
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.5008052007761009e-05,
      "loss": 0.4333,
      "step": 1640
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.5002013407647801e-05,
      "loss": 0.4017,
      "step": 1641
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.4995972373942334e-05,
      "loss": 0.4188,
      "step": 1642
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.4989928909583701e-05,
      "loss": 0.4387,
      "step": 1643
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.498388301751219e-05,
      "loss": 0.435,
      "step": 1644
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.4977834700669265e-05,
      "loss": 0.4283,
      "step": 1645
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.4971783961997561e-05,
      "loss": 0.4484,
      "step": 1646
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.4965730804440905e-05,
      "loss": 0.4497,
      "step": 1647
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.495967523094429e-05,
      "loss": 0.4303,
      "step": 1648
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.4953617244453895e-05,
      "loss": 0.4389,
      "step": 1649
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4947556847917062e-05,
      "loss": 0.4151,
      "step": 1650
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4941494044282313e-05,
      "loss": 0.4287,
      "step": 1651
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4935428836499333e-05,
      "loss": 0.4211,
      "step": 1652
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4929361227518989e-05,
      "loss": 0.4279,
      "step": 1653
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4923291220293307e-05,
      "loss": 0.4511,
      "step": 1654
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4917218817775485e-05,
      "loss": 0.4054,
      "step": 1655
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4911144022919879e-05,
      "loss": 0.4259,
      "step": 1656
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4905066838682017e-05,
      "loss": 0.4055,
      "step": 1657
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4898987268018586e-05,
      "loss": 0.3893,
      "step": 1658
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4892905313887432e-05,
      "loss": 0.4126,
      "step": 1659
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4886820979247561e-05,
      "loss": 0.4176,
      "step": 1660
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4880734267059142e-05,
      "loss": 0.4571,
      "step": 1661
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.48746451802835e-05,
      "loss": 0.4408,
      "step": 1662
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4868553721883113e-05,
      "loss": 0.4488,
      "step": 1663
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.4862459894821606e-05,
      "loss": 0.4455,
      "step": 1664
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.485636370206377e-05,
      "loss": 0.4229,
      "step": 1665
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4850265146575535e-05,
      "loss": 0.4197,
      "step": 1666
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4844164231323992e-05,
      "loss": 0.4417,
      "step": 1667
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.483806095927737e-05,
      "loss": 0.4093,
      "step": 1668
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4831955333405044e-05,
      "loss": 0.4296,
      "step": 1669
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4825847356677546e-05,
      "loss": 0.4231,
      "step": 1670
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.481973703206654e-05,
      "loss": 0.43,
      "step": 1671
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.481362436254484e-05,
      "loss": 0.4271,
      "step": 1672
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4807509351086395e-05,
      "loss": 0.4426,
      "step": 1673
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4801392000666297e-05,
      "loss": 0.4248,
      "step": 1674
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4795272314260776e-05,
      "loss": 0.421,
      "step": 1675
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4789150294847192e-05,
      "loss": 0.4428,
      "step": 1676
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4783025945404053e-05,
      "loss": 0.4245,
      "step": 1677
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4776899268910985e-05,
      "loss": 0.4227,
      "step": 1678
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.477077026834876e-05,
      "loss": 0.4246,
      "step": 1679
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.4764638946699275e-05,
      "loss": 0.4097,
      "step": 1680
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4758505306945551e-05,
      "loss": 0.4201,
      "step": 1681
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.475236935207175e-05,
      "loss": 0.4363,
      "step": 1682
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4746231085063146e-05,
      "loss": 0.4354,
      "step": 1683
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4740090508906147e-05,
      "loss": 0.4017,
      "step": 1684
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4733947626588279e-05,
      "loss": 0.4261,
      "step": 1685
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4727802441098193e-05,
      "loss": 0.4361,
      "step": 1686
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4721654955425662e-05,
      "loss": 0.4315,
      "step": 1687
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4715505172561577e-05,
      "loss": 0.4531,
      "step": 1688
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4709353095497947e-05,
      "loss": 0.4449,
      "step": 1689
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4703198727227892e-05,
      "loss": 0.4278,
      "step": 1690
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4697042070745654e-05,
      "loss": 0.4389,
      "step": 1691
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4690883129046585e-05,
      "loss": 0.4188,
      "step": 1692
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4684721905127147e-05,
      "loss": 0.4251,
      "step": 1693
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4678558401984915e-05,
      "loss": 0.4103,
      "step": 1694
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.4672392622618571e-05,
      "loss": 0.4007,
      "step": 1695
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.466622457002791e-05,
      "loss": 0.3978,
      "step": 1696
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4660054247213826e-05,
      "loss": 0.4144,
      "step": 1697
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4653881657178317e-05,
      "loss": 0.428,
      "step": 1698
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4647706802924493e-05,
      "loss": 0.4441,
      "step": 1699
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4641529687456558e-05,
      "loss": 0.3971,
      "step": 1700
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4635350313779815e-05,
      "loss": 0.4287,
      "step": 1701
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.462916868490067e-05,
      "loss": 0.4084,
      "step": 1702
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4622984803826625e-05,
      "loss": 0.4152,
      "step": 1703
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4616798673566276e-05,
      "loss": 0.4461,
      "step": 1704
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.461061029712932e-05,
      "loss": 0.431,
      "step": 1705
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4604419677526536e-05,
      "loss": 0.4169,
      "step": 1706
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4598226817769803e-05,
      "loss": 0.4247,
      "step": 1707
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4592031720872086e-05,
      "loss": 0.4144,
      "step": 1708
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4585834389847439e-05,
      "loss": 0.4397,
      "step": 1709
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4579634827711004e-05,
      "loss": 0.4234,
      "step": 1710
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.4573433037479006e-05,
      "loss": 0.4144,
      "step": 1711
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4567229022168756e-05,
      "loss": 0.4489,
      "step": 1712
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4561022784798648e-05,
      "loss": 0.4332,
      "step": 1713
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4554814328388158e-05,
      "loss": 0.4347,
      "step": 1714
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4548603655957839e-05,
      "loss": 0.4232,
      "step": 1715
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.454239077052932e-05,
      "loss": 0.4221,
      "step": 1716
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4536175675125315e-05,
      "loss": 0.412,
      "step": 1717
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4529958372769603e-05,
      "loss": 0.4394,
      "step": 1718
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4523738866487048e-05,
      "loss": 0.4208,
      "step": 1719
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4517517159303573e-05,
      "loss": 0.4212,
      "step": 1720
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4511293254246179e-05,
      "loss": 0.4202,
      "step": 1721
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4505067154342944e-05,
      "loss": 0.4165,
      "step": 1722
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4498838862622997e-05,
      "loss": 0.4459,
      "step": 1723
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4492608382116548e-05,
      "loss": 0.436,
      "step": 1724
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4486375715854866e-05,
      "loss": 0.4473,
      "step": 1725
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.4480140866870281e-05,
      "loss": 0.4286,
      "step": 1726
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4473903838196188e-05,
      "loss": 0.4274,
      "step": 1727
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4467664632867042e-05,
      "loss": 0.4466,
      "step": 1728
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4461423253918358e-05,
      "loss": 0.4257,
      "step": 1729
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4455179704386706e-05,
      "loss": 0.4555,
      "step": 1730
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4448933987309715e-05,
      "loss": 0.4351,
      "step": 1731
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4442686105726066e-05,
      "loss": 0.4299,
      "step": 1732
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4436436062675498e-05,
      "loss": 0.4439,
      "step": 1733
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4430183861198792e-05,
      "loss": 0.4385,
      "step": 1734
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.442392950433779e-05,
      "loss": 0.4536,
      "step": 1735
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4417672995135372e-05,
      "loss": 0.4551,
      "step": 1736
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4411414336635476e-05,
      "loss": 0.4175,
      "step": 1737
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.440515353188308e-05,
      "loss": 0.4151,
      "step": 1738
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4398890583924202e-05,
      "loss": 0.4208,
      "step": 1739
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4392625495805913e-05,
      "loss": 0.4148,
      "step": 1740
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4386358270576315e-05,
      "loss": 0.421,
      "step": 1741
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.4380088911284557e-05,
      "loss": 0.4215,
      "step": 1742
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4373817420980824e-05,
      "loss": 0.4091,
      "step": 1743
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4367543802716334e-05,
      "loss": 0.4236,
      "step": 1744
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4361268059543347e-05,
      "loss": 0.4137,
      "step": 1745
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4354990194515155e-05,
      "loss": 0.4419,
      "step": 1746
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4348710210686074e-05,
      "loss": 0.4236,
      "step": 1747
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4342428111111461e-05,
      "loss": 0.4119,
      "step": 1748
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4336143898847705e-05,
      "loss": 0.4058,
      "step": 1749
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4329857576952212e-05,
      "loss": 0.4165,
      "step": 1750
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4323569148483417e-05,
      "loss": 0.4505,
      "step": 1751
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4317278616500785e-05,
      "loss": 0.4335,
      "step": 1752
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.43109859840648e-05,
      "loss": 0.4254,
      "step": 1753
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.430469125423697e-05,
      "loss": 0.4323,
      "step": 1754
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4298394430079826e-05,
      "loss": 0.4393,
      "step": 1755
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4292095514656907e-05,
      "loss": 0.4096,
      "step": 1756
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.4285794511032788e-05,
      "loss": 0.4137,
      "step": 1757
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4279491422273043e-05,
      "loss": 0.43,
      "step": 1758
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4273186251444267e-05,
      "loss": 0.4446,
      "step": 1759
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4266879001614067e-05,
      "loss": 0.4163,
      "step": 1760
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4260569675851064e-05,
      "loss": 0.4288,
      "step": 1761
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4254258277224888e-05,
      "loss": 0.4148,
      "step": 1762
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4247944808806176e-05,
      "loss": 0.4248,
      "step": 1763
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.424162927366657e-05,
      "loss": 0.428,
      "step": 1764
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4235311674878723e-05,
      "loss": 0.432,
      "step": 1765
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4228992015516287e-05,
      "loss": 0.422,
      "step": 1766
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4222670298653922e-05,
      "loss": 0.4355,
      "step": 1767
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4216346527367284e-05,
      "loss": 0.4433,
      "step": 1768
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4210020704733033e-05,
      "loss": 0.416,
      "step": 1769
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4203692833828817e-05,
      "loss": 0.4416,
      "step": 1770
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4197362917733295e-05,
      "loss": 0.4247,
      "step": 1771
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4191030959526106e-05,
      "loss": 0.4045,
      "step": 1772
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.4184696962287901e-05,
      "loss": 0.4213,
      "step": 1773
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4178360929100303e-05,
      "loss": 0.4494,
      "step": 1774
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4172022863045942e-05,
      "loss": 0.45,
      "step": 1775
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4165682767208426e-05,
      "loss": 0.4267,
      "step": 1776
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4159340644672358e-05,
      "loss": 0.4263,
      "step": 1777
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4152996498523317e-05,
      "loss": 0.456,
      "step": 1778
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4146650331847879e-05,
      "loss": 0.4182,
      "step": 1779
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4140302147733596e-05,
      "loss": 0.4237,
      "step": 1780
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4133951949269e-05,
      "loss": 0.4633,
      "step": 1781
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4127599739543606e-05,
      "loss": 0.4266,
      "step": 1782
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4121245521647913e-05,
      "loss": 0.444,
      "step": 1783
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4114889298673383e-05,
      "loss": 0.4167,
      "step": 1784
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4108531073712468e-05,
      "loss": 0.4189,
      "step": 1785
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4102170849858583e-05,
      "loss": 0.4451,
      "step": 1786
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4095808630206125e-05,
      "loss": 0.4512,
      "step": 1787
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.4089444417850455e-05,
      "loss": 0.4174,
      "step": 1788
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4083078215887904e-05,
      "loss": 0.4143,
      "step": 1789
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4076710027415776e-05,
      "loss": 0.4251,
      "step": 1790
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4070339855532333e-05,
      "loss": 0.4487,
      "step": 1791
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4063967703336814e-05,
      "loss": 0.436,
      "step": 1792
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4057593573929413e-05,
      "loss": 0.4334,
      "step": 1793
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4051217470411284e-05,
      "loss": 0.434,
      "step": 1794
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4044839395884546e-05,
      "loss": 0.4387,
      "step": 1795
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.403845935345228e-05,
      "loss": 0.441,
      "step": 1796
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4032077346218518e-05,
      "loss": 0.4367,
      "step": 1797
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4025693377288246e-05,
      "loss": 0.4298,
      "step": 1798
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4019307449767414e-05,
      "loss": 0.424,
      "step": 1799
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.401291956676292e-05,
      "loss": 0.4249,
      "step": 1800
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4006529731382609e-05,
      "loss": 0.4078,
      "step": 1801
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.4000137946735284e-05,
      "loss": 0.4224,
      "step": 1802
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.3993744215930694e-05,
      "loss": 0.4473,
      "step": 1803
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.3987348542079526e-05,
      "loss": 0.4017,
      "step": 1804
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.3980950928293424e-05,
      "loss": 0.4287,
      "step": 1805
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.397455137768497e-05,
      "loss": 0.4507,
      "step": 1806
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.396814989336769e-05,
      "loss": 0.4349,
      "step": 1807
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.396174647845605e-05,
      "loss": 0.4451,
      "step": 1808
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.395534113606546e-05,
      "loss": 0.4325,
      "step": 1809
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.3948933869312258e-05,
      "loss": 0.4233,
      "step": 1810
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.3942524681313722e-05,
      "loss": 0.4138,
      "step": 1811
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.3936113575188074e-05,
      "loss": 0.4304,
      "step": 1812
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.3929700554054457e-05,
      "loss": 0.4239,
      "step": 1813
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.392328562103295e-05,
      "loss": 0.4251,
      "step": 1814
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.3916868779244563e-05,
      "loss": 0.4382,
      "step": 1815
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.3910450031811235e-05,
      "loss": 0.4424,
      "step": 1816
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.3904029381855831e-05,
      "loss": 0.4181,
      "step": 1817
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.389760683250214e-05,
      "loss": 0.3904,
      "step": 1818
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.389118238687488e-05,
      "loss": 0.4131,
      "step": 1819
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3884756048099688e-05,
      "loss": 0.4324,
      "step": 1820
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3878327819303121e-05,
      "loss": 0.4285,
      "step": 1821
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3871897703612658e-05,
      "loss": 0.4259,
      "step": 1822
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3865465704156696e-05,
      "loss": 0.4444,
      "step": 1823
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3859031824064543e-05,
      "loss": 0.4103,
      "step": 1824
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3852596066466437e-05,
      "loss": 0.433,
      "step": 1825
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3846158434493507e-05,
      "loss": 0.4573,
      "step": 1826
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3839718931277814e-05,
      "loss": 0.4033,
      "step": 1827
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3833277559952323e-05,
      "loss": 0.4026,
      "step": 1828
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3826834323650899e-05,
      "loss": 0.4317,
      "step": 1829
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3820389225508327e-05,
      "loss": 0.4102,
      "step": 1830
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.381394226866029e-05,
      "loss": 0.4184,
      "step": 1831
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.380749345624338e-05,
      "loss": 0.4288,
      "step": 1832
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3801042791395084e-05,
      "loss": 0.4268,
      "step": 1833
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3794590277253803e-05,
      "loss": 0.4421,
      "step": 1834
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.3788135916958827e-05,
      "loss": 0.4318,
      "step": 1835
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3781679713650349e-05,
      "loss": 0.4157,
      "step": 1836
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3775221670469457e-05,
      "loss": 0.4109,
      "step": 1837
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3768761790558134e-05,
      "loss": 0.4314,
      "step": 1838
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3762300077059258e-05,
      "loss": 0.4345,
      "step": 1839
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3755836533116597e-05,
      "loss": 0.4199,
      "step": 1840
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3749371161874813e-05,
      "loss": 0.4242,
      "step": 1841
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.374290396647945e-05,
      "loss": 0.445,
      "step": 1842
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.373643495007695e-05,
      "loss": 0.4404,
      "step": 1843
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3729964115814636e-05,
      "loss": 0.4236,
      "step": 1844
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.372349146684071e-05,
      "loss": 0.4211,
      "step": 1845
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.371701700630426e-05,
      "loss": 0.4167,
      "step": 1846
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3710540737355265e-05,
      "loss": 0.4061,
      "step": 1847
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3704062663144569e-05,
      "loss": 0.4296,
      "step": 1848
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3697582786823902e-05,
      "loss": 0.4353,
      "step": 1849
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.3691101111545873e-05,
      "loss": 0.4424,
      "step": 1850
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.368461764046396e-05,
      "loss": 0.4295,
      "step": 1851
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3678132376732518e-05,
      "loss": 0.4576,
      "step": 1852
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3671645323506776e-05,
      "loss": 0.417,
      "step": 1853
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3665156483942834e-05,
      "loss": 0.4234,
      "step": 1854
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3658665861197653e-05,
      "loss": 0.4353,
      "step": 1855
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3652173458429068e-05,
      "loss": 0.4433,
      "step": 1856
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3645679278795783e-05,
      "loss": 0.4451,
      "step": 1857
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.363918332545736e-05,
      "loss": 0.4224,
      "step": 1858
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3632685601574225e-05,
      "loss": 0.4191,
      "step": 1859
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3626186110307673e-05,
      "loss": 0.4235,
      "step": 1860
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3619684854819845e-05,
      "loss": 0.418,
      "step": 1861
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3613181838273758e-05,
      "loss": 0.439,
      "step": 1862
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3606677063833269e-05,
      "loss": 0.4185,
      "step": 1863
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3600170534663097e-05,
      "loss": 0.4339,
      "step": 1864
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.3593662253928822e-05,
      "loss": 0.4443,
      "step": 1865
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.358715222479686e-05,
      "loss": 0.4144,
      "step": 1866
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3580640450434494e-05,
      "loss": 0.4302,
      "step": 1867
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3574126934009843e-05,
      "loss": 0.4452,
      "step": 1868
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3567611678691888e-05,
      "loss": 0.4109,
      "step": 1869
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.356109468765044e-05,
      "loss": 0.4378,
      "step": 1870
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3554575964056165e-05,
      "loss": 0.4331,
      "step": 1871
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3548055511080568e-05,
      "loss": 0.4448,
      "step": 1872
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3541533331895997e-05,
      "loss": 0.4383,
      "step": 1873
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3535009429675641e-05,
      "loss": 0.4276,
      "step": 1874
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3528483807593522e-05,
      "loss": 0.4409,
      "step": 1875
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3521956468824505e-05,
      "loss": 0.4052,
      "step": 1876
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3515427416544283e-05,
      "loss": 0.4433,
      "step": 1877
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3508896653929392e-05,
      "loss": 0.4254,
      "step": 1878
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3502364184157195e-05,
      "loss": 0.4589,
      "step": 1879
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3495830010405884e-05,
      "loss": 0.4225,
      "step": 1880
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.3489294135854481e-05,
      "loss": 0.423,
      "step": 1881
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3482756563682837e-05,
      "loss": 0.4398,
      "step": 1882
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.347621729707163e-05,
      "loss": 0.4085,
      "step": 1883
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.346967633920236e-05,
      "loss": 0.4178,
      "step": 1884
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3463133693257345e-05,
      "loss": 0.4221,
      "step": 1885
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3456589362419739e-05,
      "loss": 0.4123,
      "step": 1886
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3450043349873499e-05,
      "loss": 0.4108,
      "step": 1887
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.344349565880341e-05,
      "loss": 0.4285,
      "step": 1888
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3436946292395075e-05,
      "loss": 0.4486,
      "step": 1889
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3430395253834902e-05,
      "loss": 0.4061,
      "step": 1890
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3423842546310123e-05,
      "loss": 0.438,
      "step": 1891
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3417288173008778e-05,
      "loss": 0.4321,
      "step": 1892
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3410732137119716e-05,
      "loss": 0.4401,
      "step": 1893
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3404174441832592e-05,
      "loss": 0.4331,
      "step": 1894
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.3397615090337881e-05,
      "loss": 0.4364,
      "step": 1895
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.339105408582685e-05,
      "loss": 0.4292,
      "step": 1896
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.338449143149158e-05,
      "loss": 0.4089,
      "step": 1897
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3377927130524943e-05,
      "loss": 0.4108,
      "step": 1898
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3371361186120627e-05,
      "loss": 0.4358,
      "step": 1899
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3364793601473105e-05,
      "loss": 0.4161,
      "step": 1900
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.335822437977766e-05,
      "loss": 0.4309,
      "step": 1901
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3351653524230366e-05,
      "loss": 0.436,
      "step": 1902
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3345081038028086e-05,
      "loss": 0.4152,
      "step": 1903
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3338506924368494e-05,
      "loss": 0.4258,
      "step": 1904
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3331931186450038e-05,
      "loss": 0.442,
      "step": 1905
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.332535382747196e-05,
      "loss": 0.4513,
      "step": 1906
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3318774850634297e-05,
      "loss": 0.4156,
      "step": 1907
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.331219425913787e-05,
      "loss": 0.4455,
      "step": 1908
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3305612056184285e-05,
      "loss": 0.4265,
      "step": 1909
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3299028244975929e-05,
      "loss": 0.4418,
      "step": 1910
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3292442828715975e-05,
      "loss": 0.4507,
      "step": 1911
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.3285855810608377e-05,
      "loss": 0.4064,
      "step": 1912
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3279267193857872e-05,
      "loss": 0.4164,
      "step": 1913
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3272676981669968e-05,
      "loss": 0.4185,
      "step": 1914
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3266085177250952e-05,
      "loss": 0.4287,
      "step": 1915
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.325949178380788e-05,
      "loss": 0.4252,
      "step": 1916
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3252896804548594e-05,
      "loss": 0.4358,
      "step": 1917
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3246300242681698e-05,
      "loss": 0.4441,
      "step": 1918
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3239702101416568e-05,
      "loss": 0.4322,
      "step": 1919
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3233102383963341e-05,
      "loss": 0.4539,
      "step": 1920
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.322650109353294e-05,
      "loss": 0.423,
      "step": 1921
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3219898233337036e-05,
      "loss": 0.4372,
      "step": 1922
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3213293806588068e-05,
      "loss": 0.4137,
      "step": 1923
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3206687816499242e-05,
      "loss": 0.4231,
      "step": 1924
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3200080266284519e-05,
      "loss": 0.4355,
      "step": 1925
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3193471159158621e-05,
      "loss": 0.4374,
      "step": 1926
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3186860498337028e-05,
      "loss": 0.4269,
      "step": 1927
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.3180248287035977e-05,
      "loss": 0.4109,
      "step": 1928
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3173634528472452e-05,
      "loss": 0.4227,
      "step": 1929
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3167019225864203e-05,
      "loss": 0.4282,
      "step": 1930
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.316040238242972e-05,
      "loss": 0.437,
      "step": 1931
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3153784001388249e-05,
      "loss": 0.4206,
      "step": 1932
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3147164085959778e-05,
      "loss": 0.4319,
      "step": 1933
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3140542639365047e-05,
      "loss": 0.4193,
      "step": 1934
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3133919664825539e-05,
      "loss": 0.4291,
      "step": 1935
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3127295165563476e-05,
      "loss": 0.4598,
      "step": 1936
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3120669144801834e-05,
      "loss": 0.4355,
      "step": 1937
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3114041605764319e-05,
      "loss": 0.4293,
      "step": 1938
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3107412551675373e-05,
      "loss": 0.4258,
      "step": 1939
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3100781985760188e-05,
      "loss": 0.4374,
      "step": 1940
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3094149911244678e-05,
      "loss": 0.4112,
      "step": 1941
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3087516331355501e-05,
      "loss": 0.4174,
      "step": 1942
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.3080881249320038e-05,
      "loss": 0.4202,
      "step": 1943
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.3074244668366412e-05,
      "loss": 0.4491,
      "step": 1944
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.3067606591723463e-05,
      "loss": 0.4179,
      "step": 1945
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.3060967022620766e-05,
      "loss": 0.4171,
      "step": 1946
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.3054325964288626e-05,
      "loss": 0.4096,
      "step": 1947
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.3047683419958062e-05,
      "loss": 0.4226,
      "step": 1948
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.3041039392860824e-05,
      "loss": 0.4188,
      "step": 1949
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.3034393886229381e-05,
      "loss": 0.4414,
      "step": 1950
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.3027746903296924e-05,
      "loss": 0.4453,
      "step": 1951
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.3021098447297358e-05,
      "loss": 0.4093,
      "step": 1952
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.3014448521465304e-05,
      "loss": 0.4308,
      "step": 1953
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.3007797129036104e-05,
      "loss": 0.4244,
      "step": 1954
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.3001144273245807e-05,
      "loss": 0.4092,
      "step": 1955
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.2994489957331183e-05,
      "loss": 0.4297,
      "step": 1956
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.2987834184529701e-05,
      "loss": 0.4276,
      "step": 1957
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.2981176958079549e-05,
      "loss": 0.444,
      "step": 1958
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.2974518281219614e-05,
      "loss": 0.436,
      "step": 1959
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2967858157189495e-05,
      "loss": 0.4308,
      "step": 1960
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.296119658922949e-05,
      "loss": 0.4268,
      "step": 1961
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2954533580580603e-05,
      "loss": 0.4543,
      "step": 1962
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2947869134484541e-05,
      "loss": 0.4184,
      "step": 1963
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.29412032541837e-05,
      "loss": 0.4278,
      "step": 1964
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2934535942921189e-05,
      "loss": 0.4362,
      "step": 1965
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.29278672039408e-05,
      "loss": 0.4522,
      "step": 1966
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2921197040487027e-05,
      "loss": 0.4205,
      "step": 1967
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2914525455805056e-05,
      "loss": 0.4325,
      "step": 1968
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.290785245314076e-05,
      "loss": 0.4154,
      "step": 1969
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2901178035740709e-05,
      "loss": 0.4341,
      "step": 1970
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2894502206852153e-05,
      "loss": 0.4174,
      "step": 1971
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2887824969723035e-05,
      "loss": 0.4308,
      "step": 1972
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2881146327601981e-05,
      "loss": 0.4416,
      "step": 1973
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.2874466283738303e-05,
      "loss": 0.3959,
      "step": 1974
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2867784841381994e-05,
      "loss": 0.433,
      "step": 1975
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2861102003783722e-05,
      "loss": 0.418,
      "step": 1976
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2854417774194841e-05,
      "loss": 0.4286,
      "step": 1977
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.284773215586738e-05,
      "loss": 0.4246,
      "step": 1978
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.284104515205404e-05,
      "loss": 0.4258,
      "step": 1979
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2834356766008198e-05,
      "loss": 0.4398,
      "step": 1980
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2827667000983907e-05,
      "loss": 0.4423,
      "step": 1981
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2820975860235892e-05,
      "loss": 0.435,
      "step": 1982
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2814283347019538e-05,
      "loss": 0.4196,
      "step": 1983
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2807589464590908e-05,
      "loss": 0.4563,
      "step": 1984
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2800894216206725e-05,
      "loss": 0.4323,
      "step": 1985
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2794197605124375e-05,
      "loss": 0.4242,
      "step": 1986
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2787499634601914e-05,
      "loss": 0.4209,
      "step": 1987
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2780800307898057e-05,
      "loss": 0.4267,
      "step": 1988
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2774099628272171e-05,
      "loss": 0.4221,
      "step": 1989
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.2767397598984293e-05,
      "loss": 0.428,
      "step": 1990
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2760694223295112e-05,
      "loss": 0.4151,
      "step": 1991
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2753989504465967e-05,
      "loss": 0.4498,
      "step": 1992
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2747283445758858e-05,
      "loss": 0.4426,
      "step": 1993
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2740576050436433e-05,
      "loss": 0.4274,
      "step": 1994
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2733867321761992e-05,
      "loss": 0.4081,
      "step": 1995
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2727157262999481e-05,
      "loss": 0.4093,
      "step": 1996
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2720445877413496e-05,
      "loss": 0.4261,
      "step": 1997
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2713733168269275e-05,
      "loss": 0.4435,
      "step": 1998
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2707019138832704e-05,
      "loss": 0.42,
      "step": 1999
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.270030379237031e-05,
      "loss": 0.4532,
      "step": 2000
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.269358713214926e-05,
      "loss": 0.4308,
      "step": 2001
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2686869161437364e-05,
      "loss": 0.4175,
      "step": 2002
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2680149883503062e-05,
      "loss": 0.4339,
      "step": 2003
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2673429301615431e-05,
      "loss": 0.4274,
      "step": 2004
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.2666707419044193e-05,
      "loss": 0.4081,
      "step": 2005
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2659984239059693e-05,
      "loss": 0.4311,
      "step": 2006
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2653259764932904e-05,
      "loss": 0.4371,
      "step": 2007
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2646533999935442e-05,
      "loss": 0.4438,
      "step": 2008
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.263980694733954e-05,
      "loss": 0.4429,
      "step": 2009
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2633078610418062e-05,
      "loss": 0.4403,
      "step": 2010
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.262634899244449e-05,
      "loss": 0.4135,
      "step": 2011
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2619618096692942e-05,
      "loss": 0.4291,
      "step": 2012
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2612885926438147e-05,
      "loss": 0.4358,
      "step": 2013
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2606152484955458e-05,
      "loss": 0.4297,
      "step": 2014
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2599417775520849e-05,
      "loss": 0.427,
      "step": 2015
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.25926818014109e-05,
      "loss": 0.4376,
      "step": 2016
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2585944565902823e-05,
      "loss": 0.4364,
      "step": 2017
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.257920607227443e-05,
      "loss": 0.4074,
      "step": 2018
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2572466323804154e-05,
      "loss": 0.3826,
      "step": 2019
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.256572532377103e-05,
      "loss": 0.4122,
      "step": 2020
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.2558983075454708e-05,
      "loss": 0.4045,
      "step": 2021
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2552239582135446e-05,
      "loss": 0.4306,
      "step": 2022
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.25454948470941e-05,
      "loss": 0.4083,
      "step": 2023
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.253874887361214e-05,
      "loss": 0.4269,
      "step": 2024
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2532001664971634e-05,
      "loss": 0.4528,
      "step": 2025
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2525253224455249e-05,
      "loss": 0.4374,
      "step": 2026
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2518503555346258e-05,
      "loss": 0.4297,
      "step": 2027
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2511752660928523e-05,
      "loss": 0.4236,
      "step": 2028
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2505000544486507e-05,
      "loss": 0.4148,
      "step": 2029
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2498247209305267e-05,
      "loss": 0.4376,
      "step": 2030
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2491492658670452e-05,
      "loss": 0.4091,
      "step": 2031
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2484736895868306e-05,
      "loss": 0.4188,
      "step": 2032
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2477979924185655e-05,
      "loss": 0.4161,
      "step": 2033
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2471221746909923e-05,
      "loss": 0.4444,
      "step": 2034
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2464462367329114e-05,
      "loss": 0.4501,
      "step": 2035
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.2457701788731812e-05,
      "loss": 0.4092,
      "step": 2036
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2450940014407204e-05,
      "loss": 0.4386,
      "step": 2037
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2444177047645036e-05,
      "loss": 0.4299,
      "step": 2038
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2437412891735648e-05,
      "loss": 0.4215,
      "step": 2039
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2430647549969949e-05,
      "loss": 0.4457,
      "step": 2040
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2423881025639438e-05,
      "loss": 0.4466,
      "step": 2041
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2417113322036172e-05,
      "loss": 0.4283,
      "step": 2042
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2410344442452797e-05,
      "loss": 0.4232,
      "step": 2043
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2403574390182529e-05,
      "loss": 0.4146,
      "step": 2044
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2396803168519147e-05,
      "loss": 0.4215,
      "step": 2045
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2390030780757e-05,
      "loss": 0.4293,
      "step": 2046
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2383257230191013e-05,
      "loss": 0.4346,
      "step": 2047
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2376482520116666e-05,
      "loss": 0.426,
      "step": 2048
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2369706653830012e-05,
      "loss": 0.4196,
      "step": 2049
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2362929634627663e-05,
      "loss": 0.4538,
      "step": 2050
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2356151465806785e-05,
      "loss": 0.426,
      "step": 2051
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.2349372150665117e-05,
      "loss": 0.4342,
      "step": 2052
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2342591692500949e-05,
      "loss": 0.4327,
      "step": 2053
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2335810094613123e-05,
      "loss": 0.4106,
      "step": 2054
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2329027360301038e-05,
      "loss": 0.4295,
      "step": 2055
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2322243492864651e-05,
      "loss": 0.4306,
      "step": 2056
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2315458495604467e-05,
      "loss": 0.4453,
      "step": 2057
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2308672371821532e-05,
      "loss": 0.4387,
      "step": 2058
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.230188512481746e-05,
      "loss": 0.4147,
      "step": 2059
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2295096757894389e-05,
      "loss": 0.4233,
      "step": 2060
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2288307274355018e-05,
      "loss": 0.4111,
      "step": 2061
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2281516677502586e-05,
      "loss": 0.4147,
      "step": 2062
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2274724970640865e-05,
      "loss": 0.4381,
      "step": 2063
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2267932157074178e-05,
      "loss": 0.4401,
      "step": 2064
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2261138240107382e-05,
      "loss": 0.4494,
      "step": 2065
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.225434322304587e-05,
      "loss": 0.4307,
      "step": 2066
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.2247547109195567e-05,
      "loss": 0.4288,
      "step": 2067
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.224074990186294e-05,
      "loss": 0.4196,
      "step": 2068
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2233951604354983e-05,
      "loss": 0.4536,
      "step": 2069
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2227152219979224e-05,
      "loss": 0.429,
      "step": 2070
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2220351752043708e-05,
      "loss": 0.4448,
      "step": 2071
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2213550203857025e-05,
      "loss": 0.4361,
      "step": 2072
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2206747578728275e-05,
      "loss": 0.4393,
      "step": 2073
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2199943879967092e-05,
      "loss": 0.4367,
      "step": 2074
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2193139110883626e-05,
      "loss": 0.4207,
      "step": 2075
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2186333274788558e-05,
      "loss": 0.419,
      "step": 2076
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2179526374993068e-05,
      "loss": 0.425,
      "step": 2077
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2172718414808877e-05,
      "loss": 0.4317,
      "step": 2078
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2165909397548208e-05,
      "loss": 0.4431,
      "step": 2079
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.21590993265238e-05,
      "loss": 0.4372,
      "step": 2080
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2152288205048909e-05,
      "loss": 0.4135,
      "step": 2081
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2145476036437294e-05,
      "loss": 0.4288,
      "step": 2082
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.2138662824003232e-05,
      "loss": 0.4395,
      "step": 2083
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2131848571061501e-05,
      "loss": 0.4621,
      "step": 2084
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2125033280927395e-05,
      "loss": 0.4232,
      "step": 2085
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.21182169569167e-05,
      "loss": 0.4043,
      "step": 2086
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2111399602345715e-05,
      "loss": 0.4035,
      "step": 2087
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2104581220531237e-05,
      "loss": 0.4058,
      "step": 2088
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2097761814790561e-05,
      "loss": 0.4058,
      "step": 2089
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2090941388441482e-05,
      "loss": 0.4194,
      "step": 2090
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2084119944802288e-05,
      "loss": 0.4188,
      "step": 2091
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2077297487191771e-05,
      "loss": 0.4258,
      "step": 2092
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2070474018929204e-05,
      "loss": 0.4232,
      "step": 2093
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2063649543334364e-05,
      "loss": 0.4352,
      "step": 2094
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2056824063727508e-05,
      "loss": 0.4206,
      "step": 2095
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2049997583429389e-05,
      "loss": 0.4601,
      "step": 2096
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2043170105761242e-05,
      "loss": 0.4073,
      "step": 2097
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.2036341634044785e-05,
      "loss": 0.4246,
      "step": 2098
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.202951217160223e-05,
      "loss": 0.4149,
      "step": 2099
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.202268172175626e-05,
      "loss": 0.4096,
      "step": 2100
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.2015850287830045e-05,
      "loss": 0.461,
      "step": 2101
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.200901787314723e-05,
      "loss": 0.4246,
      "step": 2102
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.2002184481031938e-05,
      "loss": 0.4161,
      "step": 2103
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.1995350114808772e-05,
      "loss": 0.4206,
      "step": 2104
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.1988514777802802e-05,
      "loss": 0.4184,
      "step": 2105
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.1981678473339576e-05,
      "loss": 0.4309,
      "step": 2106
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.1974841204745107e-05,
      "loss": 0.4245,
      "step": 2107
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.1968002975345882e-05,
      "loss": 0.4242,
      "step": 2108
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.1961163788468854e-05,
      "loss": 0.4388,
      "step": 2109
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.1954323647441439e-05,
      "loss": 0.442,
      "step": 2110
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.1947482555591525e-05,
      "loss": 0.4521,
      "step": 2111
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.194064051624745e-05,
      "loss": 0.4427,
      "step": 2112
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.1933797532738029e-05,
      "loss": 0.4253,
      "step": 2113
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.1926953608392522e-05,
      "loss": 0.4489,
      "step": 2114
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1920108746540655e-05,
      "loss": 0.427,
      "step": 2115
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1913262950512605e-05,
      "loss": 0.441,
      "step": 2116
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1906416223639008e-05,
      "loss": 0.4593,
      "step": 2117
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1899568569250951e-05,
      "loss": 0.4321,
      "step": 2118
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1892719990679968e-05,
      "loss": 0.4235,
      "step": 2119
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1885870491258054e-05,
      "loss": 0.4341,
      "step": 2120
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1879020074317641e-05,
      "loss": 0.4356,
      "step": 2121
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1872168743191613e-05,
      "loss": 0.4333,
      "step": 2122
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1865316501213293e-05,
      "loss": 0.4352,
      "step": 2123
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.185846335171645e-05,
      "loss": 0.4127,
      "step": 2124
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1851609298035305e-05,
      "loss": 0.4274,
      "step": 2125
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1844754343504503e-05,
      "loss": 0.4287,
      "step": 2126
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1837898491459135e-05,
      "loss": 0.4248,
      "step": 2127
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1831041745234728e-05,
      "loss": 0.4318,
      "step": 2128
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.1824184108167238e-05,
      "loss": 0.429,
      "step": 2129
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.181732558359307e-05,
      "loss": 0.4283,
      "step": 2130
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1810466174849044e-05,
      "loss": 0.4444,
      "step": 2131
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.180360588527242e-05,
      "loss": 0.4356,
      "step": 2132
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1796744718200881e-05,
      "loss": 0.4079,
      "step": 2133
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1789882676972541e-05,
      "loss": 0.4299,
      "step": 2134
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1783019764925938e-05,
      "loss": 0.4073,
      "step": 2135
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.177615598540003e-05,
      "loss": 0.4423,
      "step": 2136
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1769291341734206e-05,
      "loss": 0.442,
      "step": 2137
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1762425837268263e-05,
      "loss": 0.4504,
      "step": 2138
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1755559475342431e-05,
      "loss": 0.4251,
      "step": 2139
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1748692259297347e-05,
      "loss": 0.4423,
      "step": 2140
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1741824192474065e-05,
      "loss": 0.4173,
      "step": 2141
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1734955278214057e-05,
      "loss": 0.4112,
      "step": 2142
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.17280855198592e-05,
      "loss": 0.4119,
      "step": 2143
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.172121492075179e-05,
      "loss": 0.4239,
      "step": 2144
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.1714343484234528e-05,
      "loss": 0.4146,
      "step": 2145
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.170747121365052e-05,
      "loss": 0.4141,
      "step": 2146
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1700598112343287e-05,
      "loss": 0.4253,
      "step": 2147
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.169372418365674e-05,
      "loss": 0.4192,
      "step": 2148
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1686849430935204e-05,
      "loss": 0.4463,
      "step": 2149
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.16799738575234e-05,
      "loss": 0.4097,
      "step": 2150
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1673097466766452e-05,
      "loss": 0.4295,
      "step": 2151
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1666220262009877e-05,
      "loss": 0.447,
      "step": 2152
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.165934224659959e-05,
      "loss": 0.4461,
      "step": 2153
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1652463423881898e-05,
      "loss": 0.4251,
      "step": 2154
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1645583797203511e-05,
      "loss": 0.451,
      "step": 2155
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1638703369911517e-05,
      "loss": 0.438,
      "step": 2156
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1631822145353398e-05,
      "loss": 0.4396,
      "step": 2157
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.1624940126877027e-05,
      "loss": 0.4546,
      "step": 2158
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.161805731783066e-05,
      "loss": 0.4187,
      "step": 2159
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.161117372156294e-05,
      "loss": 0.4226,
      "step": 2160
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1604289341422889e-05,
      "loss": 0.4361,
      "step": 2161
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1597404180759917e-05,
      "loss": 0.4258,
      "step": 2162
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1590518242923807e-05,
      "loss": 0.429,
      "step": 2163
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1583631531264723e-05,
      "loss": 0.4326,
      "step": 2164
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1576744049133209e-05,
      "loss": 0.4261,
      "step": 2165
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1569855799880174e-05,
      "loss": 0.4465,
      "step": 2166
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1562966786856916e-05,
      "loss": 0.4252,
      "step": 2167
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1556077013415084e-05,
      "loss": 0.4284,
      "step": 2168
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1549186482906715e-05,
      "loss": 0.4416,
      "step": 2169
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1542295198684206e-05,
      "loss": 0.4374,
      "step": 2170
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.153540316410032e-05,
      "loss": 0.4483,
      "step": 2171
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.152851038250819e-05,
      "loss": 0.425,
      "step": 2172
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.152161685726131e-05,
      "loss": 0.4552,
      "step": 2173
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1514722591713529e-05,
      "loss": 0.4016,
      "step": 2174
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.1507827589219072e-05,
      "loss": 0.4156,
      "step": 2175
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.150093185313251e-05,
      "loss": 0.4351,
      "step": 2176
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.149403538680877e-05,
      "loss": 0.4558,
      "step": 2177
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1487138193603142e-05,
      "loss": 0.4501,
      "step": 2178
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1480240276871264e-05,
      "loss": 0.465,
      "step": 2179
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.147334163996913e-05,
      "loss": 0.4207,
      "step": 2180
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.146644228625308e-05,
      "loss": 0.4066,
      "step": 2181
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1459542219079808e-05,
      "loss": 0.4499,
      "step": 2182
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1452641441806348e-05,
      "loss": 0.4082,
      "step": 2183
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1445739957790087e-05,
      "loss": 0.4345,
      "step": 2184
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1438837770388753e-05,
      "loss": 0.4425,
      "step": 2185
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1431934882960412e-05,
      "loss": 0.4226,
      "step": 2186
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1425031298863478e-05,
      "loss": 0.4112,
      "step": 2187
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.14181270214567e-05,
      "loss": 0.4274,
      "step": 2188
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.141122205409916e-05,
      "loss": 0.4209,
      "step": 2189
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1404316400150288e-05,
      "loss": 0.4546,
      "step": 2190
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.1397410062969833e-05,
      "loss": 0.4144,
      "step": 2191
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1390503045917892e-05,
      "loss": 0.4325,
      "step": 2192
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1383595352354874e-05,
      "loss": 0.4419,
      "step": 2193
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1376686985641536e-05,
      "loss": 0.4277,
      "step": 2194
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1369777949138948e-05,
      "loss": 0.432,
      "step": 2195
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1362868246208519e-05,
      "loss": 0.4255,
      "step": 2196
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1355957880211968e-05,
      "loss": 0.4191,
      "step": 2197
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1349046854511347e-05,
      "loss": 0.4292,
      "step": 2198
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1342135172469024e-05,
      "loss": 0.4237,
      "step": 2199
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1335222837447692e-05,
      "loss": 0.4277,
      "step": 2200
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1328309852810357e-05,
      "loss": 0.4359,
      "step": 2201
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.132139622192034e-05,
      "loss": 0.4327,
      "step": 2202
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1314481948141277e-05,
      "loss": 0.4337,
      "step": 2203
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1307567034837123e-05,
      "loss": 0.4173,
      "step": 2204
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1300651485372133e-05,
      "loss": 0.4298,
      "step": 2205
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.129373530311088e-05,
      "loss": 0.3972,
      "step": 2206
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.1286818491418246e-05,
      "loss": 0.4238,
      "step": 2207
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.127990105365941e-05,
      "loss": 0.4247,
      "step": 2208
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1272982993199868e-05,
      "loss": 0.42,
      "step": 2209
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1266064313405404e-05,
      "loss": 0.4307,
      "step": 2210
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.125914501764212e-05,
      "loss": 0.4442,
      "step": 2211
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1252225109276404e-05,
      "loss": 0.4535,
      "step": 2212
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.124530459167495e-05,
      "loss": 0.4112,
      "step": 2213
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1238383468204744e-05,
      "loss": 0.4447,
      "step": 2214
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1231461742233068e-05,
      "loss": 0.4326,
      "step": 2215
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.12245394171275e-05,
      "loss": 0.4229,
      "step": 2216
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1217616496255907e-05,
      "loss": 0.4473,
      "step": 2217
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1210692982986447e-05,
      "loss": 0.4137,
      "step": 2218
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1203768880687562e-05,
      "loss": 0.43,
      "step": 2219
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1196844192727984e-05,
      "loss": 0.4262,
      "step": 2220
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1189918922476734e-05,
      "loss": 0.4383,
      "step": 2221
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.1182993073303107e-05,
      "loss": 0.4052,
      "step": 2222
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1176066648576687e-05,
      "loss": 0.4396,
      "step": 2223
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1169139651667334e-05,
      "loss": 0.4297,
      "step": 2224
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1162212085945192e-05,
      "loss": 0.4511,
      "step": 2225
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1155283954780676e-05,
      "loss": 0.4364,
      "step": 2226
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1148355261544477e-05,
      "loss": 0.4333,
      "step": 2227
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1141426009607562e-05,
      "loss": 0.4391,
      "step": 2228
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1134496202341166e-05,
      "loss": 0.4312,
      "step": 2229
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1127565843116798e-05,
      "loss": 0.4129,
      "step": 2230
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1120634935306233e-05,
      "loss": 0.4435,
      "step": 2231
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1113703482281515e-05,
      "loss": 0.4212,
      "step": 2232
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1106771487414951e-05,
      "loss": 0.4344,
      "step": 2233
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1099838954079117e-05,
      "loss": 0.4174,
      "step": 2234
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1092905885646839e-05,
      "loss": 0.4095,
      "step": 2235
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1085972285491213e-05,
      "loss": 0.4349,
      "step": 2236
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1079038156985599e-05,
      "loss": 0.4342,
      "step": 2237
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.1072103503503599e-05,
      "loss": 0.43,
      "step": 2238
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.1065168328419083e-05,
      "loss": 0.4189,
      "step": 2239
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.1058232635106167e-05,
      "loss": 0.4195,
      "step": 2240
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.1051296426939221e-05,
      "loss": 0.4066,
      "step": 2241
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.104435970729287e-05,
      "loss": 0.4174,
      "step": 2242
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.1037422479541986e-05,
      "loss": 0.4261,
      "step": 2243
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.103048474706168e-05,
      "loss": 0.4136,
      "step": 2244
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.1023546513227322e-05,
      "loss": 0.4238,
      "step": 2245
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.1016607781414514e-05,
      "loss": 0.4252,
      "step": 2246
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.1009668554999109e-05,
      "loss": 0.4352,
      "step": 2247
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.1002728837357192e-05,
      "loss": 0.4289,
      "step": 2248
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.0995788631865099e-05,
      "loss": 0.4431,
      "step": 2249
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.098884794189939e-05,
      "loss": 0.4489,
      "step": 2250
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.0981906770836872e-05,
      "loss": 0.416,
      "step": 2251
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.097496512205458e-05,
      "loss": 0.4389,
      "step": 2252
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.096802299892978e-05,
      "loss": 0.4525,
      "step": 2253
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0961080404839974e-05,
      "loss": 0.4428,
      "step": 2254
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0954137343162892e-05,
      "loss": 0.4387,
      "step": 2255
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0947193817276485e-05,
      "loss": 0.4506,
      "step": 2256
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0940249830558937e-05,
      "loss": 0.4141,
      "step": 2257
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0933305386388656e-05,
      "loss": 0.4101,
      "step": 2258
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0926360488144269e-05,
      "loss": 0.4423,
      "step": 2259
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0919415139204625e-05,
      "loss": 0.423,
      "step": 2260
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0912469342948793e-05,
      "loss": 0.4375,
      "step": 2261
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0905523102756061e-05,
      "loss": 0.4355,
      "step": 2262
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0898576422005931e-05,
      "loss": 0.4253,
      "step": 2263
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.089162930407812e-05,
      "loss": 0.4319,
      "step": 2264
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0884681752352555e-05,
      "loss": 0.4516,
      "step": 2265
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.087773377020938e-05,
      "loss": 0.4629,
      "step": 2266
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0870785361028939e-05,
      "loss": 0.4308,
      "step": 2267
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0863836528191795e-05,
      "loss": 0.4223,
      "step": 2268
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.0856887275078712e-05,
      "loss": 0.4195,
      "step": 2269
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0849937605070658e-05,
      "loss": 0.4207,
      "step": 2270
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.08429875215488e-05,
      "loss": 0.4168,
      "step": 2271
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0836037027894515e-05,
      "loss": 0.4262,
      "step": 2272
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0829086127489373e-05,
      "loss": 0.4466,
      "step": 2273
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0822134823715139e-05,
      "loss": 0.4293,
      "step": 2274
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0815183119953787e-05,
      "loss": 0.4344,
      "step": 2275
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0808231019587472e-05,
      "loss": 0.4166,
      "step": 2276
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.080127852599855e-05,
      "loss": 0.4125,
      "step": 2277
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.079432564256957e-05,
      "loss": 0.4259,
      "step": 2278
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0787372372683258e-05,
      "loss": 0.4281,
      "step": 2279
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0780418719722544e-05,
      "loss": 0.4286,
      "step": 2280
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0773464687070529e-05,
      "loss": 0.4163,
      "step": 2281
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0766510278110514e-05,
      "loss": 0.4269,
      "step": 2282
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.0759555496225973e-05,
      "loss": 0.4179,
      "step": 2283
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.075260034480056e-05,
      "loss": 0.4658,
      "step": 2284
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0745644827218119e-05,
      "loss": 0.4263,
      "step": 2285
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0738688946862661e-05,
      "loss": 0.423,
      "step": 2286
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0731732707118385e-05,
      "loss": 0.4318,
      "step": 2287
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0724776111369654e-05,
      "loss": 0.4226,
      "step": 2288
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.071781916300101e-05,
      "loss": 0.4424,
      "step": 2289
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0710861865397166e-05,
      "loss": 0.4371,
      "step": 2290
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0703904221943e-05,
      "loss": 0.4699,
      "step": 2291
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0696946236023566e-05,
      "loss": 0.4236,
      "step": 2292
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.068998791102408e-05,
      "loss": 0.3996,
      "step": 2293
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0683029250329924e-05,
      "loss": 0.4376,
      "step": 2294
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0676070257326643e-05,
      "loss": 0.4494,
      "step": 2295
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0669110935399944e-05,
      "loss": 0.4223,
      "step": 2296
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0662151287935696e-05,
      "loss": 0.4123,
      "step": 2297
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0655191318319921e-05,
      "loss": 0.4299,
      "step": 2298
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.06482310299388e-05,
      "loss": 0.4544,
      "step": 2299
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.0641270426178677e-05,
      "loss": 0.4538,
      "step": 2300
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0634309510426037e-05,
      "loss": 0.4344,
      "step": 2301
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0627348286067521e-05,
      "loss": 0.4537,
      "step": 2302
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0620386756489927e-05,
      "loss": 0.4405,
      "step": 2303
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0613424925080194e-05,
      "loss": 0.4199,
      "step": 2304
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0606462795225409e-05,
      "loss": 0.4343,
      "step": 2305
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0599500370312805e-05,
      "loss": 0.4314,
      "step": 2306
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.059253765372976e-05,
      "loss": 0.4158,
      "step": 2307
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.058557464886379e-05,
      "loss": 0.4466,
      "step": 2308
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0578611359102555e-05,
      "loss": 0.4236,
      "step": 2309
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0571647787833853e-05,
      "loss": 0.42,
      "step": 2310
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0564683938445616e-05,
      "loss": 0.43,
      "step": 2311
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.055771981432592e-05,
      "loss": 0.4326,
      "step": 2312
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0550755418862962e-05,
      "loss": 0.4192,
      "step": 2313
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.054379075544508e-05,
      "loss": 0.4059,
      "step": 2314
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.0536825827460739e-05,
      "loss": 0.437,
      "step": 2315
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0529860638298535e-05,
      "loss": 0.4249,
      "step": 2316
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0522895191347186e-05,
      "loss": 0.4035,
      "step": 2317
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0515929489995544e-05,
      "loss": 0.4231,
      "step": 2318
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0508963537632573e-05,
      "loss": 0.4082,
      "step": 2319
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0501997337647372e-05,
      "loss": 0.4255,
      "step": 2320
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0495030893429152e-05,
      "loss": 0.4043,
      "step": 2321
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0488064208367246e-05,
      "loss": 0.4268,
      "step": 2322
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0481097285851101e-05,
      "loss": 0.423,
      "step": 2323
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0474130129270281e-05,
      "loss": 0.4212,
      "step": 2324
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.046716274201447e-05,
      "loss": 0.4388,
      "step": 2325
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0460195127473456e-05,
      "loss": 0.4583,
      "step": 2326
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.045322728903714e-05,
      "loss": 0.4313,
      "step": 2327
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0446259230095531e-05,
      "loss": 0.433,
      "step": 2328
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.043929095403875e-05,
      "loss": 0.4292,
      "step": 2329
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0432322464257019e-05,
      "loss": 0.4373,
      "step": 2330
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.0425353764140665e-05,
      "loss": 0.4321,
      "step": 2331
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0418384857080118e-05,
      "loss": 0.3905,
      "step": 2332
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0411415746465907e-05,
      "loss": 0.4367,
      "step": 2333
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0404446435688665e-05,
      "loss": 0.4155,
      "step": 2334
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0397476928139113e-05,
      "loss": 0.4448,
      "step": 2335
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.039050722720808e-05,
      "loss": 0.419,
      "step": 2336
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0383537336286476e-05,
      "loss": 0.4331,
      "step": 2337
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0376567258765316e-05,
      "loss": 0.4344,
      "step": 2338
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0369596998035698e-05,
      "loss": 0.43,
      "step": 2339
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0362626557488811e-05,
      "loss": 0.4255,
      "step": 2340
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.035565594051593e-05,
      "loss": 0.426,
      "step": 2341
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0348685150508417e-05,
      "loss": 0.4458,
      "step": 2342
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0341714190857719e-05,
      "loss": 0.4432,
      "step": 2343
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0334743064955367e-05,
      "loss": 0.4096,
      "step": 2344
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0327771776192968e-05,
      "loss": 0.3971,
      "step": 2345
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.0320800327962212e-05,
      "loss": 0.4154,
      "step": 2346
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0313828723654868e-05,
      "loss": 0.4308,
      "step": 2347
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0306856966662776e-05,
      "loss": 0.4024,
      "step": 2348
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.029988506037785e-05,
      "loss": 0.4361,
      "step": 2349
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0292913008192088e-05,
      "loss": 0.401,
      "step": 2350
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0285940813497544e-05,
      "loss": 0.4204,
      "step": 2351
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.027896847968635e-05,
      "loss": 0.414,
      "step": 2352
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0271996010150704e-05,
      "loss": 0.4312,
      "step": 2353
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0265023408282866e-05,
      "loss": 0.4215,
      "step": 2354
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0258050677475167e-05,
      "loss": 0.4141,
      "step": 2355
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0251077821119998e-05,
      "loss": 0.3989,
      "step": 2356
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0244104842609812e-05,
      "loss": 0.4123,
      "step": 2357
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0237131745337117e-05,
      "loss": 0.4386,
      "step": 2358
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0230158532694483e-05,
      "loss": 0.4149,
      "step": 2359
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0223185208074538e-05,
      "loss": 0.4236,
      "step": 2360
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0216211774869952e-05,
      "loss": 0.4166,
      "step": 2361
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.0209238236473472e-05,
      "loss": 0.4121,
      "step": 2362
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0202264596277866e-05,
      "loss": 0.4145,
      "step": 2363
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0195290857675982e-05,
      "loss": 0.4245,
      "step": 2364
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0188317024060693e-05,
      "loss": 0.4348,
      "step": 2365
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0181343098824928e-05,
      "loss": 0.431,
      "step": 2366
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0174369085361655e-05,
      "loss": 0.4241,
      "step": 2367
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0167394987063894e-05,
      "loss": 0.4166,
      "step": 2368
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.01604208073247e-05,
      "loss": 0.4305,
      "step": 2369
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0153446549537164e-05,
      "loss": 0.4066,
      "step": 2370
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0146472217094428e-05,
      "loss": 0.4231,
      "step": 2371
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0139497813389654e-05,
      "loss": 0.4196,
      "step": 2372
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0132523341816053e-05,
      "loss": 0.4189,
      "step": 2373
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0125548805766852e-05,
      "loss": 0.4462,
      "step": 2374
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0118574208635332e-05,
      "loss": 0.4115,
      "step": 2375
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0111599553814788e-05,
      "loss": 0.4124,
      "step": 2376
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.0104624844698545e-05,
      "loss": 0.4266,
      "step": 2377
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0097650084679957e-05,
      "loss": 0.4168,
      "step": 2378
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0090675277152404e-05,
      "loss": 0.4301,
      "step": 2379
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.008370042550928e-05,
      "loss": 0.4086,
      "step": 2380
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0076725533144016e-05,
      "loss": 0.4281,
      "step": 2381
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.006975060345005e-05,
      "loss": 0.4426,
      "step": 2382
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0062775639820843e-05,
      "loss": 0.4382,
      "step": 2383
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0055800645649874e-05,
      "loss": 0.444,
      "step": 2384
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0048825624330631e-05,
      "loss": 0.4347,
      "step": 2385
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0041850579256623e-05,
      "loss": 0.4045,
      "step": 2386
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0034875513821363e-05,
      "loss": 0.4212,
      "step": 2387
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.002790043141838e-05,
      "loss": 0.4197,
      "step": 2388
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.002092533544121e-05,
      "loss": 0.4275,
      "step": 2389
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.0013950229283392e-05,
      "loss": 0.4219,
      "step": 2390
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.000697511633847e-05,
      "loss": 0.4227,
      "step": 2391
    },
    {
      "epoch": 1.54,
      "learning_rate": 1e-05,
      "loss": 0.4318,
      "step": 2392
    },
    {
      "epoch": 1.54,
      "learning_rate": 9.993024883661528e-06,
      "loss": 0.4417,
      "step": 2393
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.986049770716613e-06,
      "loss": 0.396,
      "step": 2394
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.979074664558792e-06,
      "loss": 0.4404,
      "step": 2395
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.972099568581621e-06,
      "loss": 0.4353,
      "step": 2396
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.96512448617864e-06,
      "loss": 0.4176,
      "step": 2397
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.958149420743382e-06,
      "loss": 0.4157,
      "step": 2398
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.95117437566937e-06,
      "loss": 0.4129,
      "step": 2399
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.94419935435013e-06,
      "loss": 0.419,
      "step": 2400
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.937224360179159e-06,
      "loss": 0.4222,
      "step": 2401
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.930249396549954e-06,
      "loss": 0.4278,
      "step": 2402
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.923274466855988e-06,
      "loss": 0.429,
      "step": 2403
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.916299574490722e-06,
      "loss": 0.4392,
      "step": 2404
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.9093247228476e-06,
      "loss": 0.4563,
      "step": 2405
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.902349915320045e-06,
      "loss": 0.4321,
      "step": 2406
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.895375155301456e-06,
      "loss": 0.4357,
      "step": 2407
    },
    {
      "epoch": 1.55,
      "learning_rate": 9.888400446185212e-06,
      "loss": 0.4369,
      "step": 2408
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.88142579136467e-06,
      "loss": 0.4253,
      "step": 2409
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.874451194233148e-06,
      "loss": 0.4395,
      "step": 2410
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.867476658183954e-06,
      "loss": 0.4332,
      "step": 2411
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.860502186610349e-06,
      "loss": 0.4463,
      "step": 2412
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.853527782905576e-06,
      "loss": 0.4065,
      "step": 2413
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.846553450462838e-06,
      "loss": 0.4378,
      "step": 2414
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.839579192675305e-06,
      "loss": 0.4235,
      "step": 2415
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.832605012936107e-06,
      "loss": 0.422,
      "step": 2416
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.825630914638348e-06,
      "loss": 0.4322,
      "step": 2417
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.818656901175076e-06,
      "loss": 0.4271,
      "step": 2418
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.811682975939309e-06,
      "loss": 0.437,
      "step": 2419
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.80470914232402e-06,
      "loss": 0.4277,
      "step": 2420
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.797735403722132e-06,
      "loss": 0.431,
      "step": 2421
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.790761763526533e-06,
      "loss": 0.4128,
      "step": 2422
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.78378822513005e-06,
      "loss": 0.4051,
      "step": 2423
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.77681479192547e-06,
      "loss": 0.4185,
      "step": 2424
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.76984146730552e-06,
      "loss": 0.4296,
      "step": 2425
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.762868254662888e-06,
      "loss": 0.4174,
      "step": 2426
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.755895157390191e-06,
      "loss": 0.4171,
      "step": 2427
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.748922178880005e-06,
      "loss": 0.4578,
      "step": 2428
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.741949322524835e-06,
      "loss": 0.4476,
      "step": 2429
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.734976591717136e-06,
      "loss": 0.4423,
      "step": 2430
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.7280039898493e-06,
      "loss": 0.4181,
      "step": 2431
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.721031520313653e-06,
      "loss": 0.4439,
      "step": 2432
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.714059186502459e-06,
      "loss": 0.4522,
      "step": 2433
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.707086991807912e-06,
      "loss": 0.4501,
      "step": 2434
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.700114939622151e-06,
      "loss": 0.4533,
      "step": 2435
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.693143033337228e-06,
      "loss": 0.4148,
      "step": 2436
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.686171276345137e-06,
      "loss": 0.4632,
      "step": 2437
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.679199672037791e-06,
      "loss": 0.4252,
      "step": 2438
    },
    {
      "epoch": 1.57,
      "learning_rate": 9.672228223807037e-06,
      "loss": 0.4358,
      "step": 2439
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.665256935044636e-06,
      "loss": 0.4292,
      "step": 2440
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.658285809142285e-06,
      "loss": 0.4117,
      "step": 2441
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.651314849491586e-06,
      "loss": 0.4107,
      "step": 2442
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.644344059484072e-06,
      "loss": 0.4623,
      "step": 2443
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.637373442511192e-06,
      "loss": 0.4317,
      "step": 2444
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.630403001964302e-06,
      "loss": 0.4463,
      "step": 2445
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.623432741234686e-06,
      "loss": 0.4142,
      "step": 2446
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.616462663713524e-06,
      "loss": 0.4252,
      "step": 2447
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.609492772791924e-06,
      "loss": 0.4176,
      "step": 2448
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.602523071860889e-06,
      "loss": 0.4322,
      "step": 2449
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.59555356431134e-06,
      "loss": 0.4304,
      "step": 2450
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.588584253534096e-06,
      "loss": 0.448,
      "step": 2451
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.581615142919887e-06,
      "loss": 0.4312,
      "step": 2452
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.574646235859339e-06,
      "loss": 0.4287,
      "step": 2453
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.567677535742988e-06,
      "loss": 0.4245,
      "step": 2454
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.560709045961254e-06,
      "loss": 0.4146,
      "step": 2455
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.55374076990447e-06,
      "loss": 0.4217,
      "step": 2456
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.546772710962864e-06,
      "loss": 0.4246,
      "step": 2457
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.539804872526547e-06,
      "loss": 0.428,
      "step": 2458
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.532837257985533e-06,
      "loss": 0.421,
      "step": 2459
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.525869870729719e-06,
      "loss": 0.4298,
      "step": 2460
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.518902714148904e-06,
      "loss": 0.4235,
      "step": 2461
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.511935791632758e-06,
      "loss": 0.4447,
      "step": 2462
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.504969106570853e-06,
      "loss": 0.4272,
      "step": 2463
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.49800266235263e-06,
      "loss": 0.4469,
      "step": 2464
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.491036462367428e-06,
      "loss": 0.42,
      "step": 2465
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.484070510004459e-06,
      "loss": 0.4263,
      "step": 2466
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.477104808652818e-06,
      "loss": 0.4177,
      "step": 2467
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.470139361701469e-06,
      "loss": 0.4379,
      "step": 2468
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.463174172539261e-06,
      "loss": 0.4264,
      "step": 2469
    },
    {
      "epoch": 1.59,
      "learning_rate": 9.456209244554923e-06,
      "loss": 0.4268,
      "step": 2470
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.449244581137038e-06,
      "loss": 0.4258,
      "step": 2471
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.442280185674084e-06,
      "loss": 0.4472,
      "step": 2472
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.435316061554384e-06,
      "loss": 0.4356,
      "step": 2473
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.428352212166149e-06,
      "loss": 0.398,
      "step": 2474
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.421388640897447e-06,
      "loss": 0.4038,
      "step": 2475
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.414425351136215e-06,
      "loss": 0.4012,
      "step": 2476
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.407462346270242e-06,
      "loss": 0.4291,
      "step": 2477
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.4004996296872e-06,
      "loss": 0.4123,
      "step": 2478
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.393537204774594e-06,
      "loss": 0.4316,
      "step": 2479
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.386575074919806e-06,
      "loss": 0.476,
      "step": 2480
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.379613243510075e-06,
      "loss": 0.4505,
      "step": 2481
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.372651713932479e-06,
      "loss": 0.4498,
      "step": 2482
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.365690489573966e-06,
      "loss": 0.4354,
      "step": 2483
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.358729573821325e-06,
      "loss": 0.4259,
      "step": 2484
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.351768970061201e-06,
      "loss": 0.409,
      "step": 2485
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.344808681680082e-06,
      "loss": 0.4221,
      "step": 2486
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.337848712064309e-06,
      "loss": 0.4508,
      "step": 2487
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.330889064600058e-06,
      "loss": 0.431,
      "step": 2488
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.323929742673362e-06,
      "loss": 0.4373,
      "step": 2489
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.31697074967008e-06,
      "loss": 0.4361,
      "step": 2490
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.310012088975924e-06,
      "loss": 0.4206,
      "step": 2491
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.303053763976436e-06,
      "loss": 0.4234,
      "step": 2492
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.296095778057001e-06,
      "loss": 0.3996,
      "step": 2493
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.289138134602837e-06,
      "loss": 0.4353,
      "step": 2494
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.28218083699899e-06,
      "loss": 0.4154,
      "step": 2495
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.275223888630348e-06,
      "loss": 0.4216,
      "step": 2496
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.268267292881615e-06,
      "loss": 0.4198,
      "step": 2497
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.26131105313734e-06,
      "loss": 0.4337,
      "step": 2498
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.254355172781883e-06,
      "loss": 0.4181,
      "step": 2499
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.247399655199444e-06,
      "loss": 0.4266,
      "step": 2500
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.24044450377403e-06,
      "loss": 0.4218,
      "step": 2501
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.233489721889491e-06,
      "loss": 0.4449,
      "step": 2502
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.226535312929473e-06,
      "loss": 0.4076,
      "step": 2503
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.219581280277463e-06,
      "loss": 0.4256,
      "step": 2504
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.212627627316744e-06,
      "loss": 0.4363,
      "step": 2505
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.205674357430432e-06,
      "loss": 0.4118,
      "step": 2506
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.198721474001451e-06,
      "loss": 0.446,
      "step": 2507
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.191768980412528e-06,
      "loss": 0.4254,
      "step": 2508
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.184816880046217e-06,
      "loss": 0.4254,
      "step": 2509
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.177865176284863e-06,
      "loss": 0.4279,
      "step": 2510
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.170913872510634e-06,
      "loss": 0.4295,
      "step": 2511
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.163962972105488e-06,
      "loss": 0.4459,
      "step": 2512
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.157012478451205e-06,
      "loss": 0.4103,
      "step": 2513
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.150062394929347e-06,
      "loss": 0.4162,
      "step": 2514
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.143112724921293e-06,
      "loss": 0.4081,
      "step": 2515
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.136163471808207e-06,
      "loss": 0.423,
      "step": 2516
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.129214638971065e-06,
      "loss": 0.4219,
      "step": 2517
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.122266229790624e-06,
      "loss": 0.4373,
      "step": 2518
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.115318247647448e-06,
      "loss": 0.4469,
      "step": 2519
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.108370695921884e-06,
      "loss": 0.4391,
      "step": 2520
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.101423577994069e-06,
      "loss": 0.4593,
      "step": 2521
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.09447689724394e-06,
      "loss": 0.4357,
      "step": 2522
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.087530657051207e-06,
      "loss": 0.4462,
      "step": 2523
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.080584860795378e-06,
      "loss": 0.4232,
      "step": 2524
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.073639511855734e-06,
      "loss": 0.4231,
      "step": 2525
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.066694613611346e-06,
      "loss": 0.4175,
      "step": 2526
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.059750169441064e-06,
      "loss": 0.4173,
      "step": 2527
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.05280618272352e-06,
      "loss": 0.4316,
      "step": 2528
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.045862656837112e-06,
      "loss": 0.4212,
      "step": 2529
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.038919595160026e-06,
      "loss": 0.4255,
      "step": 2530
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.031977001070222e-06,
      "loss": 0.4409,
      "step": 2531
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.025034877945422e-06,
      "loss": 0.4423,
      "step": 2532
    },
    {
      "epoch": 1.64,
      "learning_rate": 9.018093229163133e-06,
      "loss": 0.431,
      "step": 2533
    },
    {
      "epoch": 1.64,
      "learning_rate": 9.011152058100612e-06,
      "loss": 0.4217,
      "step": 2534
    },
    {
      "epoch": 1.64,
      "learning_rate": 9.004211368134906e-06,
      "loss": 0.4174,
      "step": 2535
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.99727116264281e-06,
      "loss": 0.427,
      "step": 2536
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.990331445000898e-06,
      "loss": 0.4481,
      "step": 2537
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.98339221858549e-06,
      "loss": 0.4177,
      "step": 2538
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.976453486772683e-06,
      "loss": 0.4379,
      "step": 2539
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.969515252938323e-06,
      "loss": 0.4458,
      "step": 2540
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.96257752045802e-06,
      "loss": 0.4245,
      "step": 2541
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.955640292707131e-06,
      "loss": 0.4278,
      "step": 2542
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.948703573060779e-06,
      "loss": 0.425,
      "step": 2543
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.941767364893836e-06,
      "loss": 0.4328,
      "step": 2544
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.934831671580919e-06,
      "loss": 0.4335,
      "step": 2545
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.927896496496402e-06,
      "loss": 0.4654,
      "step": 2546
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.920961843014403e-06,
      "loss": 0.4082,
      "step": 2547
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.914027714508788e-06,
      "loss": 0.4559,
      "step": 2548
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.907094114353165e-06,
      "loss": 0.4421,
      "step": 2549
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.90016104592089e-06,
      "loss": 0.4186,
      "step": 2550
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.89322851258505e-06,
      "loss": 0.4188,
      "step": 2551
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.88629651771849e-06,
      "loss": 0.4166,
      "step": 2552
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.879365064693769e-06,
      "loss": 0.4193,
      "step": 2553
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.872434156883207e-06,
      "loss": 0.4256,
      "step": 2554
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.865503797658836e-06,
      "loss": 0.4198,
      "step": 2555
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.85857399039244e-06,
      "loss": 0.4363,
      "step": 2556
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.851644738455526e-06,
      "loss": 0.4192,
      "step": 2557
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.844716045219325e-06,
      "loss": 0.4143,
      "step": 2558
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.837787914054812e-06,
      "loss": 0.4138,
      "step": 2559
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.830860348332666e-06,
      "loss": 0.4127,
      "step": 2560
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.823933351423316e-06,
      "loss": 0.422,
      "step": 2561
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.817006926696894e-06,
      "loss": 0.4262,
      "step": 2562
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.810081077523271e-06,
      "loss": 0.4226,
      "step": 2563
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.803155807272019e-06,
      "loss": 0.4253,
      "step": 2564
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.796231119312443e-06,
      "loss": 0.414,
      "step": 2565
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.789307017013556e-06,
      "loss": 0.4272,
      "step": 2566
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.782383503744096e-06,
      "loss": 0.4346,
      "step": 2567
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.775460582872502e-06,
      "loss": 0.4306,
      "step": 2568
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.768538257766933e-06,
      "loss": 0.4101,
      "step": 2569
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.761616531795258e-06,
      "loss": 0.4125,
      "step": 2570
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.754695408325052e-06,
      "loss": 0.4276,
      "step": 2571
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.7477748907236e-06,
      "loss": 0.4397,
      "step": 2572
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.74085498235788e-06,
      "loss": 0.4041,
      "step": 2573
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.733935686594599e-06,
      "loss": 0.4144,
      "step": 2574
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.727017006800136e-06,
      "loss": 0.4299,
      "step": 2575
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.720098946340594e-06,
      "loss": 0.4395,
      "step": 2576
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.713181508581758e-06,
      "loss": 0.4296,
      "step": 2577
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.706264696889122e-06,
      "loss": 0.4375,
      "step": 2578
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.699348514627869e-06,
      "loss": 0.4385,
      "step": 2579
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.69243296516288e-06,
      "loss": 0.4278,
      "step": 2580
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.685518051858724e-06,
      "loss": 0.4295,
      "step": 2581
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.678603778079661e-06,
      "loss": 0.4201,
      "step": 2582
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.671690147189647e-06,
      "loss": 0.4056,
      "step": 2583
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.664777162552308e-06,
      "loss": 0.4261,
      "step": 2584
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.657864827530978e-06,
      "loss": 0.4715,
      "step": 2585
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.650953145488656e-06,
      "loss": 0.4093,
      "step": 2586
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.644042119788037e-06,
      "loss": 0.4268,
      "step": 2587
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.637131753791485e-06,
      "loss": 0.4616,
      "step": 2588
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.630222050861055e-06,
      "loss": 0.4324,
      "step": 2589
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.623313014358467e-06,
      "loss": 0.4424,
      "step": 2590
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.61640464764513e-06,
      "loss": 0.4013,
      "step": 2591
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.609496954082113e-06,
      "loss": 0.4302,
      "step": 2592
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.602589937030165e-06,
      "loss": 0.4389,
      "step": 2593
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.595683599849716e-06,
      "loss": 0.4108,
      "step": 2594
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.58877794590084e-06,
      "loss": 0.4417,
      "step": 2595
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.581872978543305e-06,
      "loss": 0.4288,
      "step": 2596
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.574968701136524e-06,
      "loss": 0.4163,
      "step": 2597
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.568065117039592e-06,
      "loss": 0.4052,
      "step": 2598
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.56116222961125e-06,
      "loss": 0.4496,
      "step": 2599
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.554260042209918e-06,
      "loss": 0.4124,
      "step": 2600
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.547358558193656e-06,
      "loss": 0.4274,
      "step": 2601
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.5404577809202e-06,
      "loss": 0.3958,
      "step": 2602
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.533557713746924e-06,
      "loss": 0.4527,
      "step": 2603
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.526658360030876e-06,
      "loss": 0.4523,
      "step": 2604
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.51975972312874e-06,
      "loss": 0.4344,
      "step": 2605
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.51286180639686e-06,
      "loss": 0.4558,
      "step": 2606
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.505964613191233e-06,
      "loss": 0.4464,
      "step": 2607
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.499068146867492e-06,
      "loss": 0.4166,
      "step": 2608
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.49217241078093e-06,
      "loss": 0.4092,
      "step": 2609
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.485277408286471e-06,
      "loss": 0.452,
      "step": 2610
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.478383142738696e-06,
      "loss": 0.4318,
      "step": 2611
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.471489617491813e-06,
      "loss": 0.4305,
      "step": 2612
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.464596835899683e-06,
      "loss": 0.4079,
      "step": 2613
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.457704801315797e-06,
      "loss": 0.4572,
      "step": 2614
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.45081351709329e-06,
      "loss": 0.4227,
      "step": 2615
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.44392298658492e-06,
      "loss": 0.4424,
      "step": 2616
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.437033213143091e-06,
      "loss": 0.4213,
      "step": 2617
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.430144200119827e-06,
      "loss": 0.4141,
      "step": 2618
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.423255950866793e-06,
      "loss": 0.4224,
      "step": 2619
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.41636846873528e-06,
      "loss": 0.4151,
      "step": 2620
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.409481757076196e-06,
      "loss": 0.4406,
      "step": 2621
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.402595819240086e-06,
      "loss": 0.4094,
      "step": 2622
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.395710658577113e-06,
      "loss": 0.4336,
      "step": 2623
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.388826278437066e-06,
      "loss": 0.4258,
      "step": 2624
    },
    {
      "epoch": 1.69,
      "learning_rate": 8.381942682169343e-06,
      "loss": 0.4462,
      "step": 2625
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.375059873122978e-06,
      "loss": 0.4102,
      "step": 2626
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.368177854646606e-06,
      "loss": 0.4088,
      "step": 2627
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.361296630088488e-06,
      "loss": 0.4061,
      "step": 2628
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.35441620279649e-06,
      "loss": 0.4134,
      "step": 2629
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.347536576118102e-06,
      "loss": 0.437,
      "step": 2630
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.340657753400411e-06,
      "loss": 0.4261,
      "step": 2631
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.333779737990124e-06,
      "loss": 0.4257,
      "step": 2632
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.326902533233551e-06,
      "loss": 0.4351,
      "step": 2633
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.320026142476601e-06,
      "loss": 0.4547,
      "step": 2634
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.3131505690648e-06,
      "loss": 0.4376,
      "step": 2635
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.306275816343262e-06,
      "loss": 0.4236,
      "step": 2636
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.29940188765672e-06,
      "loss": 0.4503,
      "step": 2637
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.292528786349481e-06,
      "loss": 0.417,
      "step": 2638
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.285656515765476e-06,
      "loss": 0.4315,
      "step": 2639
    },
    {
      "epoch": 1.7,
      "learning_rate": 8.278785079248211e-06,
      "loss": 0.429,
      "step": 2640
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.271914480140804e-06,
      "loss": 0.4249,
      "step": 2641
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.265044721785946e-06,
      "loss": 0.4223,
      "step": 2642
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.258175807525935e-06,
      "loss": 0.4329,
      "step": 2643
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.251307740702656e-06,
      "loss": 0.4267,
      "step": 2644
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.244440524657569e-06,
      "loss": 0.4263,
      "step": 2645
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.237574162731739e-06,
      "loss": 0.4381,
      "step": 2646
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.230708658265798e-06,
      "loss": 0.4134,
      "step": 2647
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.223844014599973e-06,
      "loss": 0.4602,
      "step": 2648
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.216980235074065e-06,
      "loss": 0.4387,
      "step": 2649
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.210117323027464e-06,
      "loss": 0.4314,
      "step": 2650
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.203255281799122e-06,
      "loss": 0.4321,
      "step": 2651
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.196394114727586e-06,
      "loss": 0.414,
      "step": 2652
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.189533825150958e-06,
      "loss": 0.4285,
      "step": 2653
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.182674416406936e-06,
      "loss": 0.4379,
      "step": 2654
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.175815891832764e-06,
      "loss": 0.4186,
      "step": 2655
    },
    {
      "epoch": 1.71,
      "learning_rate": 8.168958254765275e-06,
      "loss": 0.4522,
      "step": 2656
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.162101508540867e-06,
      "loss": 0.4029,
      "step": 2657
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.155245656495497e-06,
      "loss": 0.4084,
      "step": 2658
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.148390701964697e-06,
      "loss": 0.44,
      "step": 2659
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.14153664828355e-06,
      "loss": 0.4398,
      "step": 2660
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.134683498786713e-06,
      "loss": 0.4347,
      "step": 2661
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.12783125680839e-06,
      "loss": 0.4231,
      "step": 2662
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.120979925682364e-06,
      "loss": 0.4406,
      "step": 2663
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.114129508741947e-06,
      "loss": 0.4278,
      "step": 2664
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.107280009320034e-06,
      "loss": 0.4274,
      "step": 2665
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.100431430749052e-06,
      "loss": 0.4169,
      "step": 2666
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.093583776360996e-06,
      "loss": 0.4452,
      "step": 2667
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.086737049487398e-06,
      "loss": 0.4317,
      "step": 2668
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.079891253459345e-06,
      "loss": 0.44,
      "step": 2669
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.07304639160748e-06,
      "loss": 0.4374,
      "step": 2670
    },
    {
      "epoch": 1.72,
      "learning_rate": 8.066202467261973e-06,
      "loss": 0.4126,
      "step": 2671
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.059359483752551e-06,
      "loss": 0.4123,
      "step": 2672
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.052517444408478e-06,
      "loss": 0.437,
      "step": 2673
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.045676352558563e-06,
      "loss": 0.4201,
      "step": 2674
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.038836211531147e-06,
      "loss": 0.4192,
      "step": 2675
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.031997024654123e-06,
      "loss": 0.4264,
      "step": 2676
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.025158795254896e-06,
      "loss": 0.4424,
      "step": 2677
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.01832152666043e-06,
      "loss": 0.443,
      "step": 2678
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.011485222197201e-06,
      "loss": 0.4233,
      "step": 2679
    },
    {
      "epoch": 1.73,
      "learning_rate": 8.00464988519123e-06,
      "loss": 0.4262,
      "step": 2680
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.997815518968063e-06,
      "loss": 0.4171,
      "step": 2681
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.990982126852772e-06,
      "loss": 0.4037,
      "step": 2682
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.984149712169957e-06,
      "loss": 0.4185,
      "step": 2683
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.977318278243742e-06,
      "loss": 0.4162,
      "step": 2684
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.970487828397774e-06,
      "loss": 0.4224,
      "step": 2685
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.963658365955216e-06,
      "loss": 0.4016,
      "step": 2686
    },
    {
      "epoch": 1.73,
      "learning_rate": 7.956829894238763e-06,
      "loss": 0.4309,
      "step": 2687
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.950002416570614e-06,
      "loss": 0.4203,
      "step": 2688
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.943175936272495e-06,
      "loss": 0.423,
      "step": 2689
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.936350456665639e-06,
      "loss": 0.4123,
      "step": 2690
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.929525981070798e-06,
      "loss": 0.4267,
      "step": 2691
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.92270251280823e-06,
      "loss": 0.4135,
      "step": 2692
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.91588005519771e-06,
      "loss": 0.4095,
      "step": 2693
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.909058611558522e-06,
      "loss": 0.4505,
      "step": 2694
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.90223818520944e-06,
      "loss": 0.4363,
      "step": 2695
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.895418779468766e-06,
      "loss": 0.4273,
      "step": 2696
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.888600397654285e-06,
      "loss": 0.4338,
      "step": 2697
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.881783043083304e-06,
      "loss": 0.4136,
      "step": 2698
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.874966719072609e-06,
      "loss": 0.4278,
      "step": 2699
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.868151428938502e-06,
      "loss": 0.4416,
      "step": 2700
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.861337175996772e-06,
      "loss": 0.4263,
      "step": 2701
    },
    {
      "epoch": 1.74,
      "learning_rate": 7.854523963562713e-06,
      "loss": 0.427,
      "step": 2702
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.847711794951097e-06,
      "loss": 0.4138,
      "step": 2703
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.840900673476204e-06,
      "loss": 0.4114,
      "step": 2704
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.834090602451795e-06,
      "loss": 0.4046,
      "step": 2705
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.827281585191123e-06,
      "loss": 0.4306,
      "step": 2706
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.820473625006935e-06,
      "loss": 0.4322,
      "step": 2707
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.813666725211445e-06,
      "loss": 0.4404,
      "step": 2708
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.806860889116375e-06,
      "loss": 0.4284,
      "step": 2709
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.800056120032908e-06,
      "loss": 0.4393,
      "step": 2710
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.793252421271729e-06,
      "loss": 0.4427,
      "step": 2711
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.786449796142979e-06,
      "loss": 0.4459,
      "step": 2712
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.779648247956295e-06,
      "loss": 0.4123,
      "step": 2713
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.77284778002078e-06,
      "loss": 0.41,
      "step": 2714
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.76604839564502e-06,
      "loss": 0.4177,
      "step": 2715
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.759250098137061e-06,
      "loss": 0.4301,
      "step": 2716
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.752452890804436e-06,
      "loss": 0.4298,
      "step": 2717
    },
    {
      "epoch": 1.75,
      "learning_rate": 7.745656776954133e-06,
      "loss": 0.436,
      "step": 2718
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.738861759892618e-06,
      "loss": 0.4182,
      "step": 2719
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.732067842925823e-06,
      "loss": 0.4139,
      "step": 2720
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.725275029359136e-06,
      "loss": 0.4139,
      "step": 2721
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.71848332249742e-06,
      "loss": 0.3978,
      "step": 2722
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.711692725644983e-06,
      "loss": 0.4288,
      "step": 2723
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.704903242105616e-06,
      "loss": 0.4264,
      "step": 2724
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.698114875182546e-06,
      "loss": 0.4311,
      "step": 2725
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.691327628178471e-06,
      "loss": 0.381,
      "step": 2726
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.684541504395538e-06,
      "loss": 0.4422,
      "step": 2727
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.677756507135354e-06,
      "loss": 0.4327,
      "step": 2728
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.670972639698963e-06,
      "loss": 0.4494,
      "step": 2729
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.664189905386878e-06,
      "loss": 0.4327,
      "step": 2730
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.657408307499054e-06,
      "loss": 0.446,
      "step": 2731
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.650627849334881e-06,
      "loss": 0.4219,
      "step": 2732
    },
    {
      "epoch": 1.76,
      "learning_rate": 7.643848534193216e-06,
      "loss": 0.4174,
      "step": 2733
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.63707036537234e-06,
      "loss": 0.423,
      "step": 2734
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.63029334616999e-06,
      "loss": 0.4144,
      "step": 2735
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.623517479883335e-06,
      "loss": 0.4094,
      "step": 2736
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.616742769808991e-06,
      "loss": 0.4252,
      "step": 2737
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.609969219243002e-06,
      "loss": 0.4277,
      "step": 2738
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.603196831480859e-06,
      "loss": 0.4192,
      "step": 2739
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.596425609817474e-06,
      "loss": 0.4358,
      "step": 2740
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.589655557547206e-06,
      "loss": 0.443,
      "step": 2741
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.5828866779638326e-06,
      "loss": 0.4373,
      "step": 2742
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.5761189743605665e-06,
      "loss": 0.4126,
      "step": 2743
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.569352450030054e-06,
      "loss": 0.4486,
      "step": 2744
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.562587108264354e-06,
      "loss": 0.407,
      "step": 2745
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.555822952354968e-06,
      "loss": 0.4225,
      "step": 2746
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.549059985592797e-06,
      "loss": 0.4112,
      "step": 2747
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.542298211268189e-06,
      "loss": 0.4332,
      "step": 2748
    },
    {
      "epoch": 1.77,
      "learning_rate": 7.535537632670891e-06,
      "loss": 0.4206,
      "step": 2749
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.528778253090082e-06,
      "loss": 0.4285,
      "step": 2750
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.522020075814348e-06,
      "loss": 0.4313,
      "step": 2751
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.515263104131699e-06,
      "loss": 0.4224,
      "step": 2752
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.50850734132955e-06,
      "loss": 0.4581,
      "step": 2753
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.501752790694738e-06,
      "loss": 0.3965,
      "step": 2754
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.494999455513496e-06,
      "loss": 0.3934,
      "step": 2755
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.488247339071478e-06,
      "loss": 0.4173,
      "step": 2756
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.481496444653746e-06,
      "loss": 0.4215,
      "step": 2757
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.47474677554475e-06,
      "loss": 0.4282,
      "step": 2758
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.46799833502837e-06,
      "loss": 0.3934,
      "step": 2759
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.461251126387863e-06,
      "loss": 0.4222,
      "step": 2760
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.454505152905904e-06,
      "loss": 0.4166,
      "step": 2761
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.447760417864558e-06,
      "loss": 0.4333,
      "step": 2762
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.4410169245452965e-06,
      "loss": 0.4425,
      "step": 2763
    },
    {
      "epoch": 1.78,
      "learning_rate": 7.434274676228973e-06,
      "loss": 0.4213,
      "step": 2764
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.427533676195852e-06,
      "loss": 0.4305,
      "step": 2765
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.420793927725572e-06,
      "loss": 0.4337,
      "step": 2766
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.414055434097181e-06,
      "loss": 0.4088,
      "step": 2767
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.407318198589102e-06,
      "loss": 0.4304,
      "step": 2768
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.4005822244791545e-06,
      "loss": 0.4205,
      "step": 2769
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.393847515044543e-06,
      "loss": 0.421,
      "step": 2770
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.387114073561854e-06,
      "loss": 0.4186,
      "step": 2771
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.380381903307061e-06,
      "loss": 0.4256,
      "step": 2772
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.373651007555511e-06,
      "loss": 0.4266,
      "step": 2773
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.366921389581945e-06,
      "loss": 0.4428,
      "step": 2774
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.360193052660463e-06,
      "loss": 0.4313,
      "step": 2775
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.353466000064563e-06,
      "loss": 0.4182,
      "step": 2776
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.346740235067099e-06,
      "loss": 0.4113,
      "step": 2777
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.340015760940313e-06,
      "loss": 0.4162,
      "step": 2778
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.333292580955808e-06,
      "loss": 0.4154,
      "step": 2779
    },
    {
      "epoch": 1.79,
      "learning_rate": 7.326570698384569e-06,
      "loss": 0.4217,
      "step": 2780
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.319850116496942e-06,
      "loss": 0.4345,
      "step": 2781
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.313130838562638e-06,
      "loss": 0.4147,
      "step": 2782
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.3064128678507415e-06,
      "loss": 0.425,
      "step": 2783
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.299696207629692e-06,
      "loss": 0.4078,
      "step": 2784
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.2929808611673e-06,
      "loss": 0.4215,
      "step": 2785
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.286266831730728e-06,
      "loss": 0.4463,
      "step": 2786
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.279554122586509e-06,
      "loss": 0.4131,
      "step": 2787
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.2728427370005205e-06,
      "loss": 0.4381,
      "step": 2788
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.266132678238013e-06,
      "loss": 0.4329,
      "step": 2789
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.259423949563569e-06,
      "loss": 0.4226,
      "step": 2790
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.252716554241146e-06,
      "loss": 0.4433,
      "step": 2791
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.246010495534036e-06,
      "loss": 0.4521,
      "step": 2792
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.23930577670489e-06,
      "loss": 0.4214,
      "step": 2793
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.232602401015711e-06,
      "loss": 0.4069,
      "step": 2794
    },
    {
      "epoch": 1.8,
      "learning_rate": 7.225900371727832e-06,
      "loss": 0.421,
      "step": 2795
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.2191996921019485e-06,
      "loss": 0.4346,
      "step": 2796
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.212500365398087e-06,
      "loss": 0.4404,
      "step": 2797
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.205802394875629e-06,
      "loss": 0.4435,
      "step": 2798
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.199105783793279e-06,
      "loss": 0.4174,
      "step": 2799
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.1924105354090955e-06,
      "loss": 0.4256,
      "step": 2800
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.185716652980463e-06,
      "loss": 0.4328,
      "step": 2801
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.17902413976411e-06,
      "loss": 0.4126,
      "step": 2802
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.172332999016095e-06,
      "loss": 0.4172,
      "step": 2803
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.165643233991806e-06,
      "loss": 0.4244,
      "step": 2804
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.1589548479459645e-06,
      "loss": 0.3945,
      "step": 2805
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.152267844132623e-06,
      "loss": 0.4235,
      "step": 2806
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.1455822258051625e-06,
      "loss": 0.4358,
      "step": 2807
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.138897996216278e-06,
      "loss": 0.4362,
      "step": 2808
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.13221515861801e-06,
      "loss": 0.4166,
      "step": 2809
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.1255337162616965e-06,
      "loss": 0.4275,
      "step": 2810
    },
    {
      "epoch": 1.81,
      "learning_rate": 7.118853672398023e-06,
      "loss": 0.4256,
      "step": 2811
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.1121750302769685e-06,
      "loss": 0.4175,
      "step": 2812
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.105497793147852e-06,
      "loss": 0.4209,
      "step": 2813
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.098821964259295e-06,
      "loss": 0.4324,
      "step": 2814
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.0921475468592435e-06,
      "loss": 0.4142,
      "step": 2815
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.085474544194946e-06,
      "loss": 0.4385,
      "step": 2816
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.078802959512976e-06,
      "loss": 0.4354,
      "step": 2817
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.072132796059201e-06,
      "loss": 0.4204,
      "step": 2818
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.065464057078812e-06,
      "loss": 0.4273,
      "step": 2819
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.058796745816303e-06,
      "loss": 0.4058,
      "step": 2820
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.052130865515463e-06,
      "loss": 0.4199,
      "step": 2821
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.0454664194194e-06,
      "loss": 0.4337,
      "step": 2822
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.038803410770512e-06,
      "loss": 0.4132,
      "step": 2823
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.03214184281051e-06,
      "loss": 0.43,
      "step": 2824
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.025481718780388e-06,
      "loss": 0.4246,
      "step": 2825
    },
    {
      "epoch": 1.82,
      "learning_rate": 7.018823041920456e-06,
      "loss": 0.4206,
      "step": 2826
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.012165815470302e-06,
      "loss": 0.4487,
      "step": 2827
    },
    {
      "epoch": 1.83,
      "learning_rate": 7.0055100426688205e-06,
      "loss": 0.4262,
      "step": 2828
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.998855726754195e-06,
      "loss": 0.4044,
      "step": 2829
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.992202870963899e-06,
      "loss": 0.422,
      "step": 2830
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.985551478534699e-06,
      "loss": 0.415,
      "step": 2831
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.978901552702643e-06,
      "loss": 0.4085,
      "step": 2832
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.972253096703079e-06,
      "loss": 0.4532,
      "step": 2833
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.965606113770617e-06,
      "loss": 0.4523,
      "step": 2834
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.958960607139178e-06,
      "loss": 0.424,
      "step": 2835
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.95231658004194e-06,
      "loss": 0.4066,
      "step": 2836
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.945674035711381e-06,
      "loss": 0.4383,
      "step": 2837
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.939032977379237e-06,
      "loss": 0.4082,
      "step": 2838
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.932393408276543e-06,
      "loss": 0.4417,
      "step": 2839
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.925755331633592e-06,
      "loss": 0.4349,
      "step": 2840
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.919118750679966e-06,
      "loss": 0.4319,
      "step": 2841
    },
    {
      "epoch": 1.83,
      "learning_rate": 6.912483668644503e-06,
      "loss": 0.4162,
      "step": 2842
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.905850088755322e-06,
      "loss": 0.4311,
      "step": 2843
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.899218014239815e-06,
      "loss": 0.4281,
      "step": 2844
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.892587448324627e-06,
      "loss": 0.4132,
      "step": 2845
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.8859583942356855e-06,
      "loss": 0.4385,
      "step": 2846
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.879330855198168e-06,
      "loss": 0.4258,
      "step": 2847
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.872704834436526e-06,
      "loss": 0.4145,
      "step": 2848
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.866080335174466e-06,
      "loss": 0.4413,
      "step": 2849
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.8594573606349575e-06,
      "loss": 0.385,
      "step": 2850
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.852835914040226e-06,
      "loss": 0.4184,
      "step": 2851
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.846215998611757e-06,
      "loss": 0.3976,
      "step": 2852
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.839597617570282e-06,
      "loss": 0.4312,
      "step": 2853
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.8329807741358e-06,
      "loss": 0.4498,
      "step": 2854
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.826365471527551e-06,
      "loss": 0.3972,
      "step": 2855
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.8197517129640265e-06,
      "loss": 0.4283,
      "step": 2856
    },
    {
      "epoch": 1.84,
      "learning_rate": 6.8131395016629745e-06,
      "loss": 0.4455,
      "step": 2857
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.80652884084138e-06,
      "loss": 0.4272,
      "step": 2858
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.799919733715485e-06,
      "loss": 0.4309,
      "step": 2859
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.79331218350076e-06,
      "loss": 0.418,
      "step": 2860
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.786706193411935e-06,
      "loss": 0.4181,
      "step": 2861
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.780101766662966e-06,
      "loss": 0.4211,
      "step": 2862
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.773498906467062e-06,
      "loss": 0.4138,
      "step": 2863
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.766897616036661e-06,
      "loss": 0.4146,
      "step": 2864
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.760297898583439e-06,
      "loss": 0.4432,
      "step": 2865
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.753699757318304e-06,
      "loss": 0.4237,
      "step": 2866
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.74710319545141e-06,
      "loss": 0.4106,
      "step": 2867
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.740508216192121e-06,
      "loss": 0.4233,
      "step": 2868
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.7339148227490505e-06,
      "loss": 0.4389,
      "step": 2869
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.727323018330034e-06,
      "loss": 0.4186,
      "step": 2870
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.7207328061421276e-06,
      "loss": 0.4115,
      "step": 2871
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.714144189391625e-06,
      "loss": 0.4329,
      "step": 2872
    },
    {
      "epoch": 1.85,
      "learning_rate": 6.7075571712840285e-06,
      "loss": 0.4236,
      "step": 2873
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.700971755024077e-06,
      "loss": 0.4252,
      "step": 2874
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.694387943815719e-06,
      "loss": 0.4296,
      "step": 2875
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.6878057408621345e-06,
      "loss": 0.414,
      "step": 2876
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.681225149365704e-06,
      "loss": 0.4205,
      "step": 2877
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.674646172528045e-06,
      "loss": 0.4033,
      "step": 2878
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.668068813549966e-06,
      "loss": 0.42,
      "step": 2879
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.661493075631506e-06,
      "loss": 0.4132,
      "step": 2880
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.6549189619719145e-06,
      "loss": 0.4145,
      "step": 2881
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.648346475769637e-06,
      "loss": 0.4044,
      "step": 2882
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.641775620222342e-06,
      "loss": 0.436,
      "step": 2883
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.635206398526895e-06,
      "loss": 0.4154,
      "step": 2884
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.628638813879378e-06,
      "loss": 0.4402,
      "step": 2885
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.622072869475058e-06,
      "loss": 0.4217,
      "step": 2886
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.615508568508425e-06,
      "loss": 0.4281,
      "step": 2887
    },
    {
      "epoch": 1.86,
      "learning_rate": 6.60894591417315e-06,
      "loss": 0.4395,
      "step": 2888
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.6023849096621205e-06,
      "loss": 0.4213,
      "step": 2889
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.59582555816741e-06,
      "loss": 0.4099,
      "step": 2890
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.589267862880289e-06,
      "loss": 0.4125,
      "step": 2891
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.582711826991226e-06,
      "loss": 0.4097,
      "step": 2892
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.576157453689877e-06,
      "loss": 0.4336,
      "step": 2893
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.569604746165101e-06,
      "loss": 0.4052,
      "step": 2894
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.563053707604927e-06,
      "loss": 0.4379,
      "step": 2895
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.556504341196592e-06,
      "loss": 0.4071,
      "step": 2896
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.5499566501265036e-06,
      "loss": 0.4129,
      "step": 2897
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.543410637580265e-06,
      "loss": 0.4306,
      "step": 2898
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.536866306742658e-06,
      "loss": 0.4199,
      "step": 2899
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.5303236607976465e-06,
      "loss": 0.4397,
      "step": 2900
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.523782702928372e-06,
      "loss": 0.4158,
      "step": 2901
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.517243436317167e-06,
      "loss": 0.4239,
      "step": 2902
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.510705864145521e-06,
      "loss": 0.4208,
      "step": 2903
    },
    {
      "epoch": 1.87,
      "learning_rate": 6.504169989594121e-06,
      "loss": 0.4561,
      "step": 2904
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.497635815842809e-06,
      "loss": 0.396,
      "step": 2905
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.491103346070608e-06,
      "loss": 0.4241,
      "step": 2906
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.48457258345572e-06,
      "loss": 0.4542,
      "step": 2907
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.4780435311754986e-06,
      "loss": 0.4458,
      "step": 2908
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.471516192406481e-06,
      "loss": 0.4245,
      "step": 2909
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.464990570324361e-06,
      "loss": 0.4507,
      "step": 2910
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.458466668104005e-06,
      "loss": 0.4083,
      "step": 2911
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.451944488919433e-06,
      "loss": 0.4253,
      "step": 2912
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.445424035943839e-06,
      "loss": 0.414,
      "step": 2913
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.438905312349563e-06,
      "loss": 0.4186,
      "step": 2914
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.432388321308115e-06,
      "loss": 0.4212,
      "step": 2915
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.425873065990158e-06,
      "loss": 0.4101,
      "step": 2916
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.41935954956551e-06,
      "loss": 0.4295,
      "step": 2917
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.412847775203141e-06,
      "loss": 0.4204,
      "step": 2918
    },
    {
      "epoch": 1.88,
      "learning_rate": 6.406337746071181e-06,
      "loss": 0.4329,
      "step": 2919
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.3998294653369046e-06,
      "loss": 0.432,
      "step": 2920
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.393322936166733e-06,
      "loss": 0.4352,
      "step": 2921
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.386818161726246e-06,
      "loss": 0.4151,
      "step": 2922
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.3803151451801545e-06,
      "loss": 0.4429,
      "step": 2923
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.373813889692331e-06,
      "loss": 0.4194,
      "step": 2924
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.367314398425777e-06,
      "loss": 0.425,
      "step": 2925
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.360816674542644e-06,
      "loss": 0.3988,
      "step": 2926
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.354320721204219e-06,
      "loss": 0.4187,
      "step": 2927
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.347826541570936e-06,
      "loss": 0.4019,
      "step": 2928
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.341334138802351e-06,
      "loss": 0.4365,
      "step": 2929
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.334843516057168e-06,
      "loss": 0.4348,
      "step": 2930
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.3283546764932245e-06,
      "loss": 0.4365,
      "step": 2931
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.3218676232674815e-06,
      "loss": 0.4197,
      "step": 2932
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.315382359536042e-06,
      "loss": 0.4299,
      "step": 2933
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.30889888845413e-06,
      "loss": 0.4163,
      "step": 2934
    },
    {
      "epoch": 1.89,
      "learning_rate": 6.302417213176101e-06,
      "loss": 0.4447,
      "step": 2935
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.295937336855433e-06,
      "loss": 0.423,
      "step": 2936
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.289459262644739e-06,
      "loss": 0.4379,
      "step": 2937
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.282982993695741e-06,
      "loss": 0.433,
      "step": 2938
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.276508533159296e-06,
      "loss": 0.4228,
      "step": 2939
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.270035884185367e-06,
      "loss": 0.444,
      "step": 2940
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.26356504992305e-06,
      "loss": 0.4349,
      "step": 2941
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.257096033520551e-06,
      "loss": 0.4632,
      "step": 2942
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.25062883812519e-06,
      "loss": 0.4188,
      "step": 2943
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.244163466883405e-06,
      "loss": 0.4342,
      "step": 2944
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.2376999229407435e-06,
      "loss": 0.4039,
      "step": 2945
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.231238209441869e-06,
      "loss": 0.4381,
      "step": 2946
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.224778329530544e-06,
      "loss": 0.4164,
      "step": 2947
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.218320286349655e-06,
      "loss": 0.4223,
      "step": 2948
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.211864083041174e-06,
      "loss": 0.4308,
      "step": 2949
    },
    {
      "epoch": 1.9,
      "learning_rate": 6.2054097227462e-06,
      "loss": 0.4417,
      "step": 2950
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.198957208604919e-06,
      "loss": 0.4128,
      "step": 2951
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.192506543756626e-06,
      "loss": 0.4269,
      "step": 2952
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.186057731339713e-06,
      "loss": 0.4386,
      "step": 2953
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.179610774491678e-06,
      "loss": 0.4051,
      "step": 2954
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.173165676349103e-06,
      "loss": 0.4036,
      "step": 2955
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.1667224400476785e-06,
      "loss": 0.4331,
      "step": 2956
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.160281068722187e-06,
      "loss": 0.447,
      "step": 2957
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.153841565506493e-06,
      "loss": 0.4119,
      "step": 2958
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.147403933533567e-06,
      "loss": 0.4448,
      "step": 2959
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.140968175935458e-06,
      "loss": 0.4161,
      "step": 2960
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.134534295843308e-06,
      "loss": 0.4256,
      "step": 2961
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.128102296387344e-06,
      "loss": 0.4282,
      "step": 2962
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.121672180696884e-06,
      "loss": 0.4297,
      "step": 2963
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.115243951900316e-06,
      "loss": 0.4077,
      "step": 2964
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.108817613125124e-06,
      "loss": 0.4218,
      "step": 2965
    },
    {
      "epoch": 1.91,
      "learning_rate": 6.102393167497862e-06,
      "loss": 0.4102,
      "step": 2966
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.095970618144173e-06,
      "loss": 0.4148,
      "step": 2967
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.089549968188767e-06,
      "loss": 0.4201,
      "step": 2968
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.0831312207554395e-06,
      "loss": 0.417,
      "step": 2969
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.0767143789670525e-06,
      "loss": 0.4168,
      "step": 2970
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.070299445945544e-06,
      "loss": 0.4055,
      "step": 2971
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.063886424811929e-06,
      "loss": 0.4113,
      "step": 2972
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.057475318686278e-06,
      "loss": 0.4145,
      "step": 2973
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.051066130687747e-06,
      "loss": 0.4085,
      "step": 2974
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.044658863934543e-06,
      "loss": 0.4227,
      "step": 2975
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.038253521543951e-06,
      "loss": 0.4043,
      "step": 2976
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.031850106632313e-06,
      "loss": 0.424,
      "step": 2977
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.025448622315033e-06,
      "loss": 0.4231,
      "step": 2978
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.01904907170658e-06,
      "loss": 0.4113,
      "step": 2979
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.01265145792048e-06,
      "loss": 0.4168,
      "step": 2980
    },
    {
      "epoch": 1.92,
      "learning_rate": 6.00625578406931e-06,
      "loss": 0.4029,
      "step": 2981
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.999862053264716e-06,
      "loss": 0.4397,
      "step": 2982
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.993470268617393e-06,
      "loss": 0.4384,
      "step": 2983
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.987080433237082e-06,
      "loss": 0.4123,
      "step": 2984
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.980692550232587e-06,
      "loss": 0.4275,
      "step": 2985
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.974306622711758e-06,
      "loss": 0.4284,
      "step": 2986
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.967922653781489e-06,
      "loss": 0.413,
      "step": 2987
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.961540646547722e-06,
      "loss": 0.4179,
      "step": 2988
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.955160604115457e-06,
      "loss": 0.4338,
      "step": 2989
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.948782529588719e-06,
      "loss": 0.4171,
      "step": 2990
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.942406426070593e-06,
      "loss": 0.4258,
      "step": 2991
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.936032296663188e-06,
      "loss": 0.413,
      "step": 2992
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.929660144467666e-06,
      "loss": 0.4218,
      "step": 2993
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.9232899725842276e-06,
      "loss": 0.4325,
      "step": 2994
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.916921784112098e-06,
      "loss": 0.4225,
      "step": 2995
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.9105555821495486e-06,
      "loss": 0.4119,
      "step": 2996
    },
    {
      "epoch": 1.93,
      "learning_rate": 5.904191369793877e-06,
      "loss": 0.4224,
      "step": 2997
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.8978291501414195e-06,
      "loss": 0.435,
      "step": 2998
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.891468926287535e-06,
      "loss": 0.4124,
      "step": 2999
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.885110701326621e-06,
      "loss": 0.4449,
      "step": 3000
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.878754478352091e-06,
      "loss": 0.4176,
      "step": 3001
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.872400260456397e-06,
      "loss": 0.4019,
      "step": 3002
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.866048050731003e-06,
      "loss": 0.4324,
      "step": 3003
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.859697852266409e-06,
      "loss": 0.4087,
      "step": 3004
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.853349668152124e-06,
      "loss": 0.4213,
      "step": 3005
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.847003501476682e-06,
      "loss": 0.4168,
      "step": 3006
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.840659355327646e-06,
      "loss": 0.4227,
      "step": 3007
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.834317232791575e-06,
      "loss": 0.4324,
      "step": 3008
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.82797713695406e-06,
      "loss": 0.4138,
      "step": 3009
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.8216390708996955e-06,
      "loss": 0.411,
      "step": 3010
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.815303037712103e-06,
      "loss": 0.4195,
      "step": 3011
    },
    {
      "epoch": 1.94,
      "learning_rate": 5.808969040473893e-06,
      "loss": 0.4047,
      "step": 3012
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.8026370822667135e-06,
      "loss": 0.4537,
      "step": 3013
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.796307166171185e-06,
      "loss": 0.4446,
      "step": 3014
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.789979295266972e-06,
      "loss": 0.4163,
      "step": 3015
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.7836534726327175e-06,
      "loss": 0.4425,
      "step": 3016
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.77732970134608e-06,
      "loss": 0.4248,
      "step": 3017
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.771007984483716e-06,
      "loss": 0.41,
      "step": 3018
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.764688325121277e-06,
      "loss": 0.4213,
      "step": 3019
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.758370726333434e-06,
      "loss": 0.4241,
      "step": 3020
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.752055191193826e-06,
      "loss": 0.4123,
      "step": 3021
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.745741722775117e-06,
      "loss": 0.4326,
      "step": 3022
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.739430324148937e-06,
      "loss": 0.4223,
      "step": 3023
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.733120998385935e-06,
      "loss": 0.4372,
      "step": 3024
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.726813748555736e-06,
      "loss": 0.4294,
      "step": 3025
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.72050857772696e-06,
      "loss": 0.4215,
      "step": 3026
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.714205488967215e-06,
      "loss": 0.4484,
      "step": 3027
    },
    {
      "epoch": 1.95,
      "learning_rate": 5.707904485343094e-06,
      "loss": 0.4055,
      "step": 3028
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.701605569920179e-06,
      "loss": 0.4059,
      "step": 3029
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.695308745763034e-06,
      "loss": 0.4193,
      "step": 3030
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.6890140159352055e-06,
      "loss": 0.4276,
      "step": 3031
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.682721383499217e-06,
      "loss": 0.4227,
      "step": 3032
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.676430851516586e-06,
      "loss": 0.4097,
      "step": 3033
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.670142423047791e-06,
      "loss": 0.4216,
      "step": 3034
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.6638561011522965e-06,
      "loss": 0.4293,
      "step": 3035
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.657571888888538e-06,
      "loss": 0.4206,
      "step": 3036
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.651289789313929e-06,
      "loss": 0.4346,
      "step": 3037
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.64500980548485e-06,
      "loss": 0.4263,
      "step": 3038
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.638731940456657e-06,
      "loss": 0.404,
      "step": 3039
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.63245619728367e-06,
      "loss": 0.4194,
      "step": 3040
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.626182579019182e-06,
      "loss": 0.4179,
      "step": 3041
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.619911088715444e-06,
      "loss": 0.4098,
      "step": 3042
    },
    {
      "epoch": 1.96,
      "learning_rate": 5.613641729423687e-06,
      "loss": 0.446,
      "step": 3043
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.60737450419409e-06,
      "loss": 0.4242,
      "step": 3044
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.601109416075799e-06,
      "loss": 0.4169,
      "step": 3045
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.594846468116923e-06,
      "loss": 0.4219,
      "step": 3046
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.588585663364526e-06,
      "loss": 0.42,
      "step": 3047
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.582327004864631e-06,
      "loss": 0.4072,
      "step": 3048
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.576070495662215e-06,
      "loss": 0.4185,
      "step": 3049
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.569816138801211e-06,
      "loss": 0.3945,
      "step": 3050
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.563563937324503e-06,
      "loss": 0.4391,
      "step": 3051
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.557313894273937e-06,
      "loss": 0.4448,
      "step": 3052
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.551066012690286e-06,
      "loss": 0.4188,
      "step": 3053
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.5448202956133004e-06,
      "loss": 0.4461,
      "step": 3054
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.5385767460816454e-06,
      "loss": 0.422,
      "step": 3055
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.532335367132962e-06,
      "loss": 0.4147,
      "step": 3056
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.526096161803816e-06,
      "loss": 0.4062,
      "step": 3057
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.519859133129723e-06,
      "loss": 0.4393,
      "step": 3058
    },
    {
      "epoch": 1.97,
      "learning_rate": 5.513624284145138e-06,
      "loss": 0.4077,
      "step": 3059
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.507391617883454e-06,
      "loss": 0.4261,
      "step": 3060
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.501161137377005e-06,
      "loss": 0.4122,
      "step": 3061
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.494932845657057e-06,
      "loss": 0.433,
      "step": 3062
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.4887067457538245e-06,
      "loss": 0.4312,
      "step": 3063
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.4824828406964305e-06,
      "loss": 0.4189,
      "step": 3064
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.476261133512959e-06,
      "loss": 0.4231,
      "step": 3065
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.470041627230398e-06,
      "loss": 0.4209,
      "step": 3066
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.463824324874687e-06,
      "loss": 0.4312,
      "step": 3067
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.457609229470681e-06,
      "loss": 0.418,
      "step": 3068
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.451396344042165e-06,
      "loss": 0.4376,
      "step": 3069
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.445185671611846e-06,
      "loss": 0.425,
      "step": 3070
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.438977215201351e-06,
      "loss": 0.3752,
      "step": 3071
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.4327709778312484e-06,
      "loss": 0.4104,
      "step": 3072
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.426566962520996e-06,
      "loss": 0.4221,
      "step": 3073
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.420365172289003e-06,
      "loss": 0.4092,
      "step": 3074
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.414165610152563e-06,
      "loss": 0.4061,
      "step": 3075
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.407968279127915e-06,
      "loss": 0.4241,
      "step": 3076
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.401773182230199e-06,
      "loss": 0.4308,
      "step": 3077
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.395580322473466e-06,
      "loss": 0.4365,
      "step": 3078
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.3893897028706835e-06,
      "loss": 0.4003,
      "step": 3079
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.383201326433727e-06,
      "loss": 0.4163,
      "step": 3080
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.37701519617338e-06,
      "loss": 0.4138,
      "step": 3081
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.370831315099332e-06,
      "loss": 0.4197,
      "step": 3082
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.364649686220192e-06,
      "loss": 0.4181,
      "step": 3083
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.358470312543445e-06,
      "loss": 0.4267,
      "step": 3084
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.352293197075507e-06,
      "loss": 0.4488,
      "step": 3085
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.346118342821683e-06,
      "loss": 0.4031,
      "step": 3086
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.339945752786177e-06,
      "loss": 0.4372,
      "step": 3087
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.3337754299720925e-06,
      "loss": 0.4367,
      "step": 3088
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.32760737738143e-06,
      "loss": 0.4285,
      "step": 3089
    },
    {
      "epoch": 1.99,
      "learning_rate": 5.321441598015089e-06,
      "loss": 0.4383,
      "step": 3090
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.315278094872857e-06,
      "loss": 0.4406,
      "step": 3091
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.30911687095342e-06,
      "loss": 0.426,
      "step": 3092
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.302957929254346e-06,
      "loss": 0.4241,
      "step": 3093
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.296801272772109e-06,
      "loss": 0.4375,
      "step": 3094
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.290646904502055e-06,
      "loss": 0.4278,
      "step": 3095
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.284494827438423e-06,
      "loss": 0.4194,
      "step": 3096
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.27834504457434e-06,
      "loss": 0.4136,
      "step": 3097
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.2721975589018105e-06,
      "loss": 0.3896,
      "step": 3098
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.266052373411726e-06,
      "loss": 0.3382,
      "step": 3099
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.25990949109386e-06,
      "loss": 0.3367,
      "step": 3100
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.253768914936858e-06,
      "loss": 0.3222,
      "step": 3101
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.247630647928254e-06,
      "loss": 0.322,
      "step": 3102
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.241494693054447e-06,
      "loss": 0.312,
      "step": 3103
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.2353610533007305e-06,
      "loss": 0.3246,
      "step": 3104
    },
    {
      "epoch": 2.0,
      "learning_rate": 5.229229731651241e-06,
      "loss": 0.3385,
      "step": 3105
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.223100731089017e-06,
      "loss": 0.3109,
      "step": 3106
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.2169740545959515e-06,
      "loss": 0.3103,
      "step": 3107
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.210849705152809e-06,
      "loss": 0.3002,
      "step": 3108
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.204727685739228e-06,
      "loss": 0.3072,
      "step": 3109
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.198607999333706e-06,
      "loss": 0.3324,
      "step": 3110
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.192490648913607e-06,
      "loss": 0.2982,
      "step": 3111
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.186375637455159e-06,
      "loss": 0.3182,
      "step": 3112
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.180262967933463e-06,
      "loss": 0.3097,
      "step": 3113
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.174152643322454e-06,
      "loss": 0.298,
      "step": 3114
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.1680446665949605e-06,
      "loss": 0.3114,
      "step": 3115
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.161939040722634e-06,
      "loss": 0.3197,
      "step": 3116
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.1558357686760144e-06,
      "loss": 0.3264,
      "step": 3117
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.149734853424466e-06,
      "loss": 0.2973,
      "step": 3118
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.143636297936233e-06,
      "loss": 0.3101,
      "step": 3119
    },
    {
      "epoch": 2.01,
      "learning_rate": 5.137540105178396e-06,
      "loss": 0.2996,
      "step": 3120
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.131446278116887e-06,
      "loss": 0.3088,
      "step": 3121
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.1253548197165016e-06,
      "loss": 0.295,
      "step": 3122
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.119265732940857e-06,
      "loss": 0.3032,
      "step": 3123
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.113179020752443e-06,
      "loss": 0.3196,
      "step": 3124
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.1070946861125726e-06,
      "loss": 0.3117,
      "step": 3125
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.101012731981422e-06,
      "loss": 0.3333,
      "step": 3126
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.094933161317985e-06,
      "loss": 0.3062,
      "step": 3127
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.088855977080123e-06,
      "loss": 0.3062,
      "step": 3128
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.082781182224518e-06,
      "loss": 0.3217,
      "step": 3129
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.076708779706695e-06,
      "loss": 0.3026,
      "step": 3130
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.070638772481014e-06,
      "loss": 0.3168,
      "step": 3131
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.064571163500667e-06,
      "loss": 0.3202,
      "step": 3132
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.058505955717694e-06,
      "loss": 0.3001,
      "step": 3133
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.05244315208294e-06,
      "loss": 0.2991,
      "step": 3134
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.04638275554611e-06,
      "loss": 0.3181,
      "step": 3135
    },
    {
      "epoch": 2.02,
      "learning_rate": 5.040324769055709e-06,
      "loss": 0.3096,
      "step": 3136
    },
    {
      "epoch": 2.03,
      "learning_rate": 5.034269195559096e-06,
      "loss": 0.3149,
      "step": 3137
    },
    {
      "epoch": 2.03,
      "learning_rate": 5.028216038002441e-06,
      "loss": 0.2996,
      "step": 3138
    },
    {
      "epoch": 2.03,
      "learning_rate": 5.022165299330738e-06,
      "loss": 0.2956,
      "step": 3139
    },
    {
      "epoch": 2.03,
      "learning_rate": 5.016116982487811e-06,
      "loss": 0.3024,
      "step": 3140
    },
    {
      "epoch": 2.03,
      "learning_rate": 5.010071090416301e-06,
      "loss": 0.3107,
      "step": 3141
    },
    {
      "epoch": 2.03,
      "learning_rate": 5.004027626057672e-06,
      "loss": 0.3008,
      "step": 3142
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.9979865923522e-06,
      "loss": 0.317,
      "step": 3143
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.991947992238997e-06,
      "loss": 0.3198,
      "step": 3144
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.985911828655962e-06,
      "loss": 0.3035,
      "step": 3145
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.979878104539836e-06,
      "loss": 0.3157,
      "step": 3146
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.973846822826158e-06,
      "loss": 0.3082,
      "step": 3147
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.967817986449284e-06,
      "loss": 0.3098,
      "step": 3148
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.961791598342377e-06,
      "loss": 0.3035,
      "step": 3149
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.955767661437412e-06,
      "loss": 0.2925,
      "step": 3150
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.949746178665168e-06,
      "loss": 0.3033,
      "step": 3151
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.943727152955235e-06,
      "loss": 0.3196,
      "step": 3152
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.937710587236001e-06,
      "loss": 0.3151,
      "step": 3153
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.9316964844346635e-06,
      "loss": 0.3188,
      "step": 3154
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.925684847477216e-06,
      "loss": 0.313,
      "step": 3155
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.9196756792884605e-06,
      "loss": 0.3094,
      "step": 3156
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.913668982791989e-06,
      "loss": 0.2767,
      "step": 3157
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.907664760910197e-06,
      "loss": 0.2843,
      "step": 3158
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.901663016564273e-06,
      "loss": 0.2943,
      "step": 3159
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.8956637526742e-06,
      "loss": 0.3048,
      "step": 3160
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.889666972158755e-06,
      "loss": 0.3107,
      "step": 3161
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.883672677935508e-06,
      "loss": 0.2947,
      "step": 3162
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.877680872920818e-06,
      "loss": 0.3113,
      "step": 3163
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.871691560029828e-06,
      "loss": 0.3039,
      "step": 3164
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.865704742176487e-06,
      "loss": 0.3293,
      "step": 3165
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.859720422273501e-06,
      "loss": 0.3066,
      "step": 3166
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.8537386032323904e-06,
      "loss": 0.3013,
      "step": 3167
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.847759287963432e-06,
      "loss": 0.2861,
      "step": 3168
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.841782479375707e-06,
      "loss": 0.3088,
      "step": 3169
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.835808180377065e-06,
      "loss": 0.3134,
      "step": 3170
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.8298363938741356e-06,
      "loss": 0.3235,
      "step": 3171
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.8238671227723285e-06,
      "loss": 0.2917,
      "step": 3172
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.817900369975826e-06,
      "loss": 0.2976,
      "step": 3173
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.811936138387597e-06,
      "loss": 0.3102,
      "step": 3174
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.805974430909361e-06,
      "loss": 0.3161,
      "step": 3175
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.800015250441638e-06,
      "loss": 0.3233,
      "step": 3176
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.794058599883687e-06,
      "loss": 0.3019,
      "step": 3177
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.78810448213357e-06,
      "loss": 0.3086,
      "step": 3178
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.7821529000880825e-06,
      "loss": 0.3018,
      "step": 3179
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.7762038566428155e-06,
      "loss": 0.3094,
      "step": 3180
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.77025735469211e-06,
      "loss": 0.3098,
      "step": 3181
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.764313397129066e-06,
      "loss": 0.2966,
      "step": 3182
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.758371986845569e-06,
      "loss": 0.3298,
      "step": 3183
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.752433126732231e-06,
      "loss": 0.3146,
      "step": 3184
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.746496819678455e-06,
      "loss": 0.3002,
      "step": 3185
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.740563068572377e-06,
      "loss": 0.2893,
      "step": 3186
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.734631876300914e-06,
      "loss": 0.3073,
      "step": 3187
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.728703245749707e-06,
      "loss": 0.3075,
      "step": 3188
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.722777179803184e-06,
      "loss": 0.2905,
      "step": 3189
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.716853681344502e-06,
      "loss": 0.3105,
      "step": 3190
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.71093275325558e-06,
      "loss": 0.2952,
      "step": 3191
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.7050143984170805e-06,
      "loss": 0.2964,
      "step": 3192
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.699098619708414e-06,
      "loss": 0.3135,
      "step": 3193
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.693185420007751e-06,
      "loss": 0.313,
      "step": 3194
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.6872748021919816e-06,
      "loss": 0.3113,
      "step": 3195
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.681366769136769e-06,
      "loss": 0.2975,
      "step": 3196
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.675461323716491e-06,
      "loss": 0.3121,
      "step": 3197
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.669558468804288e-06,
      "loss": 0.3049,
      "step": 3198
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.663658207272032e-06,
      "loss": 0.3101,
      "step": 3199
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.657760541990329e-06,
      "loss": 0.3143,
      "step": 3200
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.651865475828531e-06,
      "loss": 0.2975,
      "step": 3201
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.645973011654717e-06,
      "loss": 0.3082,
      "step": 3202
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.640083152335706e-06,
      "loss": 0.3142,
      "step": 3203
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.634195900737045e-06,
      "loss": 0.3028,
      "step": 3204
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.628311259723015e-06,
      "loss": 0.3166,
      "step": 3205
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.622429232156624e-06,
      "loss": 0.2941,
      "step": 3206
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.616549820899618e-06,
      "loss": 0.3249,
      "step": 3207
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.610673028812459e-06,
      "loss": 0.2981,
      "step": 3208
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.604798858754339e-06,
      "loss": 0.2994,
      "step": 3209
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.598927313583175e-06,
      "loss": 0.3334,
      "step": 3210
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.593058396155604e-06,
      "loss": 0.3142,
      "step": 3211
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.587192109326988e-06,
      "loss": 0.3083,
      "step": 3212
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.581328455951408e-06,
      "loss": 0.311,
      "step": 3213
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.5754674388816625e-06,
      "loss": 0.3133,
      "step": 3214
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.569609060969269e-06,
      "loss": 0.2897,
      "step": 3215
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.563753325064457e-06,
      "loss": 0.3083,
      "step": 3216
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.557900234016185e-06,
      "loss": 0.2928,
      "step": 3217
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.552049790672099e-06,
      "loss": 0.3097,
      "step": 3218
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.546201997878581e-06,
      "loss": 0.28,
      "step": 3219
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.540356858480711e-06,
      "loss": 0.2992,
      "step": 3220
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.534514375322283e-06,
      "loss": 0.2972,
      "step": 3221
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.528674551245796e-06,
      "loss": 0.3088,
      "step": 3222
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.522837389092455e-06,
      "loss": 0.3168,
      "step": 3223
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.5170028917021705e-06,
      "loss": 0.3097,
      "step": 3224
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.511171061913555e-06,
      "loss": 0.315,
      "step": 3225
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.5053419025639355e-06,
      "loss": 0.3077,
      "step": 3226
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.499515416489314e-06,
      "loss": 0.3055,
      "step": 3227
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.493691606524423e-06,
      "loss": 0.3139,
      "step": 3228
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.487870475502661e-06,
      "loss": 0.3242,
      "step": 3229
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.482052026256156e-06,
      "loss": 0.3135,
      "step": 3230
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.4762362616157e-06,
      "loss": 0.3012,
      "step": 3231
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.470423184410804e-06,
      "loss": 0.3068,
      "step": 3232
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.464612797469656e-06,
      "loss": 0.2995,
      "step": 3233
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.4588051036191395e-06,
      "loss": 0.3132,
      "step": 3234
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.453000105684836e-06,
      "loss": 0.3023,
      "step": 3235
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.447197806490996e-06,
      "loss": 0.3096,
      "step": 3236
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.441398208860582e-06,
      "loss": 0.3023,
      "step": 3237
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.4356013156152154e-06,
      "loss": 0.2979,
      "step": 3238
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.429807129575229e-06,
      "loss": 0.3027,
      "step": 3239
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.424015653559611e-06,
      "loss": 0.2952,
      "step": 3240
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.418226890386056e-06,
      "loss": 0.3131,
      "step": 3241
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.4124408428709235e-06,
      "loss": 0.2984,
      "step": 3242
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.406657513829251e-06,
      "loss": 0.3005,
      "step": 3243
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.400876906074772e-06,
      "loss": 0.3142,
      "step": 3244
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.395099022419869e-06,
      "loss": 0.3072,
      "step": 3245
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.389323865675623e-06,
      "loss": 0.3035,
      "step": 3246
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.383551438651765e-06,
      "loss": 0.298,
      "step": 3247
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.377781744156727e-06,
      "loss": 0.3037,
      "step": 3248
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.372014784997579e-06,
      "loss": 0.3054,
      "step": 3249
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.366250563980089e-06,
      "loss": 0.3212,
      "step": 3250
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.360489083908675e-06,
      "loss": 0.3052,
      "step": 3251
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.35473034758643e-06,
      "loss": 0.3002,
      "step": 3252
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.348974357815106e-06,
      "loss": 0.2952,
      "step": 3253
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.343221117395122e-06,
      "loss": 0.3195,
      "step": 3254
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.3374706291255596e-06,
      "loss": 0.3109,
      "step": 3255
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.331722895804158e-06,
      "loss": 0.3158,
      "step": 3256
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.32597792022733e-06,
      "loss": 0.302,
      "step": 3257
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.32023570519012e-06,
      "loss": 0.3036,
      "step": 3258
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.314496253486258e-06,
      "loss": 0.3252,
      "step": 3259
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.30875956790811e-06,
      "loss": 0.3049,
      "step": 3260
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.303025651246705e-06,
      "loss": 0.2962,
      "step": 3261
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.297294506291721e-06,
      "loss": 0.3244,
      "step": 3262
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.291566135831492e-06,
      "loss": 0.326,
      "step": 3263
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.285840542652997e-06,
      "loss": 0.3026,
      "step": 3264
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.280117729541866e-06,
      "loss": 0.309,
      "step": 3265
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.274397699282377e-06,
      "loss": 0.3061,
      "step": 3266
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.268680454657456e-06,
      "loss": 0.3085,
      "step": 3267
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.262965998448665e-06,
      "loss": 0.3151,
      "step": 3268
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.257254333436223e-06,
      "loss": 0.3029,
      "step": 3269
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.25154546239898e-06,
      "loss": 0.32,
      "step": 3270
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.2458393881144335e-06,
      "loss": 0.2896,
      "step": 3271
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.240136113358714e-06,
      "loss": 0.3067,
      "step": 3272
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.234435640906593e-06,
      "loss": 0.3204,
      "step": 3273
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.228737973531481e-06,
      "loss": 0.3132,
      "step": 3274
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.2230431140054205e-06,
      "loss": 0.3137,
      "step": 3275
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.2173510650990905e-06,
      "loss": 0.3101,
      "step": 3276
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.211661829581793e-06,
      "loss": 0.3078,
      "step": 3277
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.205975410221486e-06,
      "loss": 0.3098,
      "step": 3278
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.200291809784722e-06,
      "loss": 0.2958,
      "step": 3279
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.194611031036718e-06,
      "loss": 0.3107,
      "step": 3280
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.188933076741284e-06,
      "loss": 0.3059,
      "step": 3281
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.183257949660885e-06,
      "loss": 0.3037,
      "step": 3282
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.177585652556595e-06,
      "loss": 0.2777,
      "step": 3283
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.171916188188113e-06,
      "loss": 0.2985,
      "step": 3284
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.166249559313762e-06,
      "loss": 0.2959,
      "step": 3285
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.1605857686904806e-06,
      "loss": 0.3098,
      "step": 3286
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.154924819073842e-06,
      "loss": 0.3082,
      "step": 3287
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.149266713218011e-06,
      "loss": 0.2915,
      "step": 3288
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.143611453875798e-06,
      "loss": 0.2906,
      "step": 3289
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.1379590437986e-06,
      "loss": 0.3202,
      "step": 3290
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.132309485736457e-06,
      "loss": 0.3019,
      "step": 3291
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.12666278243799e-06,
      "loss": 0.3049,
      "step": 3292
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.1210189366504596e-06,
      "loss": 0.3046,
      "step": 3293
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.11537795111972e-06,
      "loss": 0.2878,
      "step": 3294
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.1097398285902344e-06,
      "loss": 0.2948,
      "step": 3295
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.104104571805088e-06,
      "loss": 0.2851,
      "step": 3296
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.098472183505944e-06,
      "loss": 0.3115,
      "step": 3297
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.092842666433101e-06,
      "loss": 0.2991,
      "step": 3298
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.087216023325433e-06,
      "loss": 0.3081,
      "step": 3299
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.0815922569204435e-06,
      "loss": 0.3274,
      "step": 3300
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.075971369954205e-06,
      "loss": 0.3122,
      "step": 3301
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.070353365161417e-06,
      "loss": 0.3022,
      "step": 3302
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.064738245275363e-06,
      "loss": 0.2886,
      "step": 3303
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.059126013027924e-06,
      "loss": 0.3281,
      "step": 3304
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.053516671149579e-06,
      "loss": 0.3135,
      "step": 3305
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.047910222369394e-06,
      "loss": 0.3061,
      "step": 3306
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.042306669415046e-06,
      "loss": 0.2957,
      "step": 3307
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.0367060150127726e-06,
      "loss": 0.3186,
      "step": 3308
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.031108261887434e-06,
      "loss": 0.3239,
      "step": 3309
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.02551341276245e-06,
      "loss": 0.305,
      "step": 3310
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.019921470359849e-06,
      "loss": 0.3093,
      "step": 3311
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.014332437400235e-06,
      "loss": 0.2962,
      "step": 3312
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.008746316602799e-06,
      "loss": 0.3147,
      "step": 3313
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.0031631106853135e-06,
      "loss": 0.2879,
      "step": 3314
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.997582822364135e-06,
      "loss": 0.3056,
      "step": 3315
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.9920054543541976e-06,
      "loss": 0.2967,
      "step": 3316
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.98643100936902e-06,
      "loss": 0.2892,
      "step": 3317
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.980859490120692e-06,
      "loss": 0.3002,
      "step": 3318
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.975290899319881e-06,
      "loss": 0.2962,
      "step": 3319
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.96972523967584e-06,
      "loss": 0.3005,
      "step": 3320
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.964162513896383e-06,
      "loss": 0.2939,
      "step": 3321
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.958602724687901e-06,
      "loss": 0.3039,
      "step": 3322
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.953045874755359e-06,
      "loss": 0.3126,
      "step": 3323
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.947491966802288e-06,
      "loss": 0.312,
      "step": 3324
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.941941003530789e-06,
      "loss": 0.3044,
      "step": 3325
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.936392987641533e-06,
      "loss": 0.3039,
      "step": 3326
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.9308479218337525e-06,
      "loss": 0.3048,
      "step": 3327
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.925305808805247e-06,
      "loss": 0.2925,
      "step": 3328
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.919766651252378e-06,
      "loss": 0.2967,
      "step": 3329
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.914230451870079e-06,
      "loss": 0.3042,
      "step": 3330
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.908697213351823e-06,
      "loss": 0.2893,
      "step": 3331
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.903166938389664e-06,
      "loss": 0.2966,
      "step": 3332
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.897639629674205e-06,
      "loss": 0.3085,
      "step": 3333
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.892115289894603e-06,
      "loss": 0.2912,
      "step": 3334
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.886593921738575e-06,
      "loss": 0.3077,
      "step": 3335
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.881075527892391e-06,
      "loss": 0.293,
      "step": 3336
    },
    {
      "epoch": 2.15,
      "learning_rate": 3.875560111040876e-06,
      "loss": 0.3177,
      "step": 3337
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.870047673867397e-06,
      "loss": 0.2937,
      "step": 3338
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.864538219053894e-06,
      "loss": 0.3103,
      "step": 3339
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.859031749280824e-06,
      "loss": 0.309,
      "step": 3340
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.853528267227223e-06,
      "loss": 0.3174,
      "step": 3341
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.848027775570645e-06,
      "loss": 0.3352,
      "step": 3342
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.842530276987215e-06,
      "loss": 0.3029,
      "step": 3343
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.837035774151585e-06,
      "loss": 0.3096,
      "step": 3344
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.831544269736955e-06,
      "loss": 0.3197,
      "step": 3345
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.826055766415067e-06,
      "loss": 0.2932,
      "step": 3346
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.820570266856195e-06,
      "loss": 0.2999,
      "step": 3347
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.815087773729171e-06,
      "loss": 0.3052,
      "step": 3348
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.8096082897013376e-06,
      "loss": 0.3144,
      "step": 3349
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.804131817438599e-06,
      "loss": 0.317,
      "step": 3350
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.7986583596053715e-06,
      "loss": 0.2933,
      "step": 3351
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.793187918864627e-06,
      "loss": 0.3197,
      "step": 3352
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.7877204978778458e-06,
      "loss": 0.2953,
      "step": 3353
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.7822560993050617e-06,
      "loss": 0.3049,
      "step": 3354
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.7767947258048243e-06,
      "loss": 0.3156,
      "step": 3355
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.771336380034211e-06,
      "loss": 0.2956,
      "step": 3356
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.765881064648841e-06,
      "loss": 0.2957,
      "step": 3357
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.760428782302833e-06,
      "loss": 0.318,
      "step": 3358
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.7549795356488607e-06,
      "loss": 0.314,
      "step": 3359
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.749533327338091e-06,
      "loss": 0.2998,
      "step": 3360
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.7440901600202404e-06,
      "loss": 0.3021,
      "step": 3361
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.7386500363435196e-06,
      "loss": 0.3048,
      "step": 3362
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.733212958954682e-06,
      "loss": 0.3054,
      "step": 3363
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.727778930498982e-06,
      "loss": 0.2865,
      "step": 3364
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.7223479536202e-06,
      "loss": 0.3178,
      "step": 3365
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.716920030960628e-06,
      "loss": 0.3044,
      "step": 3366
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.711495165161072e-06,
      "loss": 0.3184,
      "step": 3367
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.706073358860851e-06,
      "loss": 0.3101,
      "step": 3368
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.700654614697794e-06,
      "loss": 0.2933,
      "step": 3369
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.695238935308252e-06,
      "loss": 0.2898,
      "step": 3370
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6898263233270603e-06,
      "loss": 0.29,
      "step": 3371
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6844167813875888e-06,
      "loss": 0.3156,
      "step": 3372
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.679010312121697e-06,
      "loss": 0.3068,
      "step": 3373
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.673606918159753e-06,
      "loss": 0.3153,
      "step": 3374
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.668206602130632e-06,
      "loss": 0.3026,
      "step": 3375
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6628093666617083e-06,
      "loss": 0.3119,
      "step": 3376
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6574152143788578e-06,
      "loss": 0.3075,
      "step": 3377
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.652024147906459e-06,
      "loss": 0.3134,
      "step": 3378
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6466361698673856e-06,
      "loss": 0.3028,
      "step": 3379
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6412512828830114e-06,
      "loss": 0.3087,
      "step": 3380
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6358694895732016e-06,
      "loss": 0.3107,
      "step": 3381
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6304907925563273e-06,
      "loss": 0.3131,
      "step": 3382
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.62511519444924e-06,
      "loss": 0.3241,
      "step": 3383
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.6197426978672923e-06,
      "loss": 0.3032,
      "step": 3384
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.6143733054243234e-06,
      "loss": 0.3276,
      "step": 3385
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.6090070197326643e-06,
      "loss": 0.3088,
      "step": 3386
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.603643843403134e-06,
      "loss": 0.3097,
      "step": 3387
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.5982837790450376e-06,
      "loss": 0.288,
      "step": 3388
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.592926829266168e-06,
      "loss": 0.3066,
      "step": 3389
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.587572996672799e-06,
      "loss": 0.2931,
      "step": 3390
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.5822222838697006e-06,
      "loss": 0.2879,
      "step": 3391
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.5768746934601007e-06,
      "loss": 0.2982,
      "step": 3392
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.5715302280457343e-06,
      "loss": 0.2998,
      "step": 3393
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.5661888902268007e-06,
      "loss": 0.3102,
      "step": 3394
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.5608506826019797e-06,
      "loss": 0.3077,
      "step": 3395
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.5555156077684317e-06,
      "loss": 0.3046,
      "step": 3396
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.550183668321788e-06,
      "loss": 0.2936,
      "step": 3397
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.54485486685616e-06,
      "loss": 0.3141,
      "step": 3398
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.539529205964126e-06,
      "loss": 0.2999,
      "step": 3399
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.5342066882367488e-06,
      "loss": 0.3179,
      "step": 3400
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.5288873162635407e-06,
      "loss": 0.2865,
      "step": 3401
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.5235710926325093e-06,
      "loss": 0.3005,
      "step": 3402
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.518258019930103e-06,
      "loss": 0.2912,
      "step": 3403
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.5129481007412634e-06,
      "loss": 0.3005,
      "step": 3404
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.507641337649381e-06,
      "loss": 0.3094,
      "step": 3405
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.502337733236316e-06,
      "loss": 0.3104,
      "step": 3406
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.4970372900823914e-06,
      "loss": 0.302,
      "step": 3407
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.491740010766389e-06,
      "loss": 0.3079,
      "step": 3408
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.4864458978655657e-06,
      "loss": 0.3073,
      "step": 3409
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.481154953955611e-06,
      "loss": 0.2889,
      "step": 3410
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.475867181610704e-06,
      "loss": 0.2985,
      "step": 3411
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.47058258340345e-06,
      "loss": 0.3165,
      "step": 3412
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.465301161904936e-06,
      "loss": 0.3061,
      "step": 3413
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.4600229196846903e-06,
      "loss": 0.2999,
      "step": 3414
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.454747859310694e-06,
      "loss": 0.3016,
      "step": 3415
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.449475983349385e-06,
      "loss": 0.3379,
      "step": 3416
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.44420729436565e-06,
      "loss": 0.2948,
      "step": 3417
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.438941794922824e-06,
      "loss": 0.31,
      "step": 3418
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.4336794875826884e-06,
      "loss": 0.3119,
      "step": 3419
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.4284203749054834e-06,
      "loss": 0.2931,
      "step": 3420
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.4231644594498724e-06,
      "loss": 0.299,
      "step": 3421
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.4179117437729915e-06,
      "loss": 0.3192,
      "step": 3422
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.4126622304303914e-06,
      "loss": 0.3061,
      "step": 3423
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.4074159219760884e-06,
      "loss": 0.3013,
      "step": 3424
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.402172820962526e-06,
      "loss": 0.3067,
      "step": 3425
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.396932929940592e-06,
      "loss": 0.3222,
      "step": 3426
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.391696251459611e-06,
      "loss": 0.3149,
      "step": 3427
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.386462788067344e-06,
      "loss": 0.3002,
      "step": 3428
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.381232542309991e-06,
      "loss": 0.2935,
      "step": 3429
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.376005516732184e-06,
      "loss": 0.3162,
      "step": 3430
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.370781713876987e-06,
      "loss": 0.2967,
      "step": 3431
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3655611362858966e-06,
      "loss": 0.2712,
      "step": 3432
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3603437864988466e-06,
      "loss": 0.3112,
      "step": 3433
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3551296670541923e-06,
      "loss": 0.2902,
      "step": 3434
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3499187804887213e-06,
      "loss": 0.3089,
      "step": 3435
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3447111293376468e-06,
      "loss": 0.3202,
      "step": 3436
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3395067161346073e-06,
      "loss": 0.3122,
      "step": 3437
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.334305543411669e-06,
      "loss": 0.3105,
      "step": 3438
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3291076136993185e-06,
      "loss": 0.3028,
      "step": 3439
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.323912929526465e-06,
      "loss": 0.3003,
      "step": 3440
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3187214934204413e-06,
      "loss": 0.3096,
      "step": 3441
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3135333079069943e-06,
      "loss": 0.32,
      "step": 3442
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.3083483755102996e-06,
      "loss": 0.3104,
      "step": 3443
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.30316669875294e-06,
      "loss": 0.3202,
      "step": 3444
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.297988280155917e-06,
      "loss": 0.3141,
      "step": 3445
    },
    {
      "epoch": 2.22,
      "learning_rate": 3.2928131222386507e-06,
      "loss": 0.3064,
      "step": 3446
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2876412275189697e-06,
      "loss": 0.309,
      "step": 3447
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.282472598513119e-06,
      "loss": 0.2836,
      "step": 3448
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2773072377357517e-06,
      "loss": 0.3046,
      "step": 3449
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2721451476999333e-06,
      "loss": 0.2973,
      "step": 3450
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2669863309171325e-06,
      "loss": 0.2904,
      "step": 3451
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2618307898972413e-06,
      "loss": 0.3095,
      "step": 3452
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2566785271485313e-06,
      "loss": 0.2919,
      "step": 3453
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2515295451777105e-06,
      "loss": 0.2983,
      "step": 3454
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.246383846489859e-06,
      "loss": 0.323,
      "step": 3455
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2412414335884866e-06,
      "loss": 0.2961,
      "step": 3456
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2361023089754894e-06,
      "loss": 0.3035,
      "step": 3457
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.230966475151166e-06,
      "loss": 0.2977,
      "step": 3458
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2258339346142176e-06,
      "loss": 0.3285,
      "step": 3459
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.2207046898617365e-06,
      "loss": 0.3137,
      "step": 3460
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.215578743389225e-06,
      "loss": 0.3075,
      "step": 3461
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.2104560976905596e-06,
      "loss": 0.2913,
      "step": 3462
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.205336755258035e-06,
      "loss": 0.3135,
      "step": 3463
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.2002207185823155e-06,
      "loss": 0.3073,
      "step": 3464
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.195107990152475e-06,
      "loss": 0.3028,
      "step": 3465
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.18999857245597e-06,
      "loss": 0.3049,
      "step": 3466
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.1848924679786476e-06,
      "loss": 0.318,
      "step": 3467
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.1797896792047422e-06,
      "loss": 0.3012,
      "step": 3468
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.1746902086168707e-06,
      "loss": 0.3027,
      "step": 3469
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.1695940586960527e-06,
      "loss": 0.2858,
      "step": 3470
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.164501231921666e-06,
      "loss": 0.3154,
      "step": 3471
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.1594117307714977e-06,
      "loss": 0.302,
      "step": 3472
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.1543255577216935e-06,
      "loss": 0.3045,
      "step": 3473
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.1492427152468e-06,
      "loss": 0.2917,
      "step": 3474
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.1441632058197315e-06,
      "loss": 0.2885,
      "step": 3475
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.1390870319117838e-06,
      "loss": 0.2912,
      "step": 3476
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.1340141959926298e-06,
      "loss": 0.3217,
      "step": 3477
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.1289447005303186e-06,
      "loss": 0.2964,
      "step": 3478
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.1238785479912747e-06,
      "loss": 0.3229,
      "step": 3479
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.118815740840294e-06,
      "loss": 0.2917,
      "step": 3480
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.1137562815405475e-06,
      "loss": 0.3017,
      "step": 3481
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.1087001725535726e-06,
      "loss": 0.3275,
      "step": 3482
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.103647416339287e-06,
      "loss": 0.3173,
      "step": 3483
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.098598015355967e-06,
      "loss": 0.3116,
      "step": 3484
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.093551972060259e-06,
      "loss": 0.2971,
      "step": 3485
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.088509288907179e-06,
      "loss": 0.3119,
      "step": 3486
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.0834699683501047e-06,
      "loss": 0.3091,
      "step": 3487
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.0784340128407786e-06,
      "loss": 0.2985,
      "step": 3488
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.0734014248293077e-06,
      "loss": 0.2977,
      "step": 3489
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.0683722067641587e-06,
      "loss": 0.3262,
      "step": 3490
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.0633463610921588e-06,
      "loss": 0.3155,
      "step": 3491
    },
    {
      "epoch": 2.25,
      "learning_rate": 3.0583238902584976e-06,
      "loss": 0.3191,
      "step": 3492
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0533047967067153e-06,
      "loss": 0.3024,
      "step": 3493
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0482890828787205e-06,
      "loss": 0.2974,
      "step": 3494
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.043276751214769e-06,
      "loss": 0.3092,
      "step": 3495
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.038267804153472e-06,
      "loss": 0.3008,
      "step": 3496
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.033262244131795e-06,
      "loss": 0.3103,
      "step": 3497
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0282600735850564e-06,
      "loss": 0.3019,
      "step": 3498
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.023261294946925e-06,
      "loss": 0.2985,
      "step": 3499
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0182659106494195e-06,
      "loss": 0.2988,
      "step": 3500
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0132739231229057e-06,
      "loss": 0.2975,
      "step": 3501
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0082853347960983e-06,
      "loss": 0.3029,
      "step": 3502
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.0033001480960567e-06,
      "loss": 0.3203,
      "step": 3503
    },
    {
      "epoch": 2.26,
      "learning_rate": 2.998318365448194e-06,
      "loss": 0.3159,
      "step": 3504
    },
    {
      "epoch": 2.26,
      "learning_rate": 2.9933399892762483e-06,
      "loss": 0.3168,
      "step": 3505
    },
    {
      "epoch": 2.26,
      "learning_rate": 2.9883650220023196e-06,
      "loss": 0.3139,
      "step": 3506
    },
    {
      "epoch": 2.26,
      "learning_rate": 2.9833934660468412e-06,
      "loss": 0.3018,
      "step": 3507
    },
    {
      "epoch": 2.26,
      "learning_rate": 2.9784253238285844e-06,
      "loss": 0.2865,
      "step": 3508
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9734605977646638e-06,
      "loss": 0.309,
      "step": 3509
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9684992902705303e-06,
      "loss": 0.2851,
      "step": 3510
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.963541403759972e-06,
      "loss": 0.3121,
      "step": 3511
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9585869406451083e-06,
      "loss": 0.3022,
      "step": 3512
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9536359033364083e-06,
      "loss": 0.3113,
      "step": 3513
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9486882942426486e-06,
      "loss": 0.3131,
      "step": 3514
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9437441157709668e-06,
      "loss": 0.3085,
      "step": 3515
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.938803370326804e-06,
      "loss": 0.304,
      "step": 3516
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9338660603139533e-06,
      "loss": 0.3141,
      "step": 3517
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9289321881345257e-06,
      "loss": 0.3231,
      "step": 3518
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.924001756188959e-06,
      "loss": 0.3034,
      "step": 3519
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9190747668760213e-06,
      "loss": 0.2911,
      "step": 3520
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9141512225927993e-06,
      "loss": 0.301,
      "step": 3521
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9092311257347182e-06,
      "loss": 0.3219,
      "step": 3522
    },
    {
      "epoch": 2.27,
      "learning_rate": 2.9043144786955048e-06,
      "loss": 0.3056,
      "step": 3523
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.899401283867229e-06,
      "loss": 0.3245,
      "step": 3524
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.89449154364026e-06,
      "loss": 0.3058,
      "step": 3525
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.8895852604033047e-06,
      "loss": 0.2909,
      "step": 3526
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.88468243654338e-06,
      "loss": 0.286,
      "step": 3527
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.8797830744458177e-06,
      "loss": 0.3009,
      "step": 3528
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.87488717649427e-06,
      "loss": 0.3004,
      "step": 3529
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.869994745070699e-06,
      "loss": 0.3169,
      "step": 3530
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.865105782555385e-06,
      "loss": 0.2829,
      "step": 3531
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.860220291326915e-06,
      "loss": 0.2981,
      "step": 3532
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.8553382737621995e-06,
      "loss": 0.3047,
      "step": 3533
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.850459732236438e-06,
      "loss": 0.3002,
      "step": 3534
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.845584669123159e-06,
      "loss": 0.3134,
      "step": 3535
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.840713086794189e-06,
      "loss": 0.2833,
      "step": 3536
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.8358449876196625e-06,
      "loss": 0.3033,
      "step": 3537
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.8309803739680184e-06,
      "loss": 0.3051,
      "step": 3538
    },
    {
      "epoch": 2.28,
      "learning_rate": 2.8261192482060017e-06,
      "loss": 0.314,
      "step": 3539
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.8212616126986604e-06,
      "loss": 0.3079,
      "step": 3540
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.8164074698093414e-06,
      "loss": 0.302,
      "step": 3541
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.811556821899696e-06,
      "loss": 0.2906,
      "step": 3542
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.8067096713296747e-06,
      "loss": 0.3033,
      "step": 3543
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.801866020457521e-06,
      "loss": 0.2993,
      "step": 3544
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.7970258716397856e-06,
      "loss": 0.3019,
      "step": 3545
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.7921892272313087e-06,
      "loss": 0.2959,
      "step": 3546
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.787356089585227e-06,
      "loss": 0.3064,
      "step": 3547
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.7825264610529703e-06,
      "loss": 0.3255,
      "step": 3548
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.777700343984261e-06,
      "loss": 0.3053,
      "step": 3549
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.7728777407271133e-06,
      "loss": 0.3094,
      "step": 3550
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.768058653627834e-06,
      "loss": 0.3215,
      "step": 3551
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.7632430850310175e-06,
      "loss": 0.3044,
      "step": 3552
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.7584310372795462e-06,
      "loss": 0.3101,
      "step": 3553
    },
    {
      "epoch": 2.29,
      "learning_rate": 2.7536225127145897e-06,
      "loss": 0.3249,
      "step": 3554
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7488175136756e-06,
      "loss": 0.3059,
      "step": 3555
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7440160425003236e-06,
      "loss": 0.3059,
      "step": 3556
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.739218101524782e-06,
      "loss": 0.3126,
      "step": 3557
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7344236930832814e-06,
      "loss": 0.3044,
      "step": 3558
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.729632819508409e-06,
      "loss": 0.3049,
      "step": 3559
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7248454831310335e-06,
      "loss": 0.3244,
      "step": 3560
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.720061686280302e-06,
      "loss": 0.2997,
      "step": 3561
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7152814312836397e-06,
      "loss": 0.3023,
      "step": 3562
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.710504720466747e-06,
      "loss": 0.3002,
      "step": 3563
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7057315561536e-06,
      "loss": 0.2978,
      "step": 3564
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.7009619406664587e-06,
      "loss": 0.2905,
      "step": 3565
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.6961958763258366e-06,
      "loss": 0.2964,
      "step": 3566
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.6914333654505444e-06,
      "loss": 0.3022,
      "step": 3567
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.686674410357637e-06,
      "loss": 0.3042,
      "step": 3568
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.681919013362464e-06,
      "loss": 0.2877,
      "step": 3569
    },
    {
      "epoch": 2.3,
      "learning_rate": 2.67716717677863e-06,
      "loss": 0.2892,
      "step": 3570
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6724189029180103e-06,
      "loss": 0.3109,
      "step": 3571
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.667674194090748e-06,
      "loss": 0.3116,
      "step": 3572
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6629330526052465e-06,
      "loss": 0.2959,
      "step": 3573
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6581954807681877e-06,
      "loss": 0.2887,
      "step": 3574
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.653461480884494e-06,
      "loss": 0.2944,
      "step": 3575
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6487310552573776e-06,
      "loss": 0.3099,
      "step": 3576
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.644004206188282e-06,
      "loss": 0.2974,
      "step": 3577
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.639280935976938e-06,
      "loss": 0.3256,
      "step": 3578
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.634561246921317e-06,
      "loss": 0.2867,
      "step": 3579
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6298451413176564e-06,
      "loss": 0.2859,
      "step": 3580
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.625132621460447e-06,
      "loss": 0.2907,
      "step": 3581
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.620423689642432e-06,
      "loss": 0.3027,
      "step": 3582
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.615718348154623e-06,
      "loss": 0.3079,
      "step": 3583
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6110165992862635e-06,
      "loss": 0.3114,
      "step": 3584
    },
    {
      "epoch": 2.31,
      "learning_rate": 2.6063184453248704e-06,
      "loss": 0.2922,
      "step": 3585
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.6016238885561907e-06,
      "loss": 0.2917,
      "step": 3586
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.596932931264241e-06,
      "loss": 0.3341,
      "step": 3587
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.592245575731274e-06,
      "loss": 0.3116,
      "step": 3588
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5875618242377944e-06,
      "loss": 0.3036,
      "step": 3589
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.582881679062552e-06,
      "loss": 0.2907,
      "step": 3590
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.578205142482545e-06,
      "loss": 0.2815,
      "step": 3591
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5735322167730116e-06,
      "loss": 0.3031,
      "step": 3592
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5688629042074377e-06,
      "loss": 0.3024,
      "step": 3593
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5641972070575462e-06,
      "loss": 0.3043,
      "step": 3594
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5595351275933032e-06,
      "loss": 0.2905,
      "step": 3595
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5548766680829207e-06,
      "loss": 0.2833,
      "step": 3596
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.550221830792842e-06,
      "loss": 0.3166,
      "step": 3597
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5455706179877483e-06,
      "loss": 0.2884,
      "step": 3598
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5409230319305613e-06,
      "loss": 0.3072,
      "step": 3599
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5362790748824363e-06,
      "loss": 0.294,
      "step": 3600
    },
    {
      "epoch": 2.32,
      "learning_rate": 2.5316387491027615e-06,
      "loss": 0.2915,
      "step": 3601
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.5270020568491628e-06,
      "loss": 0.3008,
      "step": 3602
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.5223690003774925e-06,
      "loss": 0.302,
      "step": 3603
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.517739581941839e-06,
      "loss": 0.2956,
      "step": 3604
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.5131138037945147e-06,
      "loss": 0.2962,
      "step": 3605
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.5084916681860716e-06,
      "loss": 0.3132,
      "step": 3606
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.503873177365279e-06,
      "loss": 0.3201,
      "step": 3607
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.4992583335791375e-06,
      "loss": 0.3106,
      "step": 3608
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.4946471390728734e-06,
      "loss": 0.301,
      "step": 3609
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.490039596089935e-06,
      "loss": 0.3005,
      "step": 3610
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.485435706871996e-06,
      "loss": 0.287,
      "step": 3611
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.4808354736589525e-06,
      "loss": 0.3239,
      "step": 3612
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.4762388986889228e-06,
      "loss": 0.3004,
      "step": 3613
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.471645984198241e-06,
      "loss": 0.2954,
      "step": 3614
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.467056732421471e-06,
      "loss": 0.2976,
      "step": 3615
    },
    {
      "epoch": 2.33,
      "learning_rate": 2.4624711455913764e-06,
      "loss": 0.2856,
      "step": 3616
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4578892259389597e-06,
      "loss": 0.2951,
      "step": 3617
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.453310975693417e-06,
      "loss": 0.2989,
      "step": 3618
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.448736397082179e-06,
      "loss": 0.3075,
      "step": 3619
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.444165492330879e-06,
      "loss": 0.3065,
      "step": 3620
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4395982636633632e-06,
      "loss": 0.3112,
      "step": 3621
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.435034713301693e-06,
      "loss": 0.2952,
      "step": 3622
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4304748434661372e-06,
      "loss": 0.2963,
      "step": 3623
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.425918656375177e-06,
      "loss": 0.2883,
      "step": 3624
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4213661542454956e-06,
      "loss": 0.2947,
      "step": 3625
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.416817339291997e-06,
      "loss": 0.2976,
      "step": 3626
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.412272213727772e-06,
      "loss": 0.2961,
      "step": 3627
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4077307797641357e-06,
      "loss": 0.3059,
      "step": 3628
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.4031930396105883e-06,
      "loss": 0.3018,
      "step": 3629
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.3986589954748505e-06,
      "loss": 0.2983,
      "step": 3630
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.3941286495628357e-06,
      "loss": 0.3206,
      "step": 3631
    },
    {
      "epoch": 2.34,
      "learning_rate": 2.389602004078657e-06,
      "loss": 0.307,
      "step": 3632
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.385079061224631e-06,
      "loss": 0.2919,
      "step": 3633
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.380559823201267e-06,
      "loss": 0.2964,
      "step": 3634
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3760442922072867e-06,
      "loss": 0.312,
      "step": 3635
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3715324704395846e-06,
      "loss": 0.3129,
      "step": 3636
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3670243600932763e-06,
      "loss": 0.3012,
      "step": 3637
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.362519963361647e-06,
      "loss": 0.3035,
      "step": 3638
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3580192824361957e-06,
      "loss": 0.3072,
      "step": 3639
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3535223195066025e-06,
      "loss": 0.2791,
      "step": 3640
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3490290767607414e-06,
      "loss": 0.2798,
      "step": 3641
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3445395563846763e-06,
      "loss": 0.2979,
      "step": 3642
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3400537605626592e-06,
      "loss": 0.3126,
      "step": 3643
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.335571691477132e-06,
      "loss": 0.2904,
      "step": 3644
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.33109335130872e-06,
      "loss": 0.3151,
      "step": 3645
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.326618742236242e-06,
      "loss": 0.3038,
      "step": 3646
    },
    {
      "epoch": 2.35,
      "learning_rate": 2.3221478664366882e-06,
      "loss": 0.3024,
      "step": 3647
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.3176807260852475e-06,
      "loss": 0.315,
      "step": 3648
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.313217323355281e-06,
      "loss": 0.3148,
      "step": 3649
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.3087576604183373e-06,
      "loss": 0.3149,
      "step": 3650
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.3043017394441394e-06,
      "loss": 0.3155,
      "step": 3651
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.2998495626005955e-06,
      "loss": 0.3114,
      "step": 3652
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.2954011320537895e-06,
      "loss": 0.305,
      "step": 3653
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.2909564499679825e-06,
      "loss": 0.3056,
      "step": 3654
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.286515518505614e-06,
      "loss": 0.3117,
      "step": 3655
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.282078339827293e-06,
      "loss": 0.3053,
      "step": 3656
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.2776449160918136e-06,
      "loss": 0.2876,
      "step": 3657
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.273215249456132e-06,
      "loss": 0.2967,
      "step": 3658
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.2687893420753824e-06,
      "loss": 0.3044,
      "step": 3659
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.264367196102869e-06,
      "loss": 0.3041,
      "step": 3660
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.259948813690064e-06,
      "loss": 0.3075,
      "step": 3661
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.25553419698661e-06,
      "loss": 0.2938,
      "step": 3662
    },
    {
      "epoch": 2.36,
      "learning_rate": 2.2511233481403173e-06,
      "loss": 0.2923,
      "step": 3663
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2467162692971655e-06,
      "loss": 0.2995,
      "step": 3664
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.242312962601295e-06,
      "loss": 0.2901,
      "step": 3665
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2379134301950133e-06,
      "loss": 0.3197,
      "step": 3666
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.233517674218799e-06,
      "loss": 0.2972,
      "step": 3667
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.229125696811275e-06,
      "loss": 0.3051,
      "step": 3668
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2247375001092475e-06,
      "loss": 0.3149,
      "step": 3669
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2203530862476684e-06,
      "loss": 0.3027,
      "step": 3670
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2159724573596552e-06,
      "loss": 0.305,
      "step": 3671
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2115956155764817e-06,
      "loss": 0.2982,
      "step": 3672
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2072225630275803e-06,
      "loss": 0.3109,
      "step": 3673
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.2028533018405396e-06,
      "loss": 0.3181,
      "step": 3674
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.198487834141102e-06,
      "loss": 0.2949,
      "step": 3675
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.1941261620531718e-06,
      "loss": 0.2852,
      "step": 3676
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.1897682876987925e-06,
      "loss": 0.3021,
      "step": 3677
    },
    {
      "epoch": 2.37,
      "learning_rate": 2.185414213198177e-06,
      "loss": 0.3116,
      "step": 3678
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.18106394066967e-06,
      "loss": 0.2962,
      "step": 3679
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.17671747222979e-06,
      "loss": 0.3161,
      "step": 3680
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1723748099931785e-06,
      "loss": 0.2981,
      "step": 3681
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.168035956072648e-06,
      "loss": 0.2937,
      "step": 3682
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1637009125791466e-06,
      "loss": 0.3018,
      "step": 3683
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1593696816217667e-06,
      "loss": 0.2865,
      "step": 3684
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1550422653077584e-06,
      "loss": 0.3018,
      "step": 3685
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1507186657424962e-06,
      "loss": 0.2958,
      "step": 3686
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1463988850295204e-06,
      "loss": 0.3072,
      "step": 3687
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.142082925270489e-06,
      "loss": 0.3163,
      "step": 3688
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1377707885652264e-06,
      "loss": 0.3122,
      "step": 3689
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.133462477011673e-06,
      "loss": 0.3034,
      "step": 3690
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.129157992705927e-06,
      "loss": 0.2875,
      "step": 3691
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1248573377422155e-06,
      "loss": 0.3007,
      "step": 3692
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1205605142129028e-06,
      "loss": 0.2944,
      "step": 3693
    },
    {
      "epoch": 2.38,
      "learning_rate": 2.1162675242084907e-06,
      "loss": 0.3055,
      "step": 3694
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.1119783698176134e-06,
      "loss": 0.3121,
      "step": 3695
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.107693053127049e-06,
      "loss": 0.3012,
      "step": 3696
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.10341157622169e-06,
      "loss": 0.2891,
      "step": 3697
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.099133941184582e-06,
      "loss": 0.2984,
      "step": 3698
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.09486015009688e-06,
      "loss": 0.2763,
      "step": 3699
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.090590205037888e-06,
      "loss": 0.3141,
      "step": 3700
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.0863241080850273e-06,
      "loss": 0.2985,
      "step": 3701
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.082061861313851e-06,
      "loss": 0.3078,
      "step": 3702
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.0778034667980375e-06,
      "loss": 0.2995,
      "step": 3703
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.0735489266093923e-06,
      "loss": 0.2983,
      "step": 3704
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.069298242817844e-06,
      "loss": 0.287,
      "step": 3705
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.065051417491446e-06,
      "loss": 0.2975,
      "step": 3706
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.06080845269638e-06,
      "loss": 0.3151,
      "step": 3707
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.056569350496933e-06,
      "loss": 0.2975,
      "step": 3708
    },
    {
      "epoch": 2.39,
      "learning_rate": 2.052334112955534e-06,
      "loss": 0.2957,
      "step": 3709
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.048102742132718e-06,
      "loss": 0.2834,
      "step": 3710
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.043875240087142e-06,
      "loss": 0.3016,
      "step": 3711
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0396516088755804e-06,
      "loss": 0.2768,
      "step": 3712
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.035431850552926e-06,
      "loss": 0.3111,
      "step": 3713
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0312159671721854e-06,
      "loss": 0.3154,
      "step": 3714
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0270039607844815e-06,
      "loss": 0.3161,
      "step": 3715
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0227958334390506e-06,
      "loss": 0.2854,
      "step": 3716
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0185915871832417e-06,
      "loss": 0.2854,
      "step": 3717
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.014391224062512e-06,
      "loss": 0.295,
      "step": 3718
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0101947461204375e-06,
      "loss": 0.3204,
      "step": 3719
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0060021553986974e-06,
      "loss": 0.3016,
      "step": 3720
    },
    {
      "epoch": 2.4,
      "learning_rate": 2.0018134539370827e-06,
      "loss": 0.3003,
      "step": 3721
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.9976286437734893e-06,
      "loss": 0.2953,
      "step": 3722
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.9934477269439225e-06,
      "loss": 0.2942,
      "step": 3723
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.989270705482492e-06,
      "loss": 0.3007,
      "step": 3724
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.9850975814214144e-06,
      "loss": 0.3376,
      "step": 3725
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9809283567910063e-06,
      "loss": 0.3006,
      "step": 3726
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9767630336196896e-06,
      "loss": 0.3172,
      "step": 3727
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9726016139339934e-06,
      "loss": 0.3026,
      "step": 3728
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.968444099758532e-06,
      "loss": 0.294,
      "step": 3729
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.964290493116041e-06,
      "loss": 0.2954,
      "step": 3730
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.960140796027333e-06,
      "loss": 0.3125,
      "step": 3731
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.955995010511338e-06,
      "loss": 0.3195,
      "step": 3732
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9518531385850693e-06,
      "loss": 0.2945,
      "step": 3733
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9477151822636418e-06,
      "loss": 0.3083,
      "step": 3734
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.943581143560266e-06,
      "loss": 0.307,
      "step": 3735
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9394510244862397e-06,
      "loss": 0.308,
      "step": 3736
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9353248270509697e-06,
      "loss": 0.2883,
      "step": 3737
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9312025532619306e-06,
      "loss": 0.3043,
      "step": 3738
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.927084205124714e-06,
      "loss": 0.3057,
      "step": 3739
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.9229697846429773e-06,
      "loss": 0.2897,
      "step": 3740
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.9188592938184913e-06,
      "loss": 0.289,
      "step": 3741
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.91475273465109e-06,
      "loss": 0.2858,
      "step": 3742
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.910650109138715e-06,
      "loss": 0.3068,
      "step": 3743
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.9065514192773848e-06,
      "loss": 0.3024,
      "step": 3744
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.9024566670611999e-06,
      "loss": 0.3065,
      "step": 3745
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8983658544823591e-06,
      "loss": 0.3049,
      "step": 3746
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8942789835311237e-06,
      "loss": 0.2984,
      "step": 3747
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8901960561958588e-06,
      "loss": 0.3224,
      "step": 3748
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8861170744629909e-06,
      "loss": 0.3019,
      "step": 3749
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8820420403170459e-06,
      "loss": 0.2905,
      "step": 3750
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.87797095574061e-06,
      "loss": 0.3059,
      "step": 3751
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8739038227143658e-06,
      "loss": 0.2803,
      "step": 3752
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8698406432170613e-06,
      "loss": 0.3137,
      "step": 3753
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8657814192255264e-06,
      "loss": 0.2959,
      "step": 3754
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8617261527146647e-06,
      "loss": 0.3073,
      "step": 3755
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.8576748456574512e-06,
      "loss": 0.2985,
      "step": 3756
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.853627500024946e-06,
      "loss": 0.3057,
      "step": 3757
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.849584117786265e-06,
      "loss": 0.2919,
      "step": 3758
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8455447009086136e-06,
      "loss": 0.3229,
      "step": 3759
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8415092513572498e-06,
      "loss": 0.3053,
      "step": 3760
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8374777710955183e-06,
      "loss": 0.3152,
      "step": 3761
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8334502620848216e-06,
      "loss": 0.293,
      "step": 3762
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8294267262846354e-06,
      "loss": 0.2962,
      "step": 3763
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8254071656524997e-06,
      "loss": 0.2921,
      "step": 3764
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8213915821440197e-06,
      "loss": 0.3032,
      "step": 3765
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8173799777128698e-06,
      "loss": 0.2949,
      "step": 3766
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.813372354310785e-06,
      "loss": 0.3167,
      "step": 3767
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8093687138875648e-06,
      "loss": 0.2931,
      "step": 3768
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8053690583910665e-06,
      "loss": 0.2812,
      "step": 3769
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.8013733897672193e-06,
      "loss": 0.3056,
      "step": 3770
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.7973817099600032e-06,
      "loss": 0.3033,
      "step": 3771
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7933940209114597e-06,
      "loss": 0.2833,
      "step": 3772
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7894103245616901e-06,
      "loss": 0.3078,
      "step": 3773
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7854306228488515e-06,
      "loss": 0.3076,
      "step": 3774
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.78145491770916e-06,
      "loss": 0.2971,
      "step": 3775
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7774832110768847e-06,
      "loss": 0.2963,
      "step": 3776
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.773515504884351e-06,
      "loss": 0.3172,
      "step": 3777
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7695518010619373e-06,
      "loss": 0.3131,
      "step": 3778
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7655921015380706e-06,
      "loss": 0.2731,
      "step": 3779
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7616364082392446e-06,
      "loss": 0.3,
      "step": 3780
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7576847230899808e-06,
      "loss": 0.3098,
      "step": 3781
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7537370480128713e-06,
      "loss": 0.3151,
      "step": 3782
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7497933849285464e-06,
      "loss": 0.2956,
      "step": 3783
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.745853735755687e-06,
      "loss": 0.3162,
      "step": 3784
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7419181024110221e-06,
      "loss": 0.3057,
      "step": 3785
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.7379864868093254e-06,
      "loss": 0.2941,
      "step": 3786
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.734058890863416e-06,
      "loss": 0.3063,
      "step": 3787
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7301353164841562e-06,
      "loss": 0.2784,
      "step": 3788
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.72621576558046e-06,
      "loss": 0.3118,
      "step": 3789
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7223002400592693e-06,
      "loss": 0.3128,
      "step": 3790
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.718388741825583e-06,
      "loss": 0.3051,
      "step": 3791
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7144812727824233e-06,
      "loss": 0.3014,
      "step": 3792
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7105778348308733e-06,
      "loss": 0.2964,
      "step": 3793
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7066784298700334e-06,
      "loss": 0.3132,
      "step": 3794
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.7027830597970596e-06,
      "loss": 0.3108,
      "step": 3795
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.6988917265071337e-06,
      "loss": 0.2956,
      "step": 3796
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.6950044318934744e-06,
      "loss": 0.2819,
      "step": 3797
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.6911211778473468e-06,
      "loss": 0.3098,
      "step": 3798
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.687241966258032e-06,
      "loss": 0.318,
      "step": 3799
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.6833667990128622e-06,
      "loss": 0.2897,
      "step": 3800
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.6794956779971838e-06,
      "loss": 0.2986,
      "step": 3801
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.6756286050943959e-06,
      "loss": 0.295,
      "step": 3802
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6717655821859058e-06,
      "loss": 0.3038,
      "step": 3803
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6679066111511677e-06,
      "loss": 0.3042,
      "step": 3804
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6640516938676566e-06,
      "loss": 0.3135,
      "step": 3805
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6602008322108742e-06,
      "loss": 0.2959,
      "step": 3806
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6563540280543588e-06,
      "loss": 0.3027,
      "step": 3807
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6525112832696576e-06,
      "loss": 0.3033,
      "step": 3808
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6486725997263619e-06,
      "loss": 0.2821,
      "step": 3809
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6448379792920677e-06,
      "loss": 0.2729,
      "step": 3810
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6410074238324148e-06,
      "loss": 0.2964,
      "step": 3811
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6371809352110446e-06,
      "loss": 0.3123,
      "step": 3812
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6333585152896381e-06,
      "loss": 0.3148,
      "step": 3813
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6295401659278843e-06,
      "loss": 0.3007,
      "step": 3814
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6257258889834993e-06,
      "loss": 0.3022,
      "step": 3815
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6219156863122121e-06,
      "loss": 0.293,
      "step": 3816
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.6181095597677732e-06,
      "loss": 0.3117,
      "step": 3817
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.6143075112019492e-06,
      "loss": 0.2846,
      "step": 3818
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.61050954246452e-06,
      "loss": 0.2985,
      "step": 3819
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.6067156554032893e-06,
      "loss": 0.3033,
      "step": 3820
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.60292585186406e-06,
      "loss": 0.2976,
      "step": 3821
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5991401336906642e-06,
      "loss": 0.3006,
      "step": 3822
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5953585027249364e-06,
      "loss": 0.2803,
      "step": 3823
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5915809608067245e-06,
      "loss": 0.2917,
      "step": 3824
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5878075097738898e-06,
      "loss": 0.281,
      "step": 3825
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5840381514622994e-06,
      "loss": 0.3103,
      "step": 3826
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5802728877058314e-06,
      "loss": 0.3078,
      "step": 3827
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5765117203363723e-06,
      "loss": 0.2928,
      "step": 3828
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5727546511838133e-06,
      "loss": 0.3041,
      "step": 3829
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5690016820760556e-06,
      "loss": 0.2938,
      "step": 3830
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5652528148389988e-06,
      "loss": 0.2919,
      "step": 3831
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5615080512965563e-06,
      "loss": 0.3032,
      "step": 3832
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5577673932706384e-06,
      "loss": 0.2944,
      "step": 3833
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.554030842581159e-06,
      "loss": 0.2946,
      "step": 3834
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5502984010460354e-06,
      "loss": 0.2851,
      "step": 3835
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5465700704811825e-06,
      "loss": 0.3123,
      "step": 3836
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5428458527005187e-06,
      "loss": 0.2823,
      "step": 3837
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5391257495159584e-06,
      "loss": 0.2852,
      "step": 3838
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5354097627374176e-06,
      "loss": 0.3168,
      "step": 3839
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5316978941728045e-06,
      "loss": 0.3072,
      "step": 3840
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5279901456280321e-06,
      "loss": 0.2987,
      "step": 3841
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5242865189069978e-06,
      "loss": 0.2695,
      "step": 3842
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5205870158116055e-06,
      "loss": 0.2867,
      "step": 3843
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5168916381417387e-06,
      "loss": 0.3028,
      "step": 3844
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5132003876952895e-06,
      "loss": 0.2965,
      "step": 3845
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5095132662681323e-06,
      "loss": 0.2987,
      "step": 3846
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5058302756541331e-06,
      "loss": 0.2786,
      "step": 3847
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5021514176451514e-06,
      "loss": 0.2956,
      "step": 3848
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4984766940310314e-06,
      "loss": 0.3114,
      "step": 3849
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4948061065996167e-06,
      "loss": 0.3373,
      "step": 3850
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4911396571367208e-06,
      "loss": 0.2798,
      "step": 3851
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.487477347426164e-06,
      "loss": 0.2996,
      "step": 3852
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4838191792497315e-06,
      "loss": 0.2919,
      "step": 3853
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4801651543872165e-06,
      "loss": 0.3009,
      "step": 3854
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4765152746163724e-06,
      "loss": 0.3077,
      "step": 3855
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4728695417129579e-06,
      "loss": 0.3045,
      "step": 3856
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4692279574506996e-06,
      "loss": 0.3154,
      "step": 3857
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4655905236013079e-06,
      "loss": 0.2937,
      "step": 3858
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4619572419344852e-06,
      "loss": 0.2907,
      "step": 3859
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.458328114217894e-06,
      "loss": 0.2988,
      "step": 3860
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4547031422171977e-06,
      "loss": 0.2934,
      "step": 3861
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4510823276960151e-06,
      "loss": 0.286,
      "step": 3862
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4474656724159653e-06,
      "loss": 0.2892,
      "step": 3863
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.4438531781366226e-06,
      "loss": 0.2984,
      "step": 3864
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.440244846615554e-06,
      "loss": 0.3067,
      "step": 3865
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.436640679608292e-06,
      "loss": 0.2776,
      "step": 3866
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4330406788683448e-06,
      "loss": 0.312,
      "step": 3867
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4294448461471943e-06,
      "loss": 0.306,
      "step": 3868
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4258531831942912e-06,
      "loss": 0.3034,
      "step": 3869
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4222656917570675e-06,
      "loss": 0.3068,
      "step": 3870
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4186823735809096e-06,
      "loss": 0.2983,
      "step": 3871
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4151032304091928e-06,
      "loss": 0.2906,
      "step": 3872
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4115282639832395e-06,
      "loss": 0.3213,
      "step": 3873
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.407957476042361e-06,
      "loss": 0.2943,
      "step": 3874
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4043908683238217e-06,
      "loss": 0.2885,
      "step": 3875
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.400828442562857e-06,
      "loss": 0.2889,
      "step": 3876
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.3972702004926675e-06,
      "loss": 0.3024,
      "step": 3877
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.3937161438444191e-06,
      "loss": 0.3004,
      "step": 3878
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.390166274347239e-06,
      "loss": 0.3081,
      "step": 3879
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3866205937282195e-06,
      "loss": 0.3008,
      "step": 3880
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3830791037124147e-06,
      "loss": 0.3019,
      "step": 3881
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3795418060228361e-06,
      "loss": 0.2926,
      "step": 3882
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3760087023804635e-06,
      "loss": 0.3123,
      "step": 3883
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.372479794504229e-06,
      "loss": 0.3043,
      "step": 3884
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3689550841110255e-06,
      "loss": 0.3245,
      "step": 3885
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3654345729157059e-06,
      "loss": 0.2866,
      "step": 3886
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3619182626310768e-06,
      "loss": 0.2944,
      "step": 3887
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3584061549679017e-06,
      "loss": 0.2812,
      "step": 3888
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3548982516349019e-06,
      "loss": 0.291,
      "step": 3889
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3513945543387497e-06,
      "loss": 0.3025,
      "step": 3890
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.347895064784075e-06,
      "loss": 0.2911,
      "step": 3891
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3443997846734535e-06,
      "loss": 0.3198,
      "step": 3892
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3409087157074264e-06,
      "loss": 0.306,
      "step": 3893
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.3374218595844668e-06,
      "loss": 0.324,
      "step": 3894
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.333939218001018e-06,
      "loss": 0.271,
      "step": 3895
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.330460792651459e-06,
      "loss": 0.2816,
      "step": 3896
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3269865852281228e-06,
      "loss": 0.3161,
      "step": 3897
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3235165974212905e-06,
      "loss": 0.3027,
      "step": 3898
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3200508309191883e-06,
      "loss": 0.3172,
      "step": 3899
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.31658928740799e-06,
      "loss": 0.3101,
      "step": 3900
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3131319685718126e-06,
      "loss": 0.3067,
      "step": 3901
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.309678876092727e-06,
      "loss": 0.3112,
      "step": 3902
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3062300116507299e-06,
      "loss": 0.3027,
      "step": 3903
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.3027853769237808e-06,
      "loss": 0.297,
      "step": 3904
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.299344973587765e-06,
      "loss": 0.2983,
      "step": 3905
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.2959088033165212e-06,
      "loss": 0.2974,
      "step": 3906
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.2924768677818222e-06,
      "loss": 0.3029,
      "step": 3907
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.2890491686533812e-06,
      "loss": 0.291,
      "step": 3908
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.285625707598852e-06,
      "loss": 0.2921,
      "step": 3909
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.2822064862838236e-06,
      "loss": 0.2951,
      "step": 3910
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2787915063718293e-06,
      "loss": 0.3039,
      "step": 3911
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2753807695243258e-06,
      "loss": 0.3088,
      "step": 3912
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2719742774007237e-06,
      "loss": 0.3066,
      "step": 3913
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2685720316583471e-06,
      "loss": 0.2845,
      "step": 3914
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2651740339524753e-06,
      "loss": 0.3141,
      "step": 3915
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2617802859363016e-06,
      "loss": 0.2984,
      "step": 3916
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2583907892609681e-06,
      "loss": 0.2982,
      "step": 3917
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2550055455755383e-06,
      "loss": 0.3091,
      "step": 3918
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.251624556527008e-06,
      "loss": 0.3162,
      "step": 3919
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2482478237603102e-06,
      "loss": 0.2893,
      "step": 3920
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.244875348918293e-06,
      "loss": 0.3112,
      "step": 3921
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2415071336417516e-06,
      "loss": 0.3008,
      "step": 3922
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2381431795693888e-06,
      "loss": 0.3149,
      "step": 3923
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2347834883378518e-06,
      "loss": 0.2944,
      "step": 3924
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2314280615816999e-06,
      "loss": 0.299,
      "step": 3925
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.2280769009334282e-06,
      "loss": 0.2788,
      "step": 3926
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2247300080234502e-06,
      "loss": 0.3081,
      "step": 3927
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2213873844801049e-06,
      "loss": 0.2935,
      "step": 3928
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2180490319296546e-06,
      "loss": 0.2986,
      "step": 3929
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.214714951996282e-06,
      "loss": 0.3046,
      "step": 3930
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.211385146302093e-06,
      "loss": 0.2607,
      "step": 3931
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2080596164671098e-06,
      "loss": 0.294,
      "step": 3932
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2047383641092858e-06,
      "loss": 0.2904,
      "step": 3933
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.2014213908444761e-06,
      "loss": 0.2941,
      "step": 3934
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.1981086982864676e-06,
      "loss": 0.2854,
      "step": 3935
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.1948002880469601e-06,
      "loss": 0.2973,
      "step": 3936
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.191496161735568e-06,
      "loss": 0.317,
      "step": 3937
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.1881963209598246e-06,
      "loss": 0.2899,
      "step": 3938
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.1849007673251755e-06,
      "loss": 0.2886,
      "step": 3939
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.1816095024349828e-06,
      "loss": 0.3111,
      "step": 3940
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.1783225278905208e-06,
      "loss": 0.3101,
      "step": 3941
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1750398452909761e-06,
      "loss": 0.2972,
      "step": 3942
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.171761456233449e-06,
      "loss": 0.3035,
      "step": 3943
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1684873623129457e-06,
      "loss": 0.2968,
      "step": 3944
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1652175651223928e-06,
      "loss": 0.3104,
      "step": 3945
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1619520662526174e-06,
      "loss": 0.3097,
      "step": 3946
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1586908672923581e-06,
      "loss": 0.293,
      "step": 3947
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1554339698282623e-06,
      "loss": 0.2962,
      "step": 3948
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.152181375444884e-06,
      "loss": 0.3006,
      "step": 3949
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1489330857246828e-06,
      "loss": 0.2848,
      "step": 3950
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1456891022480254e-06,
      "loss": 0.2981,
      "step": 3951
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1424494265931829e-06,
      "loss": 0.32,
      "step": 3952
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1392140603363288e-06,
      "loss": 0.2883,
      "step": 3953
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1359830050515475e-06,
      "loss": 0.3112,
      "step": 3954
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1327562623108123e-06,
      "loss": 0.3197,
      "step": 3955
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1295338336840113e-06,
      "loss": 0.3165,
      "step": 3956
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.1263157207389274e-06,
      "loss": 0.2945,
      "step": 3957
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1231019250412467e-06,
      "loss": 0.309,
      "step": 3958
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1198924481545503e-06,
      "loss": 0.2888,
      "step": 3959
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1166872916403226e-06,
      "loss": 0.3085,
      "step": 3960
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.113486457057944e-06,
      "loss": 0.3201,
      "step": 3961
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.110289945964692e-06,
      "loss": 0.304,
      "step": 3962
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.107097759915745e-06,
      "loss": 0.3065,
      "step": 3963
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1039099004641684e-06,
      "loss": 0.2948,
      "step": 3964
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.1007263691609326e-06,
      "loss": 0.3059,
      "step": 3965
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.097547167554891e-06,
      "loss": 0.2931,
      "step": 3966
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.0943722971928027e-06,
      "loss": 0.3118,
      "step": 3967
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.0912017596193115e-06,
      "loss": 0.2951,
      "step": 3968
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.0880355563769562e-06,
      "loss": 0.2942,
      "step": 3969
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.084873689006164e-06,
      "loss": 0.3099,
      "step": 3970
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.081716159045255e-06,
      "loss": 0.2983,
      "step": 3971
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.0785629680304433e-06,
      "loss": 0.2976,
      "step": 3972
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0754141174958188e-06,
      "loss": 0.283,
      "step": 3973
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0722696089733787e-06,
      "loss": 0.2871,
      "step": 3974
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0691294439929867e-06,
      "loss": 0.3118,
      "step": 3975
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.065993624082411e-06,
      "loss": 0.3044,
      "step": 3976
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0628621507672964e-06,
      "loss": 0.3143,
      "step": 3977
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0597350255711758e-06,
      "loss": 0.3061,
      "step": 3978
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0566122500154653e-06,
      "loss": 0.3005,
      "step": 3979
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.053493825619467e-06,
      "loss": 0.2914,
      "step": 3980
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0503797539003624e-06,
      "loss": 0.3056,
      "step": 3981
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.047270036373217e-06,
      "loss": 0.293,
      "step": 3982
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0441646745509848e-06,
      "loss": 0.3175,
      "step": 3983
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0410636699444855e-06,
      "loss": 0.2863,
      "step": 3984
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0379670240624361e-06,
      "loss": 0.2818,
      "step": 3985
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0348747384114167e-06,
      "loss": 0.2979,
      "step": 3986
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0317868144958987e-06,
      "loss": 0.3005,
      "step": 3987
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.0287032538182262e-06,
      "loss": 0.3014,
      "step": 3988
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0256240578786214e-06,
      "loss": 0.3046,
      "step": 3989
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0225492281751815e-06,
      "loss": 0.3131,
      "step": 3990
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0194787662038796e-06,
      "loss": 0.2822,
      "step": 3991
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0164126734585667e-06,
      "loss": 0.3094,
      "step": 3992
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0133509514309648e-06,
      "loss": 0.295,
      "step": 3993
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0102936016106702e-06,
      "loss": 0.3178,
      "step": 3994
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0072406254851518e-06,
      "loss": 0.3174,
      "step": 3995
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0041920245397552e-06,
      "loss": 0.2968,
      "step": 3996
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.0011478002576903e-06,
      "loss": 0.2992,
      "step": 3997
    },
    {
      "epoch": 2.58,
      "learning_rate": 9.981079541200412e-07,
      "loss": 0.3044,
      "step": 3998
    },
    {
      "epoch": 2.58,
      "learning_rate": 9.950724876057616e-07,
      "loss": 0.2981,
      "step": 3999
    },
    {
      "epoch": 2.58,
      "learning_rate": 9.920414021916747e-07,
      "loss": 0.305,
      "step": 4000
    },
    {
      "epoch": 2.58,
      "learning_rate": 9.89014699352473e-07,
      "loss": 0.2892,
      "step": 4001
    },
    {
      "epoch": 2.58,
      "learning_rate": 9.85992380560713e-07,
      "loss": 0.3047,
      "step": 4002
    },
    {
      "epoch": 2.58,
      "learning_rate": 9.82974447286823e-07,
      "loss": 0.2815,
      "step": 4003
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.79960900999094e-07,
      "loss": 0.2947,
      "step": 4004
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.769517431636843e-07,
      "loss": 0.2919,
      "step": 4005
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.739469752446197e-07,
      "loss": 0.3058,
      "step": 4006
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.709465987037848e-07,
      "loss": 0.3013,
      "step": 4007
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.67950615000931e-07,
      "loss": 0.3173,
      "step": 4008
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.649590255936735e-07,
      "loss": 0.319,
      "step": 4009
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.61971831937485e-07,
      "loss": 0.2952,
      "step": 4010
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.589890354857056e-07,
      "loss": 0.3228,
      "step": 4011
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.560106376895305e-07,
      "loss": 0.2828,
      "step": 4012
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.530366399980184e-07,
      "loss": 0.2918,
      "step": 4013
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.50067043858085e-07,
      "loss": 0.3166,
      "step": 4014
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.471018507145102e-07,
      "loss": 0.2923,
      "step": 4015
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.441410620099201e-07,
      "loss": 0.285,
      "step": 4016
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.411846791848123e-07,
      "loss": 0.2977,
      "step": 4017
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.382327036775274e-07,
      "loss": 0.2942,
      "step": 4018
    },
    {
      "epoch": 2.59,
      "learning_rate": 9.352851369242733e-07,
      "loss": 0.2861,
      "step": 4019
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.32341980359105e-07,
      "loss": 0.2898,
      "step": 4020
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.294032354139349e-07,
      "loss": 0.3173,
      "step": 4021
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.264689035185282e-07,
      "loss": 0.3086,
      "step": 4022
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.235389861005029e-07,
      "loss": 0.2991,
      "step": 4023
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.206134845853343e-07,
      "loss": 0.2997,
      "step": 4024
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.176924003963383e-07,
      "loss": 0.3159,
      "step": 4025
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.147757349546948e-07,
      "loss": 0.3065,
      "step": 4026
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.118634896794199e-07,
      "loss": 0.3105,
      "step": 4027
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.089556659873921e-07,
      "loss": 0.314,
      "step": 4028
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.060522652933313e-07,
      "loss": 0.2968,
      "step": 4029
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.031532890098083e-07,
      "loss": 0.3014,
      "step": 4030
    },
    {
      "epoch": 2.6,
      "learning_rate": 9.002587385472394e-07,
      "loss": 0.2786,
      "step": 4031
    },
    {
      "epoch": 2.6,
      "learning_rate": 8.973686153138872e-07,
      "loss": 0.3099,
      "step": 4032
    },
    {
      "epoch": 2.6,
      "learning_rate": 8.944829207158678e-07,
      "loss": 0.3075,
      "step": 4033
    },
    {
      "epoch": 2.6,
      "learning_rate": 8.916016561571284e-07,
      "loss": 0.2647,
      "step": 4034
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.887248230394763e-07,
      "loss": 0.2944,
      "step": 4035
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.858524227625498e-07,
      "loss": 0.2932,
      "step": 4036
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.829844567238399e-07,
      "loss": 0.3183,
      "step": 4037
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.801209263186761e-07,
      "loss": 0.2962,
      "step": 4038
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.772618329402305e-07,
      "loss": 0.3081,
      "step": 4039
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.744071779795171e-07,
      "loss": 0.2975,
      "step": 4040
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.715569628253884e-07,
      "loss": 0.2952,
      "step": 4041
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.687111888645383e-07,
      "loss": 0.2823,
      "step": 4042
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.658698574815017e-07,
      "loss": 0.3048,
      "step": 4043
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.630329700586481e-07,
      "loss": 0.3137,
      "step": 4044
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.602005279761871e-07,
      "loss": 0.3,
      "step": 4045
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.573725326121696e-07,
      "loss": 0.29,
      "step": 4046
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.545489853424749e-07,
      "loss": 0.2897,
      "step": 4047
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.517298875408253e-07,
      "loss": 0.3089,
      "step": 4048
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.489152405787726e-07,
      "loss": 0.3168,
      "step": 4049
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.461050458257081e-07,
      "loss": 0.3058,
      "step": 4050
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.432993046488536e-07,
      "loss": 0.3092,
      "step": 4051
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.40498018413266e-07,
      "loss": 0.288,
      "step": 4052
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.377011884818353e-07,
      "loss": 0.2999,
      "step": 4053
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.349088162152796e-07,
      "loss": 0.2884,
      "step": 4054
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.321209029721533e-07,
      "loss": 0.3224,
      "step": 4055
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.293374501088358e-07,
      "loss": 0.3146,
      "step": 4056
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.26558458979545e-07,
      "loss": 0.2939,
      "step": 4057
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.237839309363183e-07,
      "loss": 0.2919,
      "step": 4058
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.210138673290291e-07,
      "loss": 0.2943,
      "step": 4059
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.182482695053728e-07,
      "loss": 0.2956,
      "step": 4060
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.154871388108776e-07,
      "loss": 0.3211,
      "step": 4061
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.127304765888943e-07,
      "loss": 0.3116,
      "step": 4062
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.099782841806014e-07,
      "loss": 0.2995,
      "step": 4063
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.072305629250033e-07,
      "loss": 0.297,
      "step": 4064
    },
    {
      "epoch": 2.62,
      "learning_rate": 8.044873141589282e-07,
      "loss": 0.2937,
      "step": 4065
    },
    {
      "epoch": 2.63,
      "learning_rate": 8.017485392170255e-07,
      "loss": 0.3164,
      "step": 4066
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.990142394317779e-07,
      "loss": 0.3092,
      "step": 4067
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.962844161334748e-07,
      "loss": 0.2951,
      "step": 4068
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.935590706502439e-07,
      "loss": 0.2931,
      "step": 4069
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.908382043080253e-07,
      "loss": 0.2992,
      "step": 4070
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.881218184305806e-07,
      "loss": 0.2924,
      "step": 4071
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.854099143394933e-07,
      "loss": 0.2998,
      "step": 4072
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.82702493354166e-07,
      "loss": 0.2941,
      "step": 4073
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.799995567918206e-07,
      "loss": 0.3017,
      "step": 4074
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.77301105967494e-07,
      "loss": 0.3158,
      "step": 4075
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.746071421940482e-07,
      "loss": 0.3152,
      "step": 4076
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.719176667821515e-07,
      "loss": 0.2991,
      "step": 4077
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.692326810403017e-07,
      "loss": 0.3016,
      "step": 4078
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.66552186274796e-07,
      "loss": 0.3027,
      "step": 4079
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.638761837897612e-07,
      "loss": 0.3065,
      "step": 4080
    },
    {
      "epoch": 2.63,
      "learning_rate": 7.612046748871327e-07,
      "loss": 0.293,
      "step": 4081
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.585376608666583e-07,
      "loss": 0.2876,
      "step": 4082
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.558751430259015e-07,
      "loss": 0.3001,
      "step": 4083
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.532171226602336e-07,
      "loss": 0.2948,
      "step": 4084
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.505636010628492e-07,
      "loss": 0.3149,
      "step": 4085
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.479145795247367e-07,
      "loss": 0.3016,
      "step": 4086
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.45270059334714e-07,
      "loss": 0.3205,
      "step": 4087
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.426300417793919e-07,
      "loss": 0.3044,
      "step": 4088
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.399945281432041e-07,
      "loss": 0.2862,
      "step": 4089
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.37363519708384e-07,
      "loss": 0.2969,
      "step": 4090
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.347370177549784e-07,
      "loss": 0.2996,
      "step": 4091
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.321150235608398e-07,
      "loss": 0.3008,
      "step": 4092
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.294975384016257e-07,
      "loss": 0.2883,
      "step": 4093
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.268845635508037e-07,
      "loss": 0.277,
      "step": 4094
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.242761002796405e-07,
      "loss": 0.3111,
      "step": 4095
    },
    {
      "epoch": 2.64,
      "learning_rate": 7.21672149857221e-07,
      "loss": 0.2861,
      "step": 4096
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.190727135504161e-07,
      "loss": 0.3161,
      "step": 4097
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.164777926239164e-07,
      "loss": 0.3088,
      "step": 4098
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.138873883402064e-07,
      "loss": 0.3102,
      "step": 4099
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.113015019595793e-07,
      "loss": 0.3092,
      "step": 4100
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.08720134740123e-07,
      "loss": 0.2915,
      "step": 4101
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.061432879377339e-07,
      "loss": 0.3189,
      "step": 4102
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.03570962806106e-07,
      "loss": 0.2863,
      "step": 4103
    },
    {
      "epoch": 2.65,
      "learning_rate": 7.010031605967316e-07,
      "loss": 0.311,
      "step": 4104
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.984398825589045e-07,
      "loss": 0.2959,
      "step": 4105
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.958811299397194e-07,
      "loss": 0.2861,
      "step": 4106
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.933269039840628e-07,
      "loss": 0.2923,
      "step": 4107
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.907772059346285e-07,
      "loss": 0.3041,
      "step": 4108
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.882320370319006e-07,
      "loss": 0.2957,
      "step": 4109
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.856913985141589e-07,
      "loss": 0.2973,
      "step": 4110
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.831552916174833e-07,
      "loss": 0.3011,
      "step": 4111
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.806237175757457e-07,
      "loss": 0.2928,
      "step": 4112
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.780966776206144e-07,
      "loss": 0.3014,
      "step": 4113
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.755741729815524e-07,
      "loss": 0.3107,
      "step": 4114
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.730562048858136e-07,
      "loss": 0.3001,
      "step": 4115
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.705427745584469e-07,
      "loss": 0.2944,
      "step": 4116
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.680338832222921e-07,
      "loss": 0.2896,
      "step": 4117
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.655295320979804e-07,
      "loss": 0.2965,
      "step": 4118
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.630297224039383e-07,
      "loss": 0.2982,
      "step": 4119
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.605344553563775e-07,
      "loss": 0.299,
      "step": 4120
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.580437321693023e-07,
      "loss": 0.2919,
      "step": 4121
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.555575540545056e-07,
      "loss": 0.2973,
      "step": 4122
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.530759222215688e-07,
      "loss": 0.3041,
      "step": 4123
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.505988378778616e-07,
      "loss": 0.2866,
      "step": 4124
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.481263022285411e-07,
      "loss": 0.2795,
      "step": 4125
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.456583164765518e-07,
      "loss": 0.2996,
      "step": 4126
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.431948818226241e-07,
      "loss": 0.2957,
      "step": 4127
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.407359994652773e-07,
      "loss": 0.3008,
      "step": 4128
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.382816706008088e-07,
      "loss": 0.3084,
      "step": 4129
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.35831896423309e-07,
      "loss": 0.3145,
      "step": 4130
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.333866781246433e-07,
      "loss": 0.3053,
      "step": 4131
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.309460168944692e-07,
      "loss": 0.307,
      "step": 4132
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.285099139202233e-07,
      "loss": 0.3009,
      "step": 4133
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.260783703871243e-07,
      "loss": 0.2898,
      "step": 4134
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.236513874781724e-07,
      "loss": 0.2878,
      "step": 4135
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.212289663741477e-07,
      "loss": 0.2886,
      "step": 4136
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.188111082536174e-07,
      "loss": 0.2858,
      "step": 4137
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.163978142929183e-07,
      "loss": 0.2858,
      "step": 4138
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.139890856661778e-07,
      "loss": 0.2888,
      "step": 4139
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.11584923545292e-07,
      "loss": 0.3072,
      "step": 4140
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.091853290999417e-07,
      "loss": 0.2835,
      "step": 4141
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.067903034975853e-07,
      "loss": 0.3075,
      "step": 4142
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.043998479034551e-07,
      "loss": 0.3227,
      "step": 4143
    },
    {
      "epoch": 2.68,
      "learning_rate": 6.020139634805622e-07,
      "loss": 0.293,
      "step": 4144
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.996326513896911e-07,
      "loss": 0.3,
      "step": 4145
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.972559127894084e-07,
      "loss": 0.2958,
      "step": 4146
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.948837488360448e-07,
      "loss": 0.3116,
      "step": 4147
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.925161606837182e-07,
      "loss": 0.309,
      "step": 4148
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.901531494843071e-07,
      "loss": 0.3146,
      "step": 4149
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.877947163874753e-07,
      "loss": 0.3155,
      "step": 4150
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.854408625406505e-07,
      "loss": 0.2894,
      "step": 4151
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.830915890890366e-07,
      "loss": 0.3019,
      "step": 4152
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.80746897175607e-07,
      "loss": 0.2973,
      "step": 4153
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.784067879411093e-07,
      "loss": 0.2771,
      "step": 4154
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.760712625240572e-07,
      "loss": 0.3048,
      "step": 4155
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.737403220607374e-07,
      "loss": 0.2993,
      "step": 4156
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.714139676852059e-07,
      "loss": 0.3033,
      "step": 4157
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.690922005292831e-07,
      "loss": 0.3004,
      "step": 4158
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.667750217225654e-07,
      "loss": 0.2954,
      "step": 4159
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.644624323924108e-07,
      "loss": 0.297,
      "step": 4160
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.621544336639451e-07,
      "loss": 0.2959,
      "step": 4161
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.598510266600632e-07,
      "loss": 0.2962,
      "step": 4162
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.575522125014243e-07,
      "loss": 0.2974,
      "step": 4163
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.552579923064516e-07,
      "loss": 0.3026,
      "step": 4164
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.529683671913377e-07,
      "loss": 0.2999,
      "step": 4165
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.506833382700349e-07,
      "loss": 0.3099,
      "step": 4166
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.484029066542618e-07,
      "loss": 0.3045,
      "step": 4167
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.461270734534973e-07,
      "loss": 0.3063,
      "step": 4168
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.438558397749905e-07,
      "loss": 0.2939,
      "step": 4169
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.415892067237449e-07,
      "loss": 0.2968,
      "step": 4170
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.393271754025287e-07,
      "loss": 0.2798,
      "step": 4171
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.370697469118713e-07,
      "loss": 0.3074,
      "step": 4172
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.348169223500621e-07,
      "loss": 0.2905,
      "step": 4173
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.325687028131521e-07,
      "loss": 0.3169,
      "step": 4174
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.303250893949508e-07,
      "loss": 0.2797,
      "step": 4175
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.28086083187025e-07,
      "loss": 0.2929,
      "step": 4176
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.25851685278701e-07,
      "loss": 0.3051,
      "step": 4177
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.236218967570694e-07,
      "loss": 0.2948,
      "step": 4178
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.213967187069657e-07,
      "loss": 0.3026,
      "step": 4179
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.191761522109939e-07,
      "loss": 0.2938,
      "step": 4180
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.169601983495065e-07,
      "loss": 0.2992,
      "step": 4181
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.147488582006166e-07,
      "loss": 0.297,
      "step": 4182
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.125421328401914e-07,
      "loss": 0.306,
      "step": 4183
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.10340023341851e-07,
      "loss": 0.2947,
      "step": 4184
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.081425307769739e-07,
      "loss": 0.3018,
      "step": 4185
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.059496562146859e-07,
      "loss": 0.3036,
      "step": 4186
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.037614007218739e-07,
      "loss": 0.3018,
      "step": 4187
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.015777653631693e-07,
      "loss": 0.2913,
      "step": 4188
    },
    {
      "epoch": 2.7,
      "learning_rate": 4.993987512009657e-07,
      "loss": 0.3099,
      "step": 4189
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.972243592953962e-07,
      "loss": 0.3096,
      "step": 4190
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.950545907043569e-07,
      "loss": 0.2996,
      "step": 4191
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.928894464834843e-07,
      "loss": 0.2978,
      "step": 4192
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.907289276861738e-07,
      "loss": 0.3114,
      "step": 4193
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.885730353635643e-07,
      "loss": 0.2939,
      "step": 4194
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.864217705645468e-07,
      "loss": 0.2947,
      "step": 4195
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.84275134335761e-07,
      "loss": 0.3041,
      "step": 4196
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.821331277215901e-07,
      "loss": 0.3141,
      "step": 4197
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.799957517641751e-07,
      "loss": 0.2986,
      "step": 4198
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.778630075033908e-07,
      "loss": 0.3019,
      "step": 4199
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.757348959768704e-07,
      "loss": 0.3013,
      "step": 4200
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.736114182199836e-07,
      "loss": 0.3013,
      "step": 4201
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.714925752658539e-07,
      "loss": 0.2895,
      "step": 4202
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.6937836814534455e-07,
      "loss": 0.2963,
      "step": 4203
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.67268797887066e-07,
      "loss": 0.2898,
      "step": 4204
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.651638655173707e-07,
      "loss": 0.2984,
      "step": 4205
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.6306357206035603e-07,
      "loss": 0.3055,
      "step": 4206
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.6096791853786237e-07,
      "loss": 0.2976,
      "step": 4207
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.5887690596946975e-07,
      "loss": 0.3033,
      "step": 4208
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.567905353725088e-07,
      "loss": 0.2999,
      "step": 4209
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.5470880776203985e-07,
      "loss": 0.305,
      "step": 4210
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.526317241508738e-07,
      "loss": 0.2857,
      "step": 4211
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.5055928554955665e-07,
      "loss": 0.304,
      "step": 4212
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.484914929663786e-07,
      "loss": 0.3079,
      "step": 4213
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.464283474073661e-07,
      "loss": 0.307,
      "step": 4214
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.4436984987628495e-07,
      "loss": 0.3049,
      "step": 4215
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.4231600137464305e-07,
      "loss": 0.299,
      "step": 4216
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.40266802901681e-07,
      "loss": 0.306,
      "step": 4217
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.3822225545438245e-07,
      "loss": 0.3033,
      "step": 4218
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.3618236002746395e-07,
      "loss": 0.2973,
      "step": 4219
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.341471176133838e-07,
      "loss": 0.3008,
      "step": 4220
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.3211652920232996e-07,
      "loss": 0.3084,
      "step": 4221
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.30090595782231e-07,
      "loss": 0.309,
      "step": 4222
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.2806931833874855e-07,
      "loss": 0.2955,
      "step": 4223
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.2605269785528037e-07,
      "loss": 0.2871,
      "step": 4224
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.240407353129572e-07,
      "loss": 0.2717,
      "step": 4225
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.220334316906438e-07,
      "loss": 0.3007,
      "step": 4226
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.200307879649401e-07,
      "loss": 0.3038,
      "step": 4227
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.1803280511017564e-07,
      "loss": 0.2785,
      "step": 4228
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.16039484098415e-07,
      "loss": 0.316,
      "step": 4229
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.1405082589945466e-07,
      "loss": 0.3051,
      "step": 4230
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.120668314808185e-07,
      "loss": 0.2853,
      "step": 4231
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.100875018077688e-07,
      "loss": 0.2984,
      "step": 4232
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.081128378432919e-07,
      "loss": 0.3003,
      "step": 4233
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.0614284054810606e-07,
      "loss": 0.2923,
      "step": 4234
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.041775108806589e-07,
      "loss": 0.3054,
      "step": 4235
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.0221684979712906e-07,
      "loss": 0.3036,
      "step": 4236
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.0026085825142225e-07,
      "loss": 0.2798,
      "step": 4237
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.983095371951706e-07,
      "loss": 0.3005,
      "step": 4238
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.9636288757773923e-07,
      "loss": 0.2905,
      "step": 4239
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.9442091034621156e-07,
      "loss": 0.2943,
      "step": 4240
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.924836064454107e-07,
      "loss": 0.2937,
      "step": 4241
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.9055097681787056e-07,
      "loss": 0.3002,
      "step": 4242
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.8862302240386674e-07,
      "loss": 0.2886,
      "step": 4243
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.8669974414138553e-07,
      "loss": 0.3116,
      "step": 4244
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.8478114296614965e-07,
      "loss": 0.2893,
      "step": 4245
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.8286721981160125e-07,
      "loss": 0.2982,
      "step": 4246
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.8095797560890547e-07,
      "loss": 0.3013,
      "step": 4247
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.7905341128695484e-07,
      "loss": 0.3037,
      "step": 4248
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.7715352777236037e-07,
      "loss": 0.2999,
      "step": 4249
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.752583259894604e-07,
      "loss": 0.3012,
      "step": 4250
    },
    {
      "epoch": 2.74,
      "learning_rate": 3.733678068603108e-07,
      "loss": 0.2963,
      "step": 4251
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.7148197130469574e-07,
      "loss": 0.2874,
      "step": 4252
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.6960082024011244e-07,
      "loss": 0.2993,
      "step": 4253
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.6772435458178546e-07,
      "loss": 0.2887,
      "step": 4254
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.658525752426578e-07,
      "loss": 0.2916,
      "step": 4255
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.639854831333911e-07,
      "loss": 0.3182,
      "step": 4256
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.6212307916236866e-07,
      "loss": 0.3078,
      "step": 4257
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.6026536423569126e-07,
      "loss": 0.3005,
      "step": 4258
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.5841233925718144e-07,
      "loss": 0.2915,
      "step": 4259
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.5656400512837365e-07,
      "loss": 0.3029,
      "step": 4260
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.547203627485274e-07,
      "loss": 0.3185,
      "step": 4261
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.528814130146141e-07,
      "loss": 0.299,
      "step": 4262
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.5104715682132583e-07,
      "loss": 0.2959,
      "step": 4263
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.4921759506106876e-07,
      "loss": 0.2838,
      "step": 4264
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.4739272862396644e-07,
      "loss": 0.2814,
      "step": 4265
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.455725583978575e-07,
      "loss": 0.3173,
      "step": 4266
    },
    {
      "epoch": 2.75,
      "learning_rate": 3.437570852682959e-07,
      "loss": 0.3082,
      "step": 4267
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.4194631011854827e-07,
      "loss": 0.2863,
      "step": 4268
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.401402338295989e-07,
      "loss": 0.2865,
      "step": 4269
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.3833885728014713e-07,
      "loss": 0.2891,
      "step": 4270
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.3654218134659857e-07,
      "loss": 0.2736,
      "step": 4271
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.347502069030795e-07,
      "loss": 0.3129,
      "step": 4272
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.3296293482142586e-07,
      "loss": 0.2921,
      "step": 4273
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.311803659711854e-07,
      "loss": 0.3161,
      "step": 4274
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.294025012196178e-07,
      "loss": 0.3016,
      "step": 4275
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.2762934143169333e-07,
      "loss": 0.3014,
      "step": 4276
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.2586088747009637e-07,
      "loss": 0.3061,
      "step": 4277
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.2409714019521865e-07,
      "loss": 0.3125,
      "step": 4278
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.2233810046516154e-07,
      "loss": 0.2793,
      "step": 4279
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.205837691357405e-07,
      "loss": 0.3116,
      "step": 4280
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.1883414706047276e-07,
      "loss": 0.2996,
      "step": 4281
    },
    {
      "epoch": 2.76,
      "learning_rate": 3.1708923509059406e-07,
      "loss": 0.299,
      "step": 4282
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.153490340750409e-07,
      "loss": 0.296,
      "step": 4283
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.136135448604594e-07,
      "loss": 0.3085,
      "step": 4284
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.118827682912062e-07,
      "loss": 0.309,
      "step": 4285
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.1015670520934237e-07,
      "loss": 0.303,
      "step": 4286
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.0843535645463606e-07,
      "loss": 0.3137,
      "step": 4287
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.067187228645618e-07,
      "loss": 0.3002,
      "step": 4288
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.050068052743016e-07,
      "loss": 0.2956,
      "step": 4289
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.0329960451674025e-07,
      "loss": 0.2945,
      "step": 4290
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.0159712142247334e-07,
      "loss": 0.3028,
      "step": 4291
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.9989935681979165e-07,
      "loss": 0.3083,
      "step": 4292
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.982063115347011e-07,
      "loss": 0.2998,
      "step": 4293
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.965179863909029e-07,
      "loss": 0.2961,
      "step": 4294
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.9483438220980653e-07,
      "loss": 0.303,
      "step": 4295
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.931554998105235e-07,
      "loss": 0.3155,
      "step": 4296
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.9148134000986815e-07,
      "loss": 0.316,
      "step": 4297
    },
    {
      "epoch": 2.77,
      "learning_rate": 2.8981190362235676e-07,
      "loss": 0.306,
      "step": 4298
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.881471914602063e-07,
      "loss": 0.3057,
      "step": 4299
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.8648720433334e-07,
      "loss": 0.2961,
      "step": 4300
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.8483194304937514e-07,
      "loss": 0.2996,
      "step": 4301
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.831814084136375e-07,
      "loss": 0.313,
      "step": 4302
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.815356012291459e-07,
      "loss": 0.2946,
      "step": 4303
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.798945222966265e-07,
      "loss": 0.2995,
      "step": 4304
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.7825817241449615e-07,
      "loss": 0.2943,
      "step": 4305
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.7662655237887916e-07,
      "loss": 0.2989,
      "step": 4306
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.7499966298359605e-07,
      "loss": 0.3104,
      "step": 4307
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.733775050201626e-07,
      "loss": 0.2853,
      "step": 4308
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.7176007927779855e-07,
      "loss": 0.3068,
      "step": 4309
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.7014738654341453e-07,
      "loss": 0.2891,
      "step": 4310
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.685394276016251e-07,
      "loss": 0.2935,
      "step": 4311
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.6693620323473556e-07,
      "loss": 0.3183,
      "step": 4312
    },
    {
      "epoch": 2.78,
      "learning_rate": 2.6533771422275313e-07,
      "loss": 0.2923,
      "step": 4313
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.6374396134337675e-07,
      "loss": 0.3095,
      "step": 4314
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.6215494537200405e-07,
      "loss": 0.2891,
      "step": 4315
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.605706670817276e-07,
      "loss": 0.3114,
      "step": 4316
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.5899112724333433e-07,
      "loss": 0.2959,
      "step": 4317
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.574163266253049e-07,
      "loss": 0.3064,
      "step": 4318
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.5584626599381545e-07,
      "loss": 0.3072,
      "step": 4319
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.5428094611273825e-07,
      "loss": 0.3038,
      "step": 4320
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.5272036774363406e-07,
      "loss": 0.3049,
      "step": 4321
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.5116453164576114e-07,
      "loss": 0.3139,
      "step": 4322
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.4961343857606626e-07,
      "loss": 0.2964,
      "step": 4323
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.480670892891934e-07,
      "loss": 0.3055,
      "step": 4324
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.465254845374765e-07,
      "loss": 0.3018,
      "step": 4325
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.4498862507094103e-07,
      "loss": 0.3024,
      "step": 4326
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.434565116373011e-07,
      "loss": 0.3133,
      "step": 4327
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.419291449819683e-07,
      "loss": 0.3056,
      "step": 4328
    },
    {
      "epoch": 2.79,
      "learning_rate": 2.404065258480381e-07,
      "loss": 0.2986,
      "step": 4329
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.388886549763003e-07,
      "loss": 0.3129,
      "step": 4330
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.3737553310523187e-07,
      "loss": 0.3041,
      "step": 4331
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.358671609710017e-07,
      "loss": 0.2801,
      "step": 4332
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.3436353930746837e-07,
      "loss": 0.3095,
      "step": 4333
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.3286466884617554e-07,
      "loss": 0.3133,
      "step": 4334
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.313705503163577e-07,
      "loss": 0.2996,
      "step": 4335
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.298811844449389e-07,
      "loss": 0.2975,
      "step": 4336
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.2839657195652732e-07,
      "loss": 0.2854,
      "step": 4337
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.2691671357342072e-07,
      "loss": 0.2853,
      "step": 4338
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.2544161001560428e-07,
      "loss": 0.2866,
      "step": 4339
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.2397126200074837e-07,
      "loss": 0.3001,
      "step": 4340
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.2250567024421076e-07,
      "loss": 0.3006,
      "step": 4341
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.2104483545903443e-07,
      "loss": 0.2879,
      "step": 4342
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.1958875835595084e-07,
      "loss": 0.3,
      "step": 4343
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.1813743964336998e-07,
      "loss": 0.3114,
      "step": 4344
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.166908800273948e-07,
      "loss": 0.3049,
      "step": 4345
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.152490802118079e-07,
      "loss": 0.3168,
      "step": 4346
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.1381204089807705e-07,
      "loss": 0.3067,
      "step": 4347
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.1237976278535522e-07,
      "loss": 0.2909,
      "step": 4348
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.109522465704761e-07,
      "loss": 0.3045,
      "step": 4349
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.0952949294796077e-07,
      "loss": 0.2971,
      "step": 4350
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.081115026100089e-07,
      "loss": 0.2949,
      "step": 4351
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.066982762465075e-07,
      "loss": 0.2906,
      "step": 4352
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.0528981454501996e-07,
      "loss": 0.3032,
      "step": 4353
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.0388611819079696e-07,
      "loss": 0.3048,
      "step": 4354
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.0248718786676558e-07,
      "loss": 0.2891,
      "step": 4355
    },
    {
      "epoch": 2.81,
      "learning_rate": 2.0109302425354139e-07,
      "loss": 0.3064,
      "step": 4356
    },
    {
      "epoch": 2.81,
      "learning_rate": 1.997036280294118e-07,
      "loss": 0.3008,
      "step": 4357
    },
    {
      "epoch": 2.81,
      "learning_rate": 1.9831899987035276e-07,
      "loss": 0.3013,
      "step": 4358
    },
    {
      "epoch": 2.81,
      "learning_rate": 1.9693914045001428e-07,
      "loss": 0.3024,
      "step": 4359
    },
    {
      "epoch": 2.81,
      "learning_rate": 1.9556405043973158e-07,
      "loss": 0.3067,
      "step": 4360
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.941937305085162e-07,
      "loss": 0.2802,
      "step": 4361
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.9282818132305813e-07,
      "loss": 0.2802,
      "step": 4362
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.9146740354773042e-07,
      "loss": 0.2896,
      "step": 4363
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.901113978445801e-07,
      "loss": 0.2935,
      "step": 4364
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.8876016487333614e-07,
      "loss": 0.294,
      "step": 4365
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.8741370529140046e-07,
      "loss": 0.2865,
      "step": 4366
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.8607201975386013e-07,
      "loss": 0.3005,
      "step": 4367
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.8473510891347412e-07,
      "loss": 0.2965,
      "step": 4368
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.834029734206777e-07,
      "loss": 0.3018,
      "step": 4369
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.8207561392358796e-07,
      "loss": 0.2905,
      "step": 4370
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.807530310679917e-07,
      "loss": 0.2886,
      "step": 4371
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.794352254973597e-07,
      "loss": 0.2915,
      "step": 4372
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.7812219785282913e-07,
      "loss": 0.2797,
      "step": 4373
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.7681394877322234e-07,
      "loss": 0.2963,
      "step": 4374
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.7551047889502793e-07,
      "loss": 0.3098,
      "step": 4375
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.742117888524153e-07,
      "loss": 0.3025,
      "step": 4376
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.7291787927722681e-07,
      "loss": 0.284,
      "step": 4377
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.7162875079897778e-07,
      "loss": 0.3017,
      "step": 4378
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.7034440404485764e-07,
      "loss": 0.3071,
      "step": 4379
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6906483963973207e-07,
      "loss": 0.3014,
      "step": 4380
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6779005820613647e-07,
      "loss": 0.3099,
      "step": 4381
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6652006036428025e-07,
      "loss": 0.2914,
      "step": 4382
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6525484673204807e-07,
      "loss": 0.3263,
      "step": 4383
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6399441792499305e-07,
      "loss": 0.2979,
      "step": 4384
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6273877455634245e-07,
      "loss": 0.3044,
      "step": 4385
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6148791723699652e-07,
      "loss": 0.2965,
      "step": 4386
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6024184657552398e-07,
      "loss": 0.2921,
      "step": 4387
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.590005631781666e-07,
      "loss": 0.2919,
      "step": 4388
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.5776406764883678e-07,
      "loss": 0.2826,
      "step": 4389
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.565323605891178e-07,
      "loss": 0.2985,
      "step": 4390
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.5530544259826254e-07,
      "loss": 0.3005,
      "step": 4391
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.5408331427319345e-07,
      "loss": 0.2925,
      "step": 4392
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.5286597620850606e-07,
      "loss": 0.3141,
      "step": 4393
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.5165342899645997e-07,
      "loss": 0.3006,
      "step": 4394
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.5044567322698768e-07,
      "loss": 0.3141,
      "step": 4395
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.4924270948769027e-07,
      "loss": 0.2971,
      "step": 4396
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.4804453836383736e-07,
      "loss": 0.3082,
      "step": 4397
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.468511604383638e-07,
      "loss": 0.3046,
      "step": 4398
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.456625762918762e-07,
      "loss": 0.3287,
      "step": 4399
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.4447878650264867e-07,
      "loss": 0.332,
      "step": 4400
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.4329979164661945e-07,
      "loss": 0.3079,
      "step": 4401
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.421255922973974e-07,
      "loss": 0.3143,
      "step": 4402
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.4095618902625562e-07,
      "loss": 0.302,
      "step": 4403
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.3979158240213787e-07,
      "loss": 0.3018,
      "step": 4404
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.3863177299164864e-07,
      "loss": 0.2903,
      "step": 4405
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.374767613590644e-07,
      "loss": 0.2941,
      "step": 4406
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.3632654806632006e-07,
      "loss": 0.3139,
      "step": 4407
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.3518113367302356e-07,
      "loss": 0.2961,
      "step": 4408
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.3404051873644464e-07,
      "loss": 0.2862,
      "step": 4409
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.3290470381151722e-07,
      "loss": 0.3025,
      "step": 4410
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.3177368945084145e-07,
      "loss": 0.3007,
      "step": 4411
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.3064747620468053e-07,
      "loss": 0.2875,
      "step": 4412
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.2952606462096394e-07,
      "loss": 0.2918,
      "step": 4413
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.2840945524528193e-07,
      "loss": 0.291,
      "step": 4414
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.2729764862089212e-07,
      "loss": 0.2957,
      "step": 4415
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.2619064528871185e-07,
      "loss": 0.3072,
      "step": 4416
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.250884457873247e-07,
      "loss": 0.2874,
      "step": 4417
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.239910506529729e-07,
      "loss": 0.3008,
      "step": 4418
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.2289846041956599e-07,
      "loss": 0.3123,
      "step": 4419
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.218106756186743e-07,
      "loss": 0.2929,
      "step": 4420
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.207276967795279e-07,
      "loss": 0.3209,
      "step": 4421
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.1964952442902188e-07,
      "loss": 0.2987,
      "step": 4422
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1857615909171005e-07,
      "loss": 0.3066,
      "step": 4423
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1750760128981131e-07,
      "loss": 0.3161,
      "step": 4424
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1644385154319982e-07,
      "loss": 0.2784,
      "step": 4425
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1538491036941602e-07,
      "loss": 0.2964,
      "step": 4426
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1433077828365779e-07,
      "loss": 0.3017,
      "step": 4427
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.132814557987838e-07,
      "loss": 0.2915,
      "step": 4428
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1223694342531455e-07,
      "loss": 0.3055,
      "step": 4429
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1119724167142687e-07,
      "loss": 0.3179,
      "step": 4430
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.1016235104296169e-07,
      "loss": 0.3062,
      "step": 4431
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.0913227204341292e-07,
      "loss": 0.297,
      "step": 4432
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.081070051739419e-07,
      "loss": 0.2931,
      "step": 4433
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.0708655093335963e-07,
      "loss": 0.2784,
      "step": 4434
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.0607090981814339e-07,
      "loss": 0.2971,
      "step": 4435
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.0506008232242348e-07,
      "loss": 0.2979,
      "step": 4436
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.0405406893799097e-07,
      "loss": 0.2964,
      "step": 4437
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.0305287015429433e-07,
      "loss": 0.2853,
      "step": 4438
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.0205648645843836e-07,
      "loss": 0.2873,
      "step": 4439
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.010649183351875e-07,
      "loss": 0.2933,
      "step": 4440
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.0007816626696032e-07,
      "loss": 0.3167,
      "step": 4441
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.909623073383612e-08,
      "loss": 0.2896,
      "step": 4442
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.81191122135472e-08,
      "loss": 0.3029,
      "step": 4443
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.714681118148329e-08,
      "loss": 0.3041,
      "step": 4444
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.617932811069153e-08,
      "loss": 0.292,
      "step": 4445
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.52166634718743e-08,
      "loss": 0.3107,
      "step": 4446
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.425881773339029e-08,
      "loss": 0.3111,
      "step": 4447
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.330579136125117e-08,
      "loss": 0.2941,
      "step": 4448
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.235758481912715e-08,
      "loss": 0.2772,
      "step": 4449
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.141419856834255e-08,
      "loss": 0.3016,
      "step": 4450
    },
    {
      "epoch": 2.87,
      "learning_rate": 9.047563306787466e-08,
      "loss": 0.2948,
      "step": 4451
    },
    {
      "epoch": 2.87,
      "learning_rate": 8.95418887743571e-08,
      "loss": 0.3023,
      "step": 4452
    },
    {
      "epoch": 2.87,
      "learning_rate": 8.861296614207871e-08,
      "loss": 0.3044,
      "step": 4453
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.768886562298018e-08,
      "loss": 0.294,
      "step": 4454
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.676958766665633e-08,
      "loss": 0.3054,
      "step": 4455
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.58551327203594e-08,
      "loss": 0.2869,
      "step": 4456
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.494550122898903e-08,
      "loss": 0.3104,
      "step": 4457
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.404069363510348e-08,
      "loss": 0.3115,
      "step": 4458
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.314071037891169e-08,
      "loss": 0.2963,
      "step": 4459
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.224555189827565e-08,
      "loss": 0.2967,
      "step": 4460
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.135521862871144e-08,
      "loss": 0.3008,
      "step": 4461
    },
    {
      "epoch": 2.88,
      "learning_rate": 8.046971100338363e-08,
      "loss": 0.3162,
      "step": 4462
    },
    {
      "epoch": 2.88,
      "learning_rate": 7.958902945311319e-08,
      "loss": 0.2936,
      "step": 4463
    },
    {
      "epoch": 2.88,
      "learning_rate": 7.871317440637072e-08,
      "loss": 0.2969,
      "step": 4464
    },
    {
      "epoch": 2.88,
      "learning_rate": 7.784214628928199e-08,
      "loss": 0.289,
      "step": 4465
    },
    {
      "epoch": 2.88,
      "learning_rate": 7.697594552561805e-08,
      "loss": 0.2899,
      "step": 4466
    },
    {
      "epoch": 2.88,
      "learning_rate": 7.611457253680843e-08,
      "loss": 0.2929,
      "step": 4467
    },
    {
      "epoch": 2.88,
      "learning_rate": 7.525802774192791e-08,
      "loss": 0.302,
      "step": 4468
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.44063115577065e-08,
      "loss": 0.2988,
      "step": 4469
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.355942439852271e-08,
      "loss": 0.2958,
      "step": 4470
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.271736667640694e-08,
      "loss": 0.288,
      "step": 4471
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.188013880103817e-08,
      "loss": 0.2978,
      "step": 4472
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.104774117974611e-08,
      "loss": 0.2877,
      "step": 4473
    },
    {
      "epoch": 2.89,
      "learning_rate": 7.022017421751348e-08,
      "loss": 0.32,
      "step": 4474
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.93974383169671e-08,
      "loss": 0.3205,
      "step": 4475
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.857953387839012e-08,
      "loss": 0.3091,
      "step": 4476
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.776646129970754e-08,
      "loss": 0.2955,
      "step": 4477
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.695822097650073e-08,
      "loss": 0.2925,
      "step": 4478
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.615481330199514e-08,
      "loss": 0.2915,
      "step": 4479
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.535623866706698e-08,
      "loss": 0.2925,
      "step": 4480
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.456249746024101e-08,
      "loss": 0.313,
      "step": 4481
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.377359006769057e-08,
      "loss": 0.2945,
      "step": 4482
    },
    {
      "epoch": 2.89,
      "learning_rate": 6.29895168732364e-08,
      "loss": 0.3075,
      "step": 4483
    },
    {
      "epoch": 2.9,
      "learning_rate": 6.22102782583478e-08,
      "loss": 0.3127,
      "step": 4484
    },
    {
      "epoch": 2.9,
      "learning_rate": 6.143587460214151e-08,
      "loss": 0.3005,
      "step": 4485
    },
    {
      "epoch": 2.9,
      "learning_rate": 6.066630628138282e-08,
      "loss": 0.277,
      "step": 4486
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.990157367048554e-08,
      "loss": 0.2888,
      "step": 4487
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.9141677141506536e-08,
      "loss": 0.2997,
      "step": 4488
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.838661706415338e-08,
      "loss": 0.3221,
      "step": 4489
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.763639380578112e-08,
      "loss": 0.2951,
      "step": 4490
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.689100773139e-08,
      "loss": 0.3105,
      "step": 4491
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.61504592036255e-08,
      "loss": 0.3079,
      "step": 4492
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.5414748582783836e-08,
      "loss": 0.3082,
      "step": 4493
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.468387622680205e-08,
      "loss": 0.3004,
      "step": 4494
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.3957842491267895e-08,
      "loss": 0.306,
      "step": 4495
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.323664772941217e-08,
      "loss": 0.2896,
      "step": 4496
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.2520292292113086e-08,
      "loss": 0.2835,
      "step": 4497
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.1808776527892955e-08,
      "loss": 0.3021,
      "step": 4498
    },
    {
      "epoch": 2.9,
      "learning_rate": 5.1102100782920437e-08,
      "loss": 0.3068,
      "step": 4499
    },
    {
      "epoch": 2.91,
      "learning_rate": 5.0400265401009395e-08,
      "loss": 0.2974,
      "step": 4500
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.9703270723618915e-08,
      "loss": 0.3026,
      "step": 4501
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.901111708985329e-08,
      "loss": 0.2951,
      "step": 4502
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.832380483645871e-08,
      "loss": 0.2998,
      "step": 4503
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.76413342978288e-08,
      "loss": 0.2932,
      "step": 4504
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.6963705806002405e-08,
      "loss": 0.2933,
      "step": 4505
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.629091969065913e-08,
      "loss": 0.2996,
      "step": 4506
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.562297627912493e-08,
      "loss": 0.3225,
      "step": 4507
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.4959875896370965e-08,
      "loss": 0.2992,
      "step": 4508
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.430161886500806e-08,
      "loss": 0.295,
      "step": 4509
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.364820550529558e-08,
      "loss": 0.2888,
      "step": 4510
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.2999636135132583e-08,
      "loss": 0.3104,
      "step": 4511
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.2355911070062205e-08,
      "loss": 0.2878,
      "step": 4512
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.17170306232717e-08,
      "loss": 0.2824,
      "step": 4513
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.108299510559022e-08,
      "loss": 0.3008,
      "step": 4514
    },
    {
      "epoch": 2.92,
      "learning_rate": 4.04538048254921e-08,
      "loss": 0.2899,
      "step": 4515
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.982946008909139e-08,
      "loss": 0.2858,
      "step": 4516
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.920996120014509e-08,
      "loss": 0.2956,
      "step": 4517
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.8595308460054326e-08,
      "loss": 0.3069,
      "step": 4518
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.798550216786212e-08,
      "loss": 0.2932,
      "step": 4519
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.738054262025226e-08,
      "loss": 0.291,
      "step": 4520
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.6780430111550415e-08,
      "loss": 0.2915,
      "step": 4521
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.618516493372637e-08,
      "loss": 0.3108,
      "step": 4522
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.559474737638846e-08,
      "loss": 0.2873,
      "step": 4523
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.500917772679025e-08,
      "loss": 0.2998,
      "step": 4524
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.4428456269821606e-08,
      "loss": 0.3022,
      "step": 4525
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.385258328801988e-08,
      "loss": 0.3136,
      "step": 4526
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.328155906155761e-08,
      "loss": 0.2879,
      "step": 4527
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.271538386825257e-08,
      "loss": 0.3163,
      "step": 4528
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.215405798356108e-08,
      "loss": 0.3093,
      "step": 4529
    },
    {
      "epoch": 2.92,
      "learning_rate": 3.1597581680581356e-08,
      "loss": 0.2969,
      "step": 4530
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.104595523005016e-08,
      "loss": 0.2944,
      "step": 4531
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.049917890034837e-08,
      "loss": 0.284,
      "step": 4532
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.9957252957493186e-08,
      "loss": 0.2875,
      "step": 4533
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.9420177665144823e-08,
      "loss": 0.2936,
      "step": 4534
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.888795328460314e-08,
      "loss": 0.3011,
      "step": 4535
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.8360580074804355e-08,
      "loss": 0.2988,
      "step": 4536
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.7838058292330995e-08,
      "loss": 0.2996,
      "step": 4537
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.7320388191399704e-08,
      "loss": 0.3143,
      "step": 4538
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.6807570023869024e-08,
      "loss": 0.2994,
      "step": 4539
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.6299604039237147e-08,
      "loss": 0.2827,
      "step": 4540
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.5796490484639723e-08,
      "loss": 0.2907,
      "step": 4541
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.5298229604854286e-08,
      "loss": 0.3023,
      "step": 4542
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.4804821642294696e-08,
      "loss": 0.3077,
      "step": 4543
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.43162668370156e-08,
      "loss": 0.3281,
      "step": 4544
    },
    {
      "epoch": 2.93,
      "learning_rate": 2.3832565426709086e-08,
      "loss": 0.304,
      "step": 4545
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.335371764670913e-08,
      "loss": 0.2929,
      "step": 4546
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.2879723729983815e-08,
      "loss": 0.3033,
      "step": 4547
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.2410583907142012e-08,
      "loss": 0.2927,
      "step": 4548
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.1946298406432255e-08,
      "loss": 0.297,
      "step": 4549
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.148686745373829e-08,
      "loss": 0.3102,
      "step": 4550
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.1032291272584658e-08,
      "loss": 0.3118,
      "step": 4551
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.0582570084132224e-08,
      "loss": 0.2975,
      "step": 4552
    },
    {
      "epoch": 2.94,
      "learning_rate": 2.013770410718152e-08,
      "loss": 0.2897,
      "step": 4553
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.9697693558168305e-08,
      "loss": 0.2994,
      "step": 4554
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.9262538651169115e-08,
      "loss": 0.2855,
      "step": 4555
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.883223959789571e-08,
      "loss": 0.293,
      "step": 4556
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.8406796607698397e-08,
      "loss": 0.291,
      "step": 4557
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.7986209887564943e-08,
      "loss": 0.3181,
      "step": 4558
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.757047964212055e-08,
      "loss": 0.3023,
      "step": 4559
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.7159606073627875e-08,
      "loss": 0.2844,
      "step": 4560
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.6753589381983682e-08,
      "loss": 0.3088,
      "step": 4561
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.635242976472773e-08,
      "loss": 0.293,
      "step": 4562
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.5956127417030565e-08,
      "loss": 0.2972,
      "step": 4563
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.5564682531702402e-08,
      "loss": 0.3053,
      "step": 4564
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.5178095299192008e-08,
      "loss": 0.2929,
      "step": 4565
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.4796365907580046e-08,
      "loss": 0.288,
      "step": 4566
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.4419494542589063e-08,
      "loss": 0.2975,
      "step": 4567
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.4047481387573503e-08,
      "loss": 0.3034,
      "step": 4568
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.3680326623526364e-08,
      "loss": 0.3051,
      "step": 4569
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.3318030429078089e-08,
      "loss": 0.3001,
      "step": 4570
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.2960592980493237e-08,
      "loss": 0.29,
      "step": 4571
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.2608014451672701e-08,
      "loss": 0.2967,
      "step": 4572
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.2260295014154822e-08,
      "loss": 0.313,
      "step": 4573
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.1917434837112052e-08,
      "loss": 0.2893,
      "step": 4574
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.1579434087354291e-08,
      "loss": 0.305,
      "step": 4575
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.1246292929325552e-08,
      "loss": 0.2999,
      "step": 4576
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.0918011525107297e-08,
      "loss": 0.308,
      "step": 4577
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.0594590034415097e-08,
      "loss": 0.2825,
      "step": 4578
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.0276028614601974e-08,
      "loss": 0.2825,
      "step": 4579
    },
    {
      "epoch": 2.96,
      "learning_rate": 9.96232742065506e-09,
      "loss": 0.3115,
      "step": 4580
    },
    {
      "epoch": 2.96,
      "learning_rate": 9.653486605195606e-09,
      "loss": 0.3016,
      "step": 4581
    },
    {
      "epoch": 2.96,
      "learning_rate": 9.349506318483414e-09,
      "loss": 0.2913,
      "step": 4582
    },
    {
      "epoch": 2.96,
      "learning_rate": 9.050386708410186e-09,
      "loss": 0.2935,
      "step": 4583
    },
    {
      "epoch": 2.96,
      "learning_rate": 8.756127920505065e-09,
      "loss": 0.3099,
      "step": 4584
    },
    {
      "epoch": 2.96,
      "learning_rate": 8.466730097932418e-09,
      "loss": 0.2966,
      "step": 4585
    },
    {
      "epoch": 2.96,
      "learning_rate": 8.182193381489622e-09,
      "loss": 0.2877,
      "step": 4586
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.902517909611496e-09,
      "loss": 0.2963,
      "step": 4587
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.627703818363642e-09,
      "loss": 0.3023,
      "step": 4588
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.357751241452438e-09,
      "loss": 0.2918,
      "step": 4589
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.092660310215049e-09,
      "loss": 0.3044,
      "step": 4590
    },
    {
      "epoch": 2.96,
      "learning_rate": 6.832431153623864e-09,
      "loss": 0.3092,
      "step": 4591
    },
    {
      "epoch": 2.96,
      "learning_rate": 6.577063898285385e-09,
      "loss": 0.3025,
      "step": 4592
    },
    {
      "epoch": 2.97,
      "learning_rate": 6.326558668442451e-09,
      "loss": 0.3097,
      "step": 4593
    },
    {
      "epoch": 2.97,
      "learning_rate": 6.080915585972014e-09,
      "loss": 0.308,
      "step": 4594
    },
    {
      "epoch": 2.97,
      "learning_rate": 5.840134770384032e-09,
      "loss": 0.2897,
      "step": 4595
    },
    {
      "epoch": 2.97,
      "learning_rate": 5.604216338824797e-09,
      "loss": 0.3153,
      "step": 4596
    },
    {
      "epoch": 2.97,
      "learning_rate": 5.373160406072498e-09,
      "loss": 0.3044,
      "step": 4597
    },
    {
      "epoch": 2.97,
      "learning_rate": 5.1469670845416545e-09,
      "loss": 0.2944,
      "step": 4598
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.9256364842809045e-09,
      "loss": 0.2942,
      "step": 4599
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.7091687129718896e-09,
      "loss": 0.296,
      "step": 4600
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.497563875932587e-09,
      "loss": 0.3079,
      "step": 4601
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.290822076111756e-09,
      "loss": 0.2971,
      "step": 4602
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.088943414094493e-09,
      "loss": 0.2967,
      "step": 4603
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.891927988098898e-09,
      "loss": 0.283,
      "step": 4604
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.699775893978297e-09,
      "loss": 0.3107,
      "step": 4605
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.512487225219019e-09,
      "loss": 0.2951,
      "step": 4606
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.3300620729403987e-09,
      "loss": 0.2947,
      "step": 4607
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.1525005258969953e-09,
      "loss": 0.3087,
      "step": 4608
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.9798026704752625e-09,
      "loss": 0.2948,
      "step": 4609
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.811968590699099e-09,
      "loss": 0.2985,
      "step": 4610
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.6489983682220777e-09,
      "loss": 0.3115,
      "step": 4611
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.490892082331886e-09,
      "loss": 0.2872,
      "step": 4612
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.337649809952547e-09,
      "loss": 0.3042,
      "step": 4613
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.1892716256388667e-09,
      "loss": 0.3027,
      "step": 4614
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.0457576015808776e-09,
      "loss": 0.2753,
      "step": 4615
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.907107807600506e-09,
      "loss": 0.307,
      "step": 4616
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.7733223111549013e-09,
      "loss": 0.2737,
      "step": 4617
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.6444011773331103e-09,
      "loss": 0.2841,
      "step": 4618
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.5203444688582924e-09,
      "loss": 0.2995,
      "step": 4619
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.4011522460866122e-09,
      "loss": 0.2846,
      "step": 4620
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.2868245670083491e-09,
      "loss": 0.2838,
      "step": 4621
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.1773614872467864e-09,
      "loss": 0.2742,
      "step": 4622
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.0727630600571027e-09,
      "loss": 0.2911,
      "step": 4623
    },
    {
      "epoch": 2.99,
      "learning_rate": 9.730293363297006e-10,
      "loss": 0.3095,
      "step": 4624
    },
    {
      "epoch": 2.99,
      "learning_rate": 8.781603645857673e-10,
      "loss": 0.3049,
      "step": 4625
    },
    {
      "epoch": 2.99,
      "learning_rate": 7.881561909828251e-10,
      "loss": 0.2957,
      "step": 4626
    },
    {
      "epoch": 2.99,
      "learning_rate": 7.030168593102904e-10,
      "loss": 0.2931,
      "step": 4627
    },
    {
      "epoch": 2.99,
      "learning_rate": 6.227424109883639e-10,
      "loss": 0.3134,
      "step": 4628
    },
    {
      "epoch": 2.99,
      "learning_rate": 5.473328850735815e-10,
      "loss": 0.3078,
      "step": 4629
    },
    {
      "epoch": 2.99,
      "learning_rate": 4.767883182554834e-10,
      "loss": 0.3003,
      "step": 4630
    },
    {
      "epoch": 2.99,
      "learning_rate": 4.111087448532836e-10,
      "loss": 0.2714,
      "step": 4631
    },
    {
      "epoch": 2.99,
      "learning_rate": 3.502941968225315e-10,
      "loss": 0.2966,
      "step": 4632
    },
    {
      "epoch": 2.99,
      "learning_rate": 2.9434470375178105e-10,
      "loss": 0.3043,
      "step": 4633
    },
    {
      "epoch": 2.99,
      "learning_rate": 2.4326029286037003e-10,
      "loss": 0.296,
      "step": 4634
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.970409890028613e-10,
      "loss": 0.3266,
      "step": 4635
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.5568681466682223e-10,
      "loss": 0.2755,
      "step": 4636
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.1919778996949406e-10,
      "loss": 0.3062,
      "step": 4637
    },
    {
      "epoch": 2.99,
      "learning_rate": 8.757393266556336e-11,
      "loss": 0.292,
      "step": 4638
    },
    {
      "epoch": 3.0,
      "learning_rate": 6.081525814050082e-11,
      "loss": 0.3055,
      "step": 4639
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.8921779411671414e-11,
      "loss": 0.3234,
      "step": 4640
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.189350713277527e-11,
      "loss": 0.2912,
      "step": 4641
    },
    {
      "epoch": 3.0,
      "learning_rate": 9.730449587186385e-12,
      "loss": 0.2979,
      "step": 4642
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.432612692393477e-12,
      "loss": 0.2934,
      "step": 4643
    },
    {
      "epoch": 3.0,
      "learning_rate": 0.0,
      "loss": 0.3003,
      "step": 4644
    },
    {
      "epoch": 3.0,
      "step": 4644,
      "total_flos": 2.3174284043616256e+16,
      "train_loss": 0.4229431632299756,
      "train_runtime": 125794.9271,
      "train_samples_per_second": 4.728,
      "train_steps_per_second": 0.037
    }
  ],
  "logging_steps": 1.0,
  "max_steps": 4644,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 10000,
  "total_flos": 2.3174284043616256e+16,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}